2023-05-15 15:28:40,538 INFO [finetune.py:1062] (0/2) Training started 2023-05-15 15:28:40,542 INFO [finetune.py:1072] (0/2) Device: cuda:0 2023-05-15 15:28:40,545 INFO [finetune.py:1081] (0/2) {'frame_shift_ms': 10.0, 'allowed_excess_duration_ratio': 0.1, '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, 'warm_step': 2000, 'env_info': {'k2-version': '1.23.4', 'k2-build-type': 'Release', 'k2-with-cuda': True, 'k2-git-sha1': 'a23383c5a381713b51e9014f3f05d096f8aceec3', 'k2-git-date': 'Wed Apr 26 15:33:33 2023', 'lhotse-version': '1.14.0.dev+git.b61b917.dirty', 'torch-version': '1.13.1', 'torch-cuda-available': True, 'torch-cuda-version': '11.6', 'python-version': '3.1', 'icefall-git-branch': 'master', 'icefall-git-sha1': '45c13e9-dirty', 'icefall-git-date': 'Mon Apr 24 15:00:02 2023', 'icefall-path': '/k2-dev/yangyifan/icefall-master', 'k2-path': '/k2-dev/yangyifan/anaconda3/envs/icefall/lib/python3.10/site-packages/k2-1.23.4.dev20230427+cuda11.6.torch1.13.1-py3.10-linux-x86_64.egg/k2/__init__.py', 'lhotse-path': '/k2-dev/yangyifan/anaconda3/envs/icefall/lib/python3.10/site-packages/lhotse-1.14.0.dev0+git.b61b917.dirty-py3.10.egg/lhotse/__init__.py', 'hostname': 'de-74279-k2-train-7-1218101249-5d97868c7c-v8ngc', 'IP address': '10.177.77.18'}, 'world_size': 2, 'master_port': 12354, 'tensorboard': True, 'num_epochs': 20, 'start_epoch': 1, 'start_batch': 0, 'exp_dir': PosixPath('pruned_transducer_stateless7/exp_giga_finetune'), 'bpe_model': 'icefall-asr-librispeech-pruned-transducer-stateless7-2022-11-11/data/lang_bpe_500/bpe.model', 'base_lr': 0.005, 'lr_batches': 100000.0, 'lr_epochs': 100.0, 'context_size': 2, 'prune_range': 5, 'lm_scale': 0.25, 'am_scale': 0.0, 'simple_loss_scale': 0.5, 'seed': 42, 'print_diagnostics': False, 'inf_check': False, 'save_every_n': 2000, 'keep_last_k': 30, 'average_period': 200, 'use_fp16': True, 'num_encoder_layers': '2,4,3,2,4', 'feedforward_dims': '1024,1024,2048,2048,1024', 'nhead': '8,8,8,8,8', 'encoder_dims': '384,384,384,384,384', 'attention_dims': '192,192,192,192,192', 'encoder_unmasked_dims': '256,256,256,256,256', 'zipformer_downsampling_factors': '1,2,4,8,2', 'cnn_module_kernels': '31,31,31,31,31', 'decoder_dim': 512, 'joiner_dim': 512, 'do_finetune': True, 'use_mux': True, 'init_modules': None, 'finetune_ckpt': 'icefall-asr-librispeech-pruned-transducer-stateless7-2022-11-11/exp/pretrained.pt', 'manifest_dir': PosixPath('data/fbank'), 'max_duration': 500, 'bucketing_sampler': True, 'num_buckets': 30, 'concatenate_cuts': False, 'duration_factor': 1.0, 'gap': 1.0, 'on_the_fly_feats': False, 'shuffle': True, 'return_cuts': True, 'num_workers': 2, 'enable_spec_aug': True, 'spec_aug_time_warp_factor': 80, 'enable_musan': True, 'subset': 'S', 'small_dev': False, 'blank_id': 0, 'vocab_size': 500} 2023-05-15 15:28:40,546 INFO [finetune.py:1083] (0/2) About to create model 2023-05-15 15:28:41,276 INFO [zipformer.py:178] (0/2) At encoder stack 4, which has downsampling_factor=2, we will combine the outputs of layers 1 and 3, with downsampling_factors=2 and 8. 2023-05-15 15:28:41,295 INFO [finetune.py:1087] (0/2) Number of model parameters: 70369391 2023-05-15 15:28:41,850 INFO [finetune.py:639] (0/2) Loading checkpoint from icefall-asr-librispeech-pruned-transducer-stateless7-2022-11-11/exp/pretrained.pt 2023-05-15 15:28:44,487 INFO [finetune.py:1109] (0/2) Using DDP 2023-05-15 15:28:44,732 INFO [asr_datamodule.py:425] (0/2) About to get the shuffled train-clean-100, train-clean-360 and train-other-500 cuts 2023-05-15 15:28:44,734 INFO [gigaspeech.py:389] (0/2) About to get train_S cuts 2023-05-15 15:28:44,735 INFO [gigaspeech.py:216] (0/2) Enable MUSAN 2023-05-15 15:28:44,735 INFO [gigaspeech.py:217] (0/2) About to get Musan cuts 2023-05-15 15:28:47,343 INFO [gigaspeech.py:241] (0/2) Enable SpecAugment 2023-05-15 15:28:47,343 INFO [gigaspeech.py:242] (0/2) Time warp factor: 80 2023-05-15 15:28:47,343 INFO [gigaspeech.py:252] (0/2) Num frame mask: 10 2023-05-15 15:28:47,343 INFO [gigaspeech.py:265] (0/2) About to create train dataset 2023-05-15 15:28:47,343 INFO [gigaspeech.py:291] (0/2) Using DynamicBucketingSampler. 2023-05-15 15:28:53,537 INFO [gigaspeech.py:306] (0/2) About to create train dataloader 2023-05-15 15:28:53,538 INFO [gigaspeech.py:396] (0/2) About to get dev cuts 2023-05-15 15:28:53,539 INFO [gigaspeech.py:337] (0/2) About to create dev dataset 2023-05-15 15:28:53,953 INFO [gigaspeech.py:354] (0/2) About to create dev dataloader 2023-05-15 15:29:16,361 INFO [finetune.py:992] (0/2) Epoch 1, batch 0, loss[loss=0.415, simple_loss=0.3941, pruned_loss=0.2098, over 12000.00 frames. ], tot_loss[loss=0.415, simple_loss=0.3941, pruned_loss=0.2098, over 12000.00 frames. ], batch size: 28, lr: 2.50e-03, grad_scale: 2.0 2023-05-15 15:29:16,362 INFO [finetune.py:1017] (0/2) Computing validation loss 2023-05-15 15:29:34,130 INFO [finetune.py:1026] (0/2) Epoch 1, validation: loss=0.4712, simple_loss=0.4526, pruned_loss=0.1872, over 1020973.00 frames. 2023-05-15 15:29:34,131 INFO [finetune.py:1027] (0/2) Maximum memory allocated so far is 9175MB 2023-05-15 15:29:39,281 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=100007.0, num_to_drop=1, layers_to_drop={0} 2023-05-15 15:29:56,809 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=100028.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 15:29:56,996 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.1665, 2.5818, 3.7428, 3.0987, 3.5273, 3.2422, 2.5322, 3.6554], device='cuda:0'), covar=tensor([0.0098, 0.0302, 0.0106, 0.0209, 0.0118, 0.0146, 0.0297, 0.0088], device='cuda:0'), in_proj_covar=tensor([0.0162, 0.0196, 0.0174, 0.0178, 0.0199, 0.0154, 0.0184, 0.0167], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-15 15:30:00,975 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-05-15 15:30:02,979 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=100036.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 15:30:14,101 INFO [finetune.py:992] (0/2) Epoch 1, batch 50, loss[loss=0.2092, simple_loss=0.2098, pruned_loss=0.01704, over 12296.00 frames. ], tot_loss[loss=0.2399, simple_loss=0.2383, pruned_loss=0.02939, over 542911.50 frames. ], batch size: 28, lr: 2.75e-03, grad_scale: 2.0 2023-05-15 15:30:27,243 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=100068.0, num_to_drop=1, layers_to_drop={0} 2023-05-15 15:30:33,285 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=100076.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 15:30:44,763 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.0312, 5.0073, 4.8308, 4.9110, 4.5681, 4.9885, 4.9778, 5.2331], device='cuda:0'), covar=tensor([0.0209, 0.0111, 0.0168, 0.0292, 0.0736, 0.0257, 0.0139, 0.0160], device='cuda:0'), in_proj_covar=tensor([0.0190, 0.0194, 0.0191, 0.0244, 0.0247, 0.0206, 0.0177, 0.0231], device='cuda:0'), out_proj_covar=tensor([0.0005, 0.0004, 0.0004, 0.0006, 0.0005, 0.0005, 0.0004, 0.0006], device='cuda:0') 2023-05-15 15:30:52,162 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.338e+02 1.779e+02 2.045e+02 2.486e+02 5.674e+02, threshold=4.090e+02, percent-clipped=0.0 2023-05-15 15:30:52,182 INFO [finetune.py:992] (0/2) Epoch 1, batch 100, loss[loss=0.2706, simple_loss=0.2734, pruned_loss=0.02792, over 12050.00 frames. ], tot_loss[loss=0.2391, simple_loss=0.2389, pruned_loss=0.02819, over 954198.37 frames. ], batch size: 42, lr: 3.00e-03, grad_scale: 2.0 2023-05-15 15:31:09,356 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=100124.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 15:31:09,473 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=100124.0, num_to_drop=1, layers_to_drop={0} 2023-05-15 15:31:11,011 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.7152, 2.9982, 3.8926, 4.7980, 4.0701, 4.7377, 4.1435, 3.4484], device='cuda:0'), covar=tensor([0.0024, 0.0293, 0.0129, 0.0025, 0.0114, 0.0048, 0.0077, 0.0275], device='cuda:0'), in_proj_covar=tensor([0.0078, 0.0120, 0.0105, 0.0075, 0.0100, 0.0108, 0.0084, 0.0135], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-15 15:31:18,543 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.5884, 3.6087, 3.3047, 3.2346, 2.8436, 2.7320, 3.6313, 2.3573], device='cuda:0'), covar=tensor([0.0323, 0.0127, 0.0126, 0.0159, 0.0334, 0.0317, 0.0095, 0.0424], device='cuda:0'), in_proj_covar=tensor([0.0190, 0.0163, 0.0154, 0.0184, 0.0209, 0.0203, 0.0161, 0.0201], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-15 15:31:29,620 INFO [finetune.py:992] (0/2) Epoch 1, batch 150, loss[loss=0.2288, simple_loss=0.2334, pruned_loss=0.02519, over 12043.00 frames. ], tot_loss[loss=0.2399, simple_loss=0.2409, pruned_loss=0.02932, over 1263525.68 frames. ], batch size: 31, lr: 3.25e-03, grad_scale: 2.0 2023-05-15 15:31:32,130 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.4943, 5.1030, 5.4802, 4.8275, 5.1095, 4.9255, 5.5144, 5.1140], device='cuda:0'), covar=tensor([0.0193, 0.0296, 0.0205, 0.0205, 0.0291, 0.0227, 0.0146, 0.0295], device='cuda:0'), in_proj_covar=tensor([0.0240, 0.0252, 0.0269, 0.0246, 0.0244, 0.0243, 0.0220, 0.0205], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0005, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-15 15:31:34,579 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.0674, 3.8657, 3.9730, 4.4002, 2.5944, 3.7585, 2.5808, 3.8225], device='cuda:0'), covar=tensor([0.1676, 0.0653, 0.0849, 0.0466, 0.1381, 0.0580, 0.1801, 0.1199], device='cuda:0'), in_proj_covar=tensor([0.0223, 0.0266, 0.0307, 0.0374, 0.0247, 0.0240, 0.0265, 0.0424], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0004], device='cuda:0') 2023-05-15 15:31:41,723 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.32 vs. limit=2.0 2023-05-15 15:31:46,486 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=100172.0, num_to_drop=1, layers_to_drop={0} 2023-05-15 15:32:06,038 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=4.69 vs. limit=5.0 2023-05-15 15:32:08,957 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.456e+02 1.926e+02 2.279e+02 2.718e+02 1.176e+03, threshold=4.559e+02, percent-clipped=2.0 2023-05-15 15:32:08,976 INFO [finetune.py:992] (0/2) Epoch 1, batch 200, loss[loss=0.2121, simple_loss=0.2195, pruned_loss=0.01935, over 12256.00 frames. ], tot_loss[loss=0.2385, simple_loss=0.2411, pruned_loss=0.02947, over 1517725.93 frames. ], batch size: 28, lr: 3.50e-03, grad_scale: 2.0 2023-05-15 15:32:26,253 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=100224.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 15:32:34,431 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.1603, 2.6005, 3.7538, 3.1097, 3.5319, 3.2749, 2.5429, 3.6475], device='cuda:0'), covar=tensor([0.0092, 0.0272, 0.0091, 0.0187, 0.0117, 0.0123, 0.0286, 0.0084], device='cuda:0'), in_proj_covar=tensor([0.0163, 0.0197, 0.0175, 0.0179, 0.0200, 0.0155, 0.0185, 0.0168], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-15 15:32:36,606 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=100238.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 15:32:41,866 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=100245.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 15:32:46,194 INFO [finetune.py:992] (0/2) Epoch 1, batch 250, loss[loss=0.2408, simple_loss=0.2503, pruned_loss=0.0292, over 12275.00 frames. ], tot_loss[loss=0.2388, simple_loss=0.2429, pruned_loss=0.03049, over 1702053.50 frames. ], batch size: 37, lr: 3.75e-03, grad_scale: 2.0 2023-05-15 15:33:01,963 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=100272.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 15:33:17,712 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=100293.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 15:33:21,537 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.2983, 5.2859, 5.1532, 5.1656, 4.8913, 5.2499, 5.2636, 5.4947], device='cuda:0'), covar=tensor([0.0130, 0.0083, 0.0110, 0.0228, 0.0605, 0.0216, 0.0094, 0.0110], device='cuda:0'), in_proj_covar=tensor([0.0187, 0.0191, 0.0188, 0.0240, 0.0244, 0.0203, 0.0175, 0.0228], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0006, 0.0005, 0.0005, 0.0004, 0.0006], device='cuda:0') 2023-05-15 15:33:22,377 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=100299.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 15:33:23,643 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.434e+02 1.969e+02 2.308e+02 2.769e+02 4.182e+02, threshold=4.615e+02, percent-clipped=0.0 2023-05-15 15:33:23,661 INFO [finetune.py:992] (0/2) Epoch 1, batch 300, loss[loss=0.2373, simple_loss=0.2499, pruned_loss=0.02632, over 12132.00 frames. ], tot_loss[loss=0.2393, simple_loss=0.2451, pruned_loss=0.03103, over 1850167.62 frames. ], batch size: 36, lr: 4.00e-03, grad_scale: 2.0 2023-05-15 15:33:49,050 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=3.53 vs. limit=5.0 2023-05-15 15:33:51,784 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=100336.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 15:34:02,758 INFO [finetune.py:992] (0/2) Epoch 1, batch 350, loss[loss=0.1856, simple_loss=0.2003, pruned_loss=0.01123, over 11992.00 frames. ], tot_loss[loss=0.237, simple_loss=0.2445, pruned_loss=0.03099, over 1972389.04 frames. ], batch size: 28, lr: 4.25e-03, grad_scale: 2.0 2023-05-15 15:34:02,973 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.0409, 4.8287, 4.9420, 4.9447, 4.8009, 5.0227, 4.9362, 2.8453], device='cuda:0'), covar=tensor([0.0096, 0.0052, 0.0066, 0.0053, 0.0045, 0.0068, 0.0061, 0.0658], device='cuda:0'), in_proj_covar=tensor([0.0066, 0.0077, 0.0079, 0.0073, 0.0061, 0.0091, 0.0078, 0.0092], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0003, 0.0002, 0.0004, 0.0003, 0.0004], device='cuda:0') 2023-05-15 15:34:12,118 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=100363.0, num_to_drop=1, layers_to_drop={2} 2023-05-15 15:34:27,749 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=100384.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 15:34:40,985 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.635e+02 2.106e+02 2.479e+02 2.883e+02 4.026e+02, threshold=4.958e+02, percent-clipped=0.0 2023-05-15 15:34:41,003 INFO [finetune.py:992] (0/2) Epoch 1, batch 400, loss[loss=0.2504, simple_loss=0.2653, pruned_loss=0.0417, over 10391.00 frames. ], tot_loss[loss=0.2355, simple_loss=0.2448, pruned_loss=0.03141, over 2064017.33 frames. ], batch size: 68, lr: 4.50e-03, grad_scale: 4.0 2023-05-15 15:34:41,975 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=100402.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 15:34:55,701 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.1678, 3.6269, 5.4070, 2.8856, 2.9334, 3.9273, 3.4246, 3.9893], device='cuda:0'), covar=tensor([0.0279, 0.0970, 0.0160, 0.1038, 0.1936, 0.1278, 0.1250, 0.1119], device='cuda:0'), in_proj_covar=tensor([0.0242, 0.0239, 0.0249, 0.0188, 0.0249, 0.0302, 0.0234, 0.0267], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-15 15:35:18,809 INFO [finetune.py:992] (0/2) Epoch 1, batch 450, loss[loss=0.2177, simple_loss=0.2335, pruned_loss=0.03491, over 12168.00 frames. ], tot_loss[loss=0.2353, simple_loss=0.2464, pruned_loss=0.03234, over 2130580.34 frames. ], batch size: 31, lr: 4.75e-03, grad_scale: 4.0 2023-05-15 15:35:28,085 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=100463.0, num_to_drop=1, layers_to_drop={2} 2023-05-15 15:35:57,653 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.556e+02 2.141e+02 2.472e+02 2.964e+02 7.480e+02, threshold=4.945e+02, percent-clipped=1.0 2023-05-15 15:35:57,675 INFO [finetune.py:992] (0/2) Epoch 1, batch 500, loss[loss=0.2346, simple_loss=0.2569, pruned_loss=0.03055, over 12200.00 frames. ], tot_loss[loss=0.2351, simple_loss=0.248, pruned_loss=0.03327, over 2177028.05 frames. ], batch size: 35, lr: 5.00e-03, grad_scale: 4.0 2023-05-15 15:36:06,995 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.4979, 2.5916, 3.7123, 4.5622, 3.9848, 4.4463, 3.9870, 3.0159], device='cuda:0'), covar=tensor([0.0026, 0.0353, 0.0122, 0.0025, 0.0100, 0.0077, 0.0073, 0.0358], device='cuda:0'), in_proj_covar=tensor([0.0079, 0.0122, 0.0107, 0.0076, 0.0102, 0.0110, 0.0086, 0.0137], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-15 15:36:17,061 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.56 vs. limit=2.0 2023-05-15 15:36:35,195 INFO [finetune.py:992] (0/2) Epoch 1, batch 550, loss[loss=0.2155, simple_loss=0.2357, pruned_loss=0.03503, over 12090.00 frames. ], tot_loss[loss=0.2344, simple_loss=0.2493, pruned_loss=0.03395, over 2219223.09 frames. ], batch size: 32, lr: 5.00e-03, grad_scale: 4.0 2023-05-15 15:36:52,400 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.0071, 5.8543, 5.3534, 5.3470, 5.9875, 5.2319, 5.5863, 5.4241], device='cuda:0'), covar=tensor([0.1453, 0.0842, 0.1038, 0.1973, 0.0836, 0.1938, 0.1701, 0.1065], device='cuda:0'), in_proj_covar=tensor([0.0352, 0.0490, 0.0385, 0.0443, 0.0470, 0.0438, 0.0408, 0.0382], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-15 15:36:56,974 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.5413, 2.6084, 3.7223, 4.5447, 3.9429, 4.4001, 3.9954, 2.8211], device='cuda:0'), covar=tensor([0.0026, 0.0358, 0.0120, 0.0026, 0.0089, 0.0073, 0.0079, 0.0403], device='cuda:0'), in_proj_covar=tensor([0.0079, 0.0123, 0.0108, 0.0076, 0.0103, 0.0110, 0.0086, 0.0138], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-15 15:36:57,658 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.1090, 5.0703, 4.8987, 4.9430, 4.5930, 5.1111, 5.0370, 5.2835], device='cuda:0'), covar=tensor([0.0199, 0.0124, 0.0200, 0.0271, 0.0758, 0.0208, 0.0145, 0.0159], device='cuda:0'), in_proj_covar=tensor([0.0188, 0.0192, 0.0188, 0.0240, 0.0245, 0.0203, 0.0176, 0.0229], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0006, 0.0005, 0.0005, 0.0004, 0.0006], device='cuda:0') 2023-05-15 15:37:07,486 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=100594.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 15:37:13,163 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.473e+02 2.361e+02 2.670e+02 3.184e+02 7.098e+02, threshold=5.341e+02, percent-clipped=2.0 2023-05-15 15:37:13,184 INFO [finetune.py:992] (0/2) Epoch 1, batch 600, loss[loss=0.2231, simple_loss=0.2491, pruned_loss=0.03069, over 11265.00 frames. ], tot_loss[loss=0.2328, simple_loss=0.2499, pruned_loss=0.03397, over 2252862.84 frames. ], batch size: 55, lr: 5.00e-03, grad_scale: 4.0 2023-05-15 15:37:25,926 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.6766, 5.0540, 4.2663, 5.3304, 4.9283, 3.3012, 4.5460, 3.3692], device='cuda:0'), covar=tensor([0.0647, 0.0527, 0.1346, 0.0223, 0.0944, 0.1371, 0.0815, 0.2927], device='cuda:0'), in_proj_covar=tensor([0.0300, 0.0380, 0.0362, 0.0266, 0.0372, 0.0272, 0.0342, 0.0374], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-15 15:37:51,399 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.3431, 4.2102, 4.0663, 4.5294, 3.1076, 4.0119, 2.6417, 4.1728], device='cuda:0'), covar=tensor([0.1441, 0.0566, 0.0922, 0.0511, 0.1063, 0.0533, 0.1776, 0.1210], device='cuda:0'), in_proj_covar=tensor([0.0223, 0.0265, 0.0306, 0.0372, 0.0250, 0.0240, 0.0266, 0.0421], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0004], device='cuda:0') 2023-05-15 15:37:51,897 INFO [finetune.py:992] (0/2) Epoch 1, batch 650, loss[loss=0.2523, simple_loss=0.2772, pruned_loss=0.05151, over 12151.00 frames. ], tot_loss[loss=0.2311, simple_loss=0.2501, pruned_loss=0.03442, over 2284197.11 frames. ], batch size: 34, lr: 5.00e-03, grad_scale: 4.0 2023-05-15 15:38:01,000 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=100663.0, num_to_drop=1, layers_to_drop={1} 2023-05-15 15:38:03,988 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.0044, 4.4476, 4.0017, 4.7737, 3.7283, 4.3086, 2.8976, 4.6730], device='cuda:0'), covar=tensor([0.1059, 0.0532, 0.1260, 0.0672, 0.0835, 0.0454, 0.1580, 0.0950], device='cuda:0'), in_proj_covar=tensor([0.0223, 0.0265, 0.0306, 0.0372, 0.0250, 0.0240, 0.0266, 0.0421], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0004], device='cuda:0') 2023-05-15 15:38:28,976 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.657e+02 2.560e+02 2.965e+02 3.612e+02 8.428e+02, threshold=5.929e+02, percent-clipped=4.0 2023-05-15 15:38:28,995 INFO [finetune.py:992] (0/2) Epoch 1, batch 700, loss[loss=0.2316, simple_loss=0.2578, pruned_loss=0.0456, over 12341.00 frames. ], tot_loss[loss=0.2297, simple_loss=0.2507, pruned_loss=0.03515, over 2302571.25 frames. ], batch size: 36, lr: 5.00e-03, grad_scale: 4.0 2023-05-15 15:38:36,540 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=100711.0, num_to_drop=1, layers_to_drop={0} 2023-05-15 15:38:36,706 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.4780, 4.2236, 4.2339, 4.6617, 3.2506, 3.9875, 2.7719, 4.2744], device='cuda:0'), covar=tensor([0.1453, 0.0633, 0.0821, 0.0486, 0.1150, 0.0585, 0.1848, 0.1257], device='cuda:0'), in_proj_covar=tensor([0.0223, 0.0266, 0.0307, 0.0373, 0.0250, 0.0241, 0.0266, 0.0422], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0004], device='cuda:0') 2023-05-15 15:38:43,902 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.5566, 5.2037, 5.5334, 4.8674, 5.1933, 4.9391, 5.5782, 5.1398], device='cuda:0'), covar=tensor([0.0175, 0.0257, 0.0200, 0.0211, 0.0251, 0.0233, 0.0145, 0.0218], device='cuda:0'), in_proj_covar=tensor([0.0238, 0.0251, 0.0267, 0.0243, 0.0242, 0.0243, 0.0222, 0.0206], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0005, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-15 15:38:50,755 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=100730.0, num_to_drop=1, layers_to_drop={1} 2023-05-15 15:39:06,034 INFO [finetune.py:992] (0/2) Epoch 1, batch 750, loss[loss=0.2493, simple_loss=0.2852, pruned_loss=0.04034, over 11844.00 frames. ], tot_loss[loss=0.2278, simple_loss=0.2509, pruned_loss=0.03529, over 2317772.59 frames. ], batch size: 44, lr: 5.00e-03, grad_scale: 4.0 2023-05-15 15:39:11,461 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=100758.0, num_to_drop=1, layers_to_drop={1} 2023-05-15 15:39:37,530 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=100791.0, num_to_drop=1, layers_to_drop={2} 2023-05-15 15:39:45,147 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.753e+02 2.417e+02 2.979e+02 3.608e+02 6.266e+02, threshold=5.958e+02, percent-clipped=1.0 2023-05-15 15:39:45,166 INFO [finetune.py:992] (0/2) Epoch 1, batch 800, loss[loss=0.1995, simple_loss=0.2338, pruned_loss=0.02714, over 12025.00 frames. ], tot_loss[loss=0.2271, simple_loss=0.2522, pruned_loss=0.03621, over 2323833.20 frames. ], batch size: 31, lr: 5.00e-03, grad_scale: 8.0 2023-05-15 15:40:22,068 INFO [finetune.py:992] (0/2) Epoch 1, batch 850, loss[loss=0.2263, simple_loss=0.2655, pruned_loss=0.03575, over 10567.00 frames. ], tot_loss[loss=0.2239, simple_loss=0.2512, pruned_loss=0.03595, over 2339084.74 frames. ], batch size: 68, lr: 5.00e-03, grad_scale: 8.0 2023-05-15 15:40:54,137 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=100894.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 15:40:56,717 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.37 vs. limit=2.0 2023-05-15 15:40:59,138 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.511e+02 2.549e+02 3.052e+02 3.672e+02 5.649e+02, threshold=6.104e+02, percent-clipped=0.0 2023-05-15 15:40:59,157 INFO [finetune.py:992] (0/2) Epoch 1, batch 900, loss[loss=0.2067, simple_loss=0.249, pruned_loss=0.02731, over 11140.00 frames. ], tot_loss[loss=0.2217, simple_loss=0.2512, pruned_loss=0.03608, over 2353781.08 frames. ], batch size: 55, lr: 5.00e-03, grad_scale: 8.0 2023-05-15 15:41:29,902 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.95 vs. limit=2.0 2023-05-15 15:41:30,882 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=100942.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 15:41:37,562 INFO [finetune.py:992] (0/2) Epoch 1, batch 950, loss[loss=0.1976, simple_loss=0.2322, pruned_loss=0.03913, over 11389.00 frames. ], tot_loss[loss=0.2192, simple_loss=0.2509, pruned_loss=0.03612, over 2362487.98 frames. ], batch size: 25, lr: 5.00e-03, grad_scale: 8.0 2023-05-15 15:42:13,383 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.7882, 2.7688, 4.7116, 4.7700, 2.9782, 2.7829, 3.0203, 2.1814], device='cuda:0'), covar=tensor([0.1348, 0.2734, 0.0350, 0.0322, 0.1038, 0.1861, 0.2279, 0.3281], device='cuda:0'), in_proj_covar=tensor([0.0292, 0.0389, 0.0273, 0.0298, 0.0259, 0.0290, 0.0370, 0.0358], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-15 15:42:15,815 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.693e+02 2.497e+02 2.927e+02 3.305e+02 5.891e+02, threshold=5.853e+02, percent-clipped=0.0 2023-05-15 15:42:15,834 INFO [finetune.py:992] (0/2) Epoch 1, batch 1000, loss[loss=0.2074, simple_loss=0.2517, pruned_loss=0.03396, over 12244.00 frames. ], tot_loss[loss=0.2166, simple_loss=0.2505, pruned_loss=0.03604, over 2365123.88 frames. ], batch size: 32, lr: 5.00e-03, grad_scale: 8.0 2023-05-15 15:42:23,335 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.2819, 2.5273, 3.5412, 4.2524, 3.7485, 4.2931, 3.8345, 3.0400], device='cuda:0'), covar=tensor([0.0031, 0.0365, 0.0158, 0.0036, 0.0118, 0.0055, 0.0096, 0.0318], device='cuda:0'), in_proj_covar=tensor([0.0079, 0.0122, 0.0108, 0.0076, 0.0103, 0.0110, 0.0086, 0.0138], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-15 15:42:52,955 INFO [finetune.py:992] (0/2) Epoch 1, batch 1050, loss[loss=0.2087, simple_loss=0.2572, pruned_loss=0.03318, over 12092.00 frames. ], tot_loss[loss=0.2147, simple_loss=0.251, pruned_loss=0.03614, over 2370070.49 frames. ], batch size: 42, lr: 5.00e-03, grad_scale: 8.0 2023-05-15 15:42:58,372 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=101058.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 15:43:01,060 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=4.84 vs. limit=5.0 2023-05-15 15:43:03,775 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-05-15 15:43:10,965 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.5340, 2.5632, 3.6596, 4.4398, 4.0570, 4.5316, 4.0279, 3.4201], device='cuda:0'), covar=tensor([0.0023, 0.0372, 0.0140, 0.0043, 0.0091, 0.0055, 0.0088, 0.0267], device='cuda:0'), in_proj_covar=tensor([0.0077, 0.0120, 0.0106, 0.0075, 0.0101, 0.0108, 0.0084, 0.0135], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-15 15:43:19,135 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.6881, 2.6222, 4.7284, 4.9641, 3.1761, 2.8104, 2.9859, 2.1295], device='cuda:0'), covar=tensor([0.1387, 0.2836, 0.0325, 0.0239, 0.0912, 0.1855, 0.2323, 0.3479], device='cuda:0'), in_proj_covar=tensor([0.0291, 0.0388, 0.0272, 0.0298, 0.0259, 0.0290, 0.0369, 0.0356], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-15 15:43:19,767 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.3131, 2.4923, 3.5942, 4.2397, 3.8241, 4.2554, 3.9004, 3.2275], device='cuda:0'), covar=tensor([0.0026, 0.0345, 0.0133, 0.0036, 0.0103, 0.0059, 0.0083, 0.0266], device='cuda:0'), in_proj_covar=tensor([0.0077, 0.0120, 0.0106, 0.0075, 0.0101, 0.0108, 0.0084, 0.0135], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-15 15:43:20,367 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=101086.0, num_to_drop=1, layers_to_drop={1} 2023-05-15 15:43:31,226 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.688e+02 2.509e+02 2.900e+02 3.477e+02 6.199e+02, threshold=5.799e+02, percent-clipped=2.0 2023-05-15 15:43:31,244 INFO [finetune.py:992] (0/2) Epoch 1, batch 1100, loss[loss=0.2129, simple_loss=0.2603, pruned_loss=0.04074, over 12112.00 frames. ], tot_loss[loss=0.2124, simple_loss=0.2509, pruned_loss=0.03607, over 2376319.46 frames. ], batch size: 38, lr: 5.00e-03, grad_scale: 8.0 2023-05-15 15:43:35,075 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=101106.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 15:43:35,933 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=101107.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 15:44:08,188 INFO [finetune.py:992] (0/2) Epoch 1, batch 1150, loss[loss=0.1891, simple_loss=0.2381, pruned_loss=0.03155, over 12183.00 frames. ], tot_loss[loss=0.2114, simple_loss=0.252, pruned_loss=0.03686, over 2371419.63 frames. ], batch size: 31, lr: 5.00e-03, grad_scale: 8.0 2023-05-15 15:44:20,947 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=101168.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 15:44:21,012 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.0169, 2.9422, 4.9239, 5.0846, 3.2427, 3.0264, 3.0934, 2.3851], device='cuda:0'), covar=tensor([0.1272, 0.2493, 0.0316, 0.0275, 0.0955, 0.1673, 0.2282, 0.3338], device='cuda:0'), in_proj_covar=tensor([0.0292, 0.0390, 0.0274, 0.0299, 0.0260, 0.0291, 0.0371, 0.0359], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-15 15:44:21,775 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.6691, 3.1715, 4.9878, 2.5744, 2.7018, 3.7366, 3.0912, 3.9264], device='cuda:0'), covar=tensor([0.0407, 0.1158, 0.0259, 0.1228, 0.2063, 0.1421, 0.1443, 0.1049], device='cuda:0'), in_proj_covar=tensor([0.0241, 0.0239, 0.0248, 0.0187, 0.0248, 0.0298, 0.0234, 0.0265], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-15 15:44:45,564 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.521e+02 2.377e+02 2.892e+02 3.477e+02 1.072e+03, threshold=5.784e+02, percent-clipped=5.0 2023-05-15 15:44:45,582 INFO [finetune.py:992] (0/2) Epoch 1, batch 1200, loss[loss=0.2228, simple_loss=0.2733, pruned_loss=0.04933, over 11627.00 frames. ], tot_loss[loss=0.2087, simple_loss=0.2514, pruned_loss=0.03692, over 2382993.92 frames. ], batch size: 48, lr: 5.00e-03, grad_scale: 8.0 2023-05-15 15:45:11,249 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.35 vs. limit=2.0 2023-05-15 15:45:13,770 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([6.0915, 6.0738, 5.9259, 5.2914, 5.2981, 5.9859, 5.6052, 5.4439], device='cuda:0'), covar=tensor([0.0633, 0.0796, 0.0477, 0.1346, 0.0563, 0.0617, 0.1486, 0.0890], device='cuda:0'), in_proj_covar=tensor([0.0609, 0.0540, 0.0483, 0.0611, 0.0386, 0.0694, 0.0768, 0.0548], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], device='cuda:0') 2023-05-15 15:45:23,999 INFO [finetune.py:992] (0/2) Epoch 1, batch 1250, loss[loss=0.2245, simple_loss=0.284, pruned_loss=0.04426, over 12197.00 frames. ], tot_loss[loss=0.2074, simple_loss=0.2525, pruned_loss=0.0374, over 2381625.67 frames. ], batch size: 35, lr: 5.00e-03, grad_scale: 8.0 2023-05-15 15:45:33,036 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.5722, 2.6429, 3.2906, 4.4896, 2.3711, 4.4507, 4.5163, 4.7352], device='cuda:0'), covar=tensor([0.0122, 0.1182, 0.0466, 0.0123, 0.1301, 0.0259, 0.0134, 0.0060], device='cuda:0'), in_proj_covar=tensor([0.0124, 0.0212, 0.0196, 0.0119, 0.0194, 0.0190, 0.0177, 0.0123], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002], device='cuda:0') 2023-05-15 15:46:00,728 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.678e+02 2.665e+02 3.108e+02 3.746e+02 6.531e+02, threshold=6.215e+02, percent-clipped=2.0 2023-05-15 15:46:00,746 INFO [finetune.py:992] (0/2) Epoch 1, batch 1300, loss[loss=0.1807, simple_loss=0.2307, pruned_loss=0.03655, over 12131.00 frames. ], tot_loss[loss=0.2054, simple_loss=0.2526, pruned_loss=0.03761, over 2376701.92 frames. ], batch size: 30, lr: 5.00e-03, grad_scale: 8.0 2023-05-15 15:46:25,462 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-05-15 15:46:37,426 INFO [finetune.py:992] (0/2) Epoch 1, batch 1350, loss[loss=0.2126, simple_loss=0.2743, pruned_loss=0.0437, over 11852.00 frames. ], tot_loss[loss=0.204, simple_loss=0.2537, pruned_loss=0.03794, over 2379507.61 frames. ], batch size: 44, lr: 5.00e-03, grad_scale: 8.0 2023-05-15 15:47:01,911 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.3141, 4.7377, 3.0311, 2.7636, 4.1227, 2.4776, 4.0709, 3.2768], device='cuda:0'), covar=tensor([0.0623, 0.0388, 0.0916, 0.1353, 0.0246, 0.1272, 0.0387, 0.0727], device='cuda:0'), in_proj_covar=tensor([0.0198, 0.0262, 0.0177, 0.0201, 0.0142, 0.0182, 0.0200, 0.0177], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-15 15:47:04,877 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=101386.0, num_to_drop=1, layers_to_drop={2} 2023-05-15 15:47:09,946 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.5631, 5.4995, 5.3175, 4.8012, 4.9263, 5.4593, 5.0987, 4.8882], device='cuda:0'), covar=tensor([0.0660, 0.0849, 0.0607, 0.1302, 0.0849, 0.0687, 0.1386, 0.1027], device='cuda:0'), in_proj_covar=tensor([0.0605, 0.0543, 0.0483, 0.0609, 0.0385, 0.0693, 0.0766, 0.0547], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], device='cuda:0') 2023-05-15 15:47:13,757 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.1197, 5.9278, 5.4260, 5.3892, 6.0568, 5.3588, 5.5276, 5.4745], device='cuda:0'), covar=tensor([0.1480, 0.0962, 0.1061, 0.1933, 0.0957, 0.2157, 0.1883, 0.1044], device='cuda:0'), in_proj_covar=tensor([0.0343, 0.0476, 0.0377, 0.0426, 0.0457, 0.0426, 0.0398, 0.0375], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-15 15:47:16,292 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.626e+02 2.435e+02 2.915e+02 3.611e+02 6.080e+02, threshold=5.831e+02, percent-clipped=0.0 2023-05-15 15:47:16,310 INFO [finetune.py:992] (0/2) Epoch 1, batch 1400, loss[loss=0.259, simple_loss=0.2998, pruned_loss=0.08795, over 7512.00 frames. ], tot_loss[loss=0.2016, simple_loss=0.2538, pruned_loss=0.03767, over 2383805.80 frames. ], batch size: 97, lr: 5.00e-03, grad_scale: 8.0 2023-05-15 15:47:40,795 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=101434.0, num_to_drop=1, layers_to_drop={0} 2023-05-15 15:47:52,834 INFO [finetune.py:992] (0/2) Epoch 1, batch 1450, loss[loss=0.1673, simple_loss=0.2268, pruned_loss=0.03022, over 12189.00 frames. ], tot_loss[loss=0.1998, simple_loss=0.2542, pruned_loss=0.03795, over 2385380.80 frames. ], batch size: 29, lr: 5.00e-03, grad_scale: 8.0 2023-05-15 15:48:01,716 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=101463.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 15:48:22,959 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.1769, 5.1729, 5.0298, 5.1213, 4.6680, 5.1179, 5.1312, 5.3518], device='cuda:0'), covar=tensor([0.0178, 0.0102, 0.0141, 0.0203, 0.0710, 0.0237, 0.0136, 0.0138], device='cuda:0'), in_proj_covar=tensor([0.0187, 0.0191, 0.0186, 0.0239, 0.0240, 0.0202, 0.0175, 0.0226], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0006, 0.0005, 0.0005, 0.0004, 0.0005], device='cuda:0') 2023-05-15 15:48:29,431 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.791e+02 2.709e+02 3.317e+02 3.771e+02 6.828e+02, threshold=6.633e+02, percent-clipped=3.0 2023-05-15 15:48:29,450 INFO [finetune.py:992] (0/2) Epoch 1, batch 1500, loss[loss=0.1784, simple_loss=0.2419, pruned_loss=0.0352, over 11771.00 frames. ], tot_loss[loss=0.1979, simple_loss=0.2547, pruned_loss=0.03817, over 2383825.43 frames. ], batch size: 26, lr: 5.00e-03, grad_scale: 8.0 2023-05-15 15:48:49,543 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.3937, 4.1435, 4.0911, 4.5697, 3.0974, 3.9937, 2.7929, 4.1194], device='cuda:0'), covar=tensor([0.1454, 0.0604, 0.0922, 0.0497, 0.1069, 0.0601, 0.1678, 0.1324], device='cuda:0'), in_proj_covar=tensor([0.0223, 0.0267, 0.0309, 0.0371, 0.0251, 0.0242, 0.0265, 0.0421], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0004], device='cuda:0') 2023-05-15 15:48:53,204 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.0202, 4.6895, 4.8375, 4.8758, 4.8581, 4.8536, 4.8047, 2.7748], device='cuda:0'), covar=tensor([0.0100, 0.0060, 0.0072, 0.0058, 0.0041, 0.0084, 0.0068, 0.0662], device='cuda:0'), in_proj_covar=tensor([0.0067, 0.0078, 0.0080, 0.0075, 0.0062, 0.0092, 0.0080, 0.0094], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0003, 0.0002, 0.0004, 0.0003, 0.0004], device='cuda:0') 2023-05-15 15:49:07,504 INFO [finetune.py:992] (0/2) Epoch 1, batch 1550, loss[loss=0.1679, simple_loss=0.2312, pruned_loss=0.03306, over 12331.00 frames. ], tot_loss[loss=0.1957, simple_loss=0.2546, pruned_loss=0.03834, over 2387334.11 frames. ], batch size: 30, lr: 5.00e-03, grad_scale: 8.0 2023-05-15 15:49:45,137 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.950e+02 2.752e+02 3.202e+02 3.781e+02 7.212e+02, threshold=6.405e+02, percent-clipped=2.0 2023-05-15 15:49:45,160 INFO [finetune.py:992] (0/2) Epoch 1, batch 1600, loss[loss=0.199, simple_loss=0.2642, pruned_loss=0.04941, over 12154.00 frames. ], tot_loss[loss=0.1933, simple_loss=0.2545, pruned_loss=0.03831, over 2389116.23 frames. ], batch size: 36, lr: 5.00e-03, grad_scale: 8.0 2023-05-15 15:50:23,083 INFO [finetune.py:992] (0/2) Epoch 1, batch 1650, loss[loss=0.2056, simple_loss=0.2816, pruned_loss=0.0478, over 12362.00 frames. ], tot_loss[loss=0.1909, simple_loss=0.2543, pruned_loss=0.03829, over 2385876.97 frames. ], batch size: 36, lr: 5.00e-03, grad_scale: 8.0 2023-05-15 15:50:28,554 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.1505, 2.4536, 3.6693, 3.1401, 3.5598, 3.1077, 2.4108, 3.5655], device='cuda:0'), covar=tensor([0.0114, 0.0296, 0.0101, 0.0178, 0.0102, 0.0149, 0.0314, 0.0098], device='cuda:0'), in_proj_covar=tensor([0.0165, 0.0197, 0.0174, 0.0178, 0.0199, 0.0156, 0.0186, 0.0165], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-15 15:50:48,442 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=101684.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 15:50:49,265 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.5457, 4.8973, 4.5220, 5.3162, 4.8267, 3.0438, 4.3616, 3.3124], device='cuda:0'), covar=tensor([0.0702, 0.0658, 0.1064, 0.0216, 0.0822, 0.1490, 0.0955, 0.2781], device='cuda:0'), in_proj_covar=tensor([0.0302, 0.0382, 0.0363, 0.0268, 0.0374, 0.0271, 0.0341, 0.0372], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-15 15:50:51,408 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.0417, 4.6875, 4.9000, 4.8477, 4.7209, 4.9584, 4.7963, 2.9602], device='cuda:0'), covar=tensor([0.0093, 0.0071, 0.0076, 0.0070, 0.0057, 0.0078, 0.0084, 0.0580], device='cuda:0'), in_proj_covar=tensor([0.0067, 0.0078, 0.0080, 0.0074, 0.0062, 0.0092, 0.0080, 0.0093], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0003, 0.0002, 0.0004, 0.0003, 0.0004], device='cuda:0') 2023-05-15 15:51:00,878 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.806e+02 2.787e+02 3.159e+02 3.724e+02 8.656e+02, threshold=6.317e+02, percent-clipped=3.0 2023-05-15 15:51:00,897 INFO [finetune.py:992] (0/2) Epoch 1, batch 1700, loss[loss=0.1734, simple_loss=0.247, pruned_loss=0.03628, over 12271.00 frames. ], tot_loss[loss=0.1888, simple_loss=0.2544, pruned_loss=0.03859, over 2382188.29 frames. ], batch size: 28, lr: 5.00e-03, grad_scale: 8.0 2023-05-15 15:51:33,534 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=101745.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 15:51:37,785 INFO [finetune.py:992] (0/2) Epoch 1, batch 1750, loss[loss=0.1735, simple_loss=0.2509, pruned_loss=0.03656, over 12085.00 frames. ], tot_loss[loss=0.1871, simple_loss=0.2547, pruned_loss=0.03906, over 2378403.56 frames. ], batch size: 32, lr: 5.00e-03, grad_scale: 8.0 2023-05-15 15:51:46,602 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=101763.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 15:51:48,430 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=4.55 vs. limit=5.0 2023-05-15 15:52:15,369 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.889e+02 2.759e+02 3.294e+02 3.844e+02 7.750e+02, threshold=6.588e+02, percent-clipped=3.0 2023-05-15 15:52:15,388 INFO [finetune.py:992] (0/2) Epoch 1, batch 1800, loss[loss=0.2054, simple_loss=0.2665, pruned_loss=0.06474, over 12401.00 frames. ], tot_loss[loss=0.185, simple_loss=0.2547, pruned_loss=0.03935, over 2379403.07 frames. ], batch size: 32, lr: 5.00e-03, grad_scale: 8.0 2023-05-15 15:52:22,680 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=101811.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 15:52:27,750 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.9884, 5.7570, 5.1911, 5.3061, 5.8602, 5.1205, 5.3968, 5.4517], device='cuda:0'), covar=tensor([0.1446, 0.0877, 0.0975, 0.1862, 0.0939, 0.2167, 0.1703, 0.1047], device='cuda:0'), in_proj_covar=tensor([0.0337, 0.0473, 0.0372, 0.0419, 0.0454, 0.0419, 0.0392, 0.0371], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-15 15:52:52,276 INFO [finetune.py:992] (0/2) Epoch 1, batch 1850, loss[loss=0.2061, simple_loss=0.2913, pruned_loss=0.05317, over 12126.00 frames. ], tot_loss[loss=0.1832, simple_loss=0.2551, pruned_loss=0.03963, over 2377513.20 frames. ], batch size: 39, lr: 5.00e-03, grad_scale: 8.0 2023-05-15 15:52:53,333 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.9667, 2.9021, 4.7433, 4.7649, 3.0920, 2.9695, 3.1479, 2.3359], device='cuda:0'), covar=tensor([0.1288, 0.2529, 0.0357, 0.0346, 0.1054, 0.1707, 0.2219, 0.3164], device='cuda:0'), in_proj_covar=tensor([0.0291, 0.0388, 0.0273, 0.0299, 0.0260, 0.0290, 0.0370, 0.0360], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-15 15:53:09,627 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.2631, 4.6673, 4.2418, 5.0258, 4.5197, 2.8326, 4.3545, 3.0970], device='cuda:0'), covar=tensor([0.0775, 0.0720, 0.1418, 0.0312, 0.1001, 0.1630, 0.0920, 0.3127], device='cuda:0'), in_proj_covar=tensor([0.0301, 0.0380, 0.0362, 0.0266, 0.0373, 0.0270, 0.0339, 0.0372], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-15 15:53:16,871 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.7284, 2.9128, 5.3300, 2.3634, 2.7403, 4.2630, 2.8846, 4.2055], device='cuda:0'), covar=tensor([0.0439, 0.1451, 0.0188, 0.1331, 0.2010, 0.1056, 0.1690, 0.0905], device='cuda:0'), in_proj_covar=tensor([0.0243, 0.0240, 0.0251, 0.0189, 0.0250, 0.0297, 0.0236, 0.0267], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-15 15:53:29,172 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.979e+02 2.693e+02 3.375e+02 4.011e+02 6.163e+02, threshold=6.750e+02, percent-clipped=0.0 2023-05-15 15:53:29,190 INFO [finetune.py:992] (0/2) Epoch 1, batch 1900, loss[loss=0.1612, simple_loss=0.2488, pruned_loss=0.03208, over 12341.00 frames. ], tot_loss[loss=0.1807, simple_loss=0.2553, pruned_loss=0.0393, over 2382193.06 frames. ], batch size: 31, lr: 5.00e-03, grad_scale: 8.0 2023-05-15 15:53:38,269 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.4469, 3.5412, 3.2514, 3.2594, 2.8715, 2.7089, 3.5630, 2.2486], device='cuda:0'), covar=tensor([0.0330, 0.0119, 0.0131, 0.0147, 0.0326, 0.0275, 0.0092, 0.0413], device='cuda:0'), in_proj_covar=tensor([0.0186, 0.0160, 0.0150, 0.0179, 0.0204, 0.0197, 0.0157, 0.0196], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-15 15:53:43,458 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=101920.0, num_to_drop=1, layers_to_drop={1} 2023-05-15 15:53:50,757 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.2346, 4.7608, 5.1729, 4.5009, 4.8126, 4.5805, 5.1910, 4.7672], device='cuda:0'), covar=tensor([0.0228, 0.0362, 0.0271, 0.0269, 0.0314, 0.0332, 0.0198, 0.0301], device='cuda:0'), in_proj_covar=tensor([0.0236, 0.0249, 0.0264, 0.0241, 0.0241, 0.0240, 0.0218, 0.0202], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0005, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-15 15:53:50,778 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.8515, 4.8898, 4.6955, 4.8251, 4.3187, 4.7920, 4.8497, 5.1178], device='cuda:0'), covar=tensor([0.0206, 0.0119, 0.0175, 0.0230, 0.0829, 0.0231, 0.0154, 0.0170], device='cuda:0'), in_proj_covar=tensor([0.0189, 0.0193, 0.0188, 0.0239, 0.0242, 0.0204, 0.0176, 0.0227], device='cuda:0'), out_proj_covar=tensor([0.0005, 0.0004, 0.0004, 0.0006, 0.0005, 0.0005, 0.0004, 0.0006], device='cuda:0') 2023-05-15 15:54:06,214 INFO [finetune.py:992] (0/2) Epoch 1, batch 1950, loss[loss=0.1487, simple_loss=0.2342, pruned_loss=0.02941, over 12415.00 frames. ], tot_loss[loss=0.1774, simple_loss=0.2547, pruned_loss=0.03863, over 2389455.68 frames. ], batch size: 32, lr: 5.00e-03, grad_scale: 8.0 2023-05-15 15:54:29,077 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=101981.0, num_to_drop=1, layers_to_drop={1} 2023-05-15 15:54:30,594 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-05-15 15:54:33,269 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=101987.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 15:54:36,469 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=2.75 vs. limit=5.0 2023-05-15 15:54:43,300 INFO [checkpoint.py:75] (0/2) Saving checkpoint to pruned_transducer_stateless7/exp_giga_finetune/checkpoint-2000.pt 2023-05-15 15:54:47,260 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.827e+02 2.821e+02 3.357e+02 4.284e+02 8.103e+02, threshold=6.713e+02, percent-clipped=5.0 2023-05-15 15:54:47,279 INFO [finetune.py:992] (0/2) Epoch 1, batch 2000, loss[loss=0.1736, simple_loss=0.2714, pruned_loss=0.0379, over 12127.00 frames. ], tot_loss[loss=0.1746, simple_loss=0.2543, pruned_loss=0.03838, over 2394136.10 frames. ], batch size: 34, lr: 5.00e-03, grad_scale: 16.0 2023-05-15 15:54:51,992 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.7786, 3.4637, 5.2016, 2.7040, 2.8090, 3.9451, 3.2730, 3.9233], device='cuda:0'), covar=tensor([0.0400, 0.1059, 0.0238, 0.1193, 0.2003, 0.1219, 0.1356, 0.1144], device='cuda:0'), in_proj_covar=tensor([0.0240, 0.0236, 0.0247, 0.0186, 0.0246, 0.0293, 0.0232, 0.0262], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-15 15:54:55,728 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-05-15 15:55:15,797 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=102040.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 15:55:18,139 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.4179, 4.7450, 2.8260, 2.6087, 4.0327, 2.5800, 4.0210, 3.3872], device='cuda:0'), covar=tensor([0.0599, 0.0426, 0.1101, 0.1402, 0.0276, 0.1187, 0.0458, 0.0680], device='cuda:0'), in_proj_covar=tensor([0.0196, 0.0261, 0.0177, 0.0200, 0.0142, 0.0182, 0.0200, 0.0177], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-15 15:55:21,750 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=102048.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 15:55:23,735 INFO [finetune.py:992] (0/2) Epoch 1, batch 2050, loss[loss=0.1998, simple_loss=0.2863, pruned_loss=0.05664, over 12041.00 frames. ], tot_loss[loss=0.1727, simple_loss=0.2541, pruned_loss=0.03854, over 2397785.59 frames. ], batch size: 37, lr: 5.00e-03, grad_scale: 16.0 2023-05-15 15:55:49,929 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2023-05-15 15:56:01,300 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.025e+02 2.708e+02 3.160e+02 3.820e+02 1.513e+03, threshold=6.321e+02, percent-clipped=2.0 2023-05-15 15:56:01,318 INFO [finetune.py:992] (0/2) Epoch 1, batch 2100, loss[loss=0.153, simple_loss=0.2421, pruned_loss=0.03193, over 12098.00 frames. ], tot_loss[loss=0.171, simple_loss=0.254, pruned_loss=0.03845, over 2399251.29 frames. ], batch size: 33, lr: 5.00e-03, grad_scale: 16.0 2023-05-15 15:56:38,905 INFO [finetune.py:992] (0/2) Epoch 1, batch 2150, loss[loss=0.1635, simple_loss=0.2599, pruned_loss=0.03357, over 12302.00 frames. ], tot_loss[loss=0.1701, simple_loss=0.2543, pruned_loss=0.0387, over 2400792.02 frames. ], batch size: 33, lr: 5.00e-03, grad_scale: 16.0 2023-05-15 15:56:40,506 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.4935, 5.4576, 5.2841, 4.7936, 4.9612, 5.4600, 5.0202, 4.9336], device='cuda:0'), covar=tensor([0.0766, 0.0890, 0.0661, 0.1490, 0.0693, 0.0669, 0.1575, 0.0847], device='cuda:0'), in_proj_covar=tensor([0.0606, 0.0537, 0.0483, 0.0610, 0.0386, 0.0695, 0.0762, 0.0545], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], device='cuda:0') 2023-05-15 15:56:50,366 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=3.16 vs. limit=5.0 2023-05-15 15:56:59,321 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.42 vs. limit=2.0 2023-05-15 15:57:15,761 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.915e+02 2.922e+02 3.570e+02 4.264e+02 7.831e+02, threshold=7.141e+02, percent-clipped=3.0 2023-05-15 15:57:15,780 INFO [finetune.py:992] (0/2) Epoch 1, batch 2200, loss[loss=0.1632, simple_loss=0.2516, pruned_loss=0.0374, over 11998.00 frames. ], tot_loss[loss=0.1696, simple_loss=0.2549, pruned_loss=0.03883, over 2404396.37 frames. ], batch size: 28, lr: 5.00e-03, grad_scale: 16.0 2023-05-15 15:57:53,022 INFO [finetune.py:992] (0/2) Epoch 1, batch 2250, loss[loss=0.1796, simple_loss=0.2679, pruned_loss=0.04568, over 12024.00 frames. ], tot_loss[loss=0.1688, simple_loss=0.255, pruned_loss=0.03873, over 2400306.67 frames. ], batch size: 37, lr: 5.00e-03, grad_scale: 16.0 2023-05-15 15:57:56,941 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.5148, 3.6895, 3.3897, 3.2787, 2.8909, 2.8345, 3.6382, 2.1142], device='cuda:0'), covar=tensor([0.0330, 0.0105, 0.0124, 0.0149, 0.0315, 0.0268, 0.0096, 0.0449], device='cuda:0'), in_proj_covar=tensor([0.0185, 0.0160, 0.0151, 0.0180, 0.0203, 0.0195, 0.0158, 0.0197], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-15 15:58:12,326 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=102276.0, num_to_drop=1, layers_to_drop={0} 2023-05-15 15:58:18,301 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([6.1646, 6.0275, 5.8995, 5.3466, 5.2964, 6.0703, 5.6625, 5.5293], device='cuda:0'), covar=tensor([0.0577, 0.0941, 0.0617, 0.1530, 0.0599, 0.0609, 0.1511, 0.0941], device='cuda:0'), in_proj_covar=tensor([0.0603, 0.0534, 0.0483, 0.0606, 0.0387, 0.0692, 0.0760, 0.0545], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], device='cuda:0') 2023-05-15 15:58:30,595 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.758e+02 2.636e+02 3.287e+02 3.960e+02 8.443e+02, threshold=6.574e+02, percent-clipped=2.0 2023-05-15 15:58:30,613 INFO [finetune.py:992] (0/2) Epoch 1, batch 2300, loss[loss=0.1751, simple_loss=0.264, pruned_loss=0.04308, over 12023.00 frames. ], tot_loss[loss=0.1677, simple_loss=0.2547, pruned_loss=0.03838, over 2399543.65 frames. ], batch size: 40, lr: 5.00e-03, grad_scale: 16.0 2023-05-15 15:58:47,703 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.9897, 5.9323, 5.7548, 5.2952, 5.1806, 5.9301, 5.4595, 5.3734], device='cuda:0'), covar=tensor([0.0692, 0.0945, 0.0611, 0.1374, 0.0637, 0.0643, 0.1499, 0.0957], device='cuda:0'), in_proj_covar=tensor([0.0605, 0.0537, 0.0484, 0.0609, 0.0388, 0.0693, 0.0763, 0.0547], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], device='cuda:0') 2023-05-15 15:58:59,518 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=102340.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 15:59:01,683 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=102343.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 15:59:06,815 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([6.1727, 6.0868, 5.9042, 5.4577, 5.2699, 6.0721, 5.6767, 5.5138], device='cuda:0'), covar=tensor([0.0507, 0.0734, 0.0641, 0.1376, 0.0598, 0.0620, 0.1287, 0.0886], device='cuda:0'), in_proj_covar=tensor([0.0602, 0.0533, 0.0482, 0.0608, 0.0387, 0.0691, 0.0760, 0.0545], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], device='cuda:0') 2023-05-15 15:59:07,414 INFO [finetune.py:992] (0/2) Epoch 1, batch 2350, loss[loss=0.1616, simple_loss=0.2374, pruned_loss=0.04284, over 12175.00 frames. ], tot_loss[loss=0.1677, simple_loss=0.255, pruned_loss=0.03863, over 2398074.20 frames. ], batch size: 31, lr: 5.00e-03, grad_scale: 16.0 2023-05-15 15:59:34,829 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=102388.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 15:59:44,625 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.711e+02 2.777e+02 3.297e+02 3.992e+02 1.048e+03, threshold=6.593e+02, percent-clipped=3.0 2023-05-15 15:59:44,643 INFO [finetune.py:992] (0/2) Epoch 1, batch 2400, loss[loss=0.169, simple_loss=0.2593, pruned_loss=0.03936, over 11622.00 frames. ], tot_loss[loss=0.1679, simple_loss=0.2552, pruned_loss=0.0391, over 2386146.37 frames. ], batch size: 48, lr: 5.00e-03, grad_scale: 16.0 2023-05-15 15:59:50,723 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.81 vs. limit=2.0 2023-05-15 16:00:21,737 INFO [finetune.py:992] (0/2) Epoch 1, batch 2450, loss[loss=0.1788, simple_loss=0.2691, pruned_loss=0.04425, over 12136.00 frames. ], tot_loss[loss=0.168, simple_loss=0.2554, pruned_loss=0.03934, over 2387837.39 frames. ], batch size: 38, lr: 5.00e-03, grad_scale: 16.0 2023-05-15 16:00:54,888 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.8539, 4.8792, 4.7372, 4.8320, 4.3257, 4.8142, 4.8993, 5.1003], device='cuda:0'), covar=tensor([0.0226, 0.0126, 0.0177, 0.0245, 0.0706, 0.0217, 0.0128, 0.0144], device='cuda:0'), in_proj_covar=tensor([0.0188, 0.0193, 0.0189, 0.0237, 0.0241, 0.0204, 0.0176, 0.0227], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0006, 0.0005, 0.0005, 0.0004, 0.0005], device='cuda:0') 2023-05-15 16:00:58,510 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.886e+02 2.773e+02 3.289e+02 4.075e+02 7.904e+02, threshold=6.577e+02, percent-clipped=1.0 2023-05-15 16:00:58,529 INFO [finetune.py:992] (0/2) Epoch 1, batch 2500, loss[loss=0.1386, simple_loss=0.2257, pruned_loss=0.02578, over 12288.00 frames. ], tot_loss[loss=0.1673, simple_loss=0.2548, pruned_loss=0.03913, over 2385655.10 frames. ], batch size: 28, lr: 5.00e-03, grad_scale: 16.0 2023-05-15 16:01:29,213 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.64 vs. limit=2.0 2023-05-15 16:01:33,096 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.39 vs. limit=2.0 2023-05-15 16:01:35,494 INFO [finetune.py:992] (0/2) Epoch 1, batch 2550, loss[loss=0.1489, simple_loss=0.2467, pruned_loss=0.02554, over 12022.00 frames. ], tot_loss[loss=0.1669, simple_loss=0.2546, pruned_loss=0.03907, over 2388792.34 frames. ], batch size: 31, lr: 5.00e-03, grad_scale: 16.0 2023-05-15 16:01:54,649 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=102576.0, num_to_drop=1, layers_to_drop={1} 2023-05-15 16:02:12,851 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.104e+02 3.014e+02 3.479e+02 4.250e+02 9.382e+02, threshold=6.957e+02, percent-clipped=5.0 2023-05-15 16:02:12,870 INFO [finetune.py:992] (0/2) Epoch 1, batch 2600, loss[loss=0.2018, simple_loss=0.3005, pruned_loss=0.05156, over 12345.00 frames. ], tot_loss[loss=0.1679, simple_loss=0.2559, pruned_loss=0.0395, over 2387224.01 frames. ], batch size: 35, lr: 5.00e-03, grad_scale: 16.0 2023-05-15 16:02:29,655 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=102624.0, num_to_drop=1, layers_to_drop={0} 2023-05-15 16:02:39,600 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.8738, 4.5574, 4.8411, 4.3462, 4.5764, 4.3672, 4.8729, 4.4557], device='cuda:0'), covar=tensor([0.0253, 0.0344, 0.0286, 0.0226, 0.0275, 0.0310, 0.0196, 0.0517], device='cuda:0'), in_proj_covar=tensor([0.0241, 0.0256, 0.0269, 0.0245, 0.0244, 0.0243, 0.0222, 0.0206], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0005, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-15 16:02:43,252 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=102643.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 16:02:48,809 INFO [finetune.py:992] (0/2) Epoch 1, batch 2650, loss[loss=0.1835, simple_loss=0.2713, pruned_loss=0.04789, over 12120.00 frames. ], tot_loss[loss=0.1679, simple_loss=0.2561, pruned_loss=0.03953, over 2393822.88 frames. ], batch size: 38, lr: 5.00e-03, grad_scale: 16.0 2023-05-15 16:03:11,954 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.89 vs. limit=2.0 2023-05-15 16:03:18,666 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=102691.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 16:03:25,830 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.077e+02 2.634e+02 3.104e+02 3.917e+02 7.520e+02, threshold=6.208e+02, percent-clipped=2.0 2023-05-15 16:03:25,849 INFO [finetune.py:992] (0/2) Epoch 1, batch 2700, loss[loss=0.1535, simple_loss=0.25, pruned_loss=0.02846, over 12296.00 frames. ], tot_loss[loss=0.1671, simple_loss=0.2556, pruned_loss=0.03904, over 2397613.58 frames. ], batch size: 33, lr: 5.00e-03, grad_scale: 16.0 2023-05-15 16:03:57,374 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([6.0414, 5.9653, 5.8128, 5.1956, 5.1902, 5.9850, 5.5364, 5.3517], device='cuda:0'), covar=tensor([0.0658, 0.0915, 0.0657, 0.1379, 0.0648, 0.0620, 0.1457, 0.0955], device='cuda:0'), in_proj_covar=tensor([0.0603, 0.0537, 0.0480, 0.0608, 0.0389, 0.0689, 0.0760, 0.0546], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], device='cuda:0') 2023-05-15 16:04:02,826 INFO [finetune.py:992] (0/2) Epoch 1, batch 2750, loss[loss=0.1692, simple_loss=0.2605, pruned_loss=0.03899, over 11628.00 frames. ], tot_loss[loss=0.1671, simple_loss=0.2553, pruned_loss=0.03926, over 2396733.17 frames. ], batch size: 48, lr: 5.00e-03, grad_scale: 16.0 2023-05-15 16:04:39,868 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.882e+02 2.779e+02 3.278e+02 3.944e+02 1.142e+03, threshold=6.557e+02, percent-clipped=3.0 2023-05-15 16:04:39,886 INFO [finetune.py:992] (0/2) Epoch 1, batch 2800, loss[loss=0.2079, simple_loss=0.2856, pruned_loss=0.06511, over 7651.00 frames. ], tot_loss[loss=0.1675, simple_loss=0.2558, pruned_loss=0.0394, over 2389694.71 frames. ], batch size: 97, lr: 5.00e-03, grad_scale: 16.0 2023-05-15 16:05:07,700 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=4.54 vs. limit=5.0 2023-05-15 16:05:16,518 INFO [finetune.py:992] (0/2) Epoch 1, batch 2850, loss[loss=0.1687, simple_loss=0.2584, pruned_loss=0.03947, over 12265.00 frames. ], tot_loss[loss=0.1687, simple_loss=0.2572, pruned_loss=0.04001, over 2386015.13 frames. ], batch size: 37, lr: 5.00e-03, grad_scale: 16.0 2023-05-15 16:05:16,718 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=102851.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 16:05:43,577 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=102887.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 16:05:53,845 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.152e+02 2.894e+02 3.348e+02 4.267e+02 1.176e+03, threshold=6.696e+02, percent-clipped=6.0 2023-05-15 16:05:53,864 INFO [finetune.py:992] (0/2) Epoch 1, batch 2900, loss[loss=0.1609, simple_loss=0.2533, pruned_loss=0.0342, over 12097.00 frames. ], tot_loss[loss=0.1689, simple_loss=0.2573, pruned_loss=0.04014, over 2381161.59 frames. ], batch size: 32, lr: 5.00e-03, grad_scale: 16.0 2023-05-15 16:06:01,928 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=102912.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 16:06:15,097 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.9333, 2.8822, 4.6793, 4.8716, 3.1600, 2.9823, 3.1370, 2.2069], device='cuda:0'), covar=tensor([0.1302, 0.2681, 0.0408, 0.0349, 0.1074, 0.1771, 0.2310, 0.3581], device='cuda:0'), in_proj_covar=tensor([0.0287, 0.0384, 0.0271, 0.0296, 0.0258, 0.0286, 0.0365, 0.0356], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-15 16:06:28,063 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=102948.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 16:06:30,073 INFO [finetune.py:992] (0/2) Epoch 1, batch 2950, loss[loss=0.1346, simple_loss=0.2212, pruned_loss=0.024, over 11782.00 frames. ], tot_loss[loss=0.1684, simple_loss=0.2565, pruned_loss=0.0401, over 2378331.88 frames. ], batch size: 26, lr: 5.00e-03, grad_scale: 16.0 2023-05-15 16:07:07,072 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.860e+02 2.662e+02 3.202e+02 3.925e+02 7.031e+02, threshold=6.403e+02, percent-clipped=1.0 2023-05-15 16:07:07,101 INFO [finetune.py:992] (0/2) Epoch 1, batch 3000, loss[loss=0.1671, simple_loss=0.2595, pruned_loss=0.03734, over 12289.00 frames. ], tot_loss[loss=0.1691, simple_loss=0.2574, pruned_loss=0.0403, over 2379555.23 frames. ], batch size: 34, lr: 5.00e-03, grad_scale: 16.0 2023-05-15 16:07:07,102 INFO [finetune.py:1017] (0/2) Computing validation loss 2023-05-15 16:07:25,014 INFO [finetune.py:1026] (0/2) Epoch 1, validation: loss=0.3762, simple_loss=0.4347, pruned_loss=0.1589, over 1020973.00 frames. 2023-05-15 16:07:25,014 INFO [finetune.py:1027] (0/2) Maximum memory allocated so far is 12608MB 2023-05-15 16:08:00,961 INFO [finetune.py:992] (0/2) Epoch 1, batch 3050, loss[loss=0.1841, simple_loss=0.2723, pruned_loss=0.0479, over 12038.00 frames. ], tot_loss[loss=0.1694, simple_loss=0.2577, pruned_loss=0.04047, over 2364443.77 frames. ], batch size: 42, lr: 5.00e-03, grad_scale: 16.0 2023-05-15 16:08:36,015 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=103098.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 16:08:38,000 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.002e+02 2.894e+02 3.540e+02 4.389e+02 8.495e+02, threshold=7.080e+02, percent-clipped=4.0 2023-05-15 16:08:38,030 INFO [finetune.py:992] (0/2) Epoch 1, batch 3100, loss[loss=0.1554, simple_loss=0.249, pruned_loss=0.03094, over 12143.00 frames. ], tot_loss[loss=0.1692, simple_loss=0.2572, pruned_loss=0.0406, over 2368655.37 frames. ], batch size: 38, lr: 5.00e-03, grad_scale: 16.0 2023-05-15 16:09:01,337 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=103132.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 16:09:15,008 INFO [finetune.py:992] (0/2) Epoch 1, batch 3150, loss[loss=0.2117, simple_loss=0.298, pruned_loss=0.06268, over 8091.00 frames. ], tot_loss[loss=0.17, simple_loss=0.2578, pruned_loss=0.04107, over 2357025.25 frames. ], batch size: 99, lr: 5.00e-03, grad_scale: 16.0 2023-05-15 16:09:21,071 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=103159.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 16:09:31,209 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.9060, 4.9306, 4.8298, 4.9640, 4.3793, 4.9937, 4.9670, 5.2040], device='cuda:0'), covar=tensor([0.0196, 0.0124, 0.0156, 0.0220, 0.0825, 0.0189, 0.0128, 0.0148], device='cuda:0'), in_proj_covar=tensor([0.0189, 0.0192, 0.0189, 0.0237, 0.0241, 0.0204, 0.0176, 0.0227], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0006, 0.0005, 0.0005, 0.0004, 0.0005], device='cuda:0') 2023-05-15 16:09:45,842 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=103193.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 16:09:51,646 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.940e+02 2.858e+02 3.403e+02 4.168e+02 1.314e+03, threshold=6.807e+02, percent-clipped=6.0 2023-05-15 16:09:51,675 INFO [finetune.py:992] (0/2) Epoch 1, batch 3200, loss[loss=0.1479, simple_loss=0.2246, pruned_loss=0.03559, over 12292.00 frames. ], tot_loss[loss=0.1705, simple_loss=0.258, pruned_loss=0.04146, over 2348796.49 frames. ], batch size: 28, lr: 5.00e-03, grad_scale: 16.0 2023-05-15 16:09:56,058 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=103207.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 16:09:57,543 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.7954, 5.7507, 5.5333, 5.2040, 5.1199, 5.7273, 5.2860, 5.1245], device='cuda:0'), covar=tensor([0.0687, 0.0847, 0.0611, 0.1391, 0.0685, 0.0707, 0.1537, 0.0971], device='cuda:0'), in_proj_covar=tensor([0.0590, 0.0529, 0.0474, 0.0598, 0.0385, 0.0679, 0.0747, 0.0535], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], device='cuda:0') 2023-05-15 16:10:13,436 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.3746, 3.4684, 3.2083, 3.1190, 2.7870, 2.8336, 3.3630, 2.1172], device='cuda:0'), covar=tensor([0.0318, 0.0127, 0.0120, 0.0136, 0.0310, 0.0243, 0.0114, 0.0408], device='cuda:0'), in_proj_covar=tensor([0.0186, 0.0160, 0.0152, 0.0181, 0.0203, 0.0197, 0.0160, 0.0197], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-15 16:10:22,757 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=103243.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 16:10:28,474 INFO [finetune.py:992] (0/2) Epoch 1, batch 3250, loss[loss=0.1931, simple_loss=0.2813, pruned_loss=0.05242, over 12021.00 frames. ], tot_loss[loss=0.1703, simple_loss=0.2582, pruned_loss=0.0412, over 2351593.84 frames. ], batch size: 40, lr: 5.00e-03, grad_scale: 16.0 2023-05-15 16:11:05,081 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.916e+02 2.862e+02 3.260e+02 4.103e+02 6.638e+02, threshold=6.520e+02, percent-clipped=1.0 2023-05-15 16:11:05,099 INFO [finetune.py:992] (0/2) Epoch 1, batch 3300, loss[loss=0.2078, simple_loss=0.281, pruned_loss=0.06728, over 7790.00 frames. ], tot_loss[loss=0.1706, simple_loss=0.2583, pruned_loss=0.04144, over 2352398.97 frames. ], batch size: 99, lr: 5.00e-03, grad_scale: 16.0 2023-05-15 16:11:10,890 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.3785, 4.8039, 3.0384, 2.7755, 4.1230, 2.6064, 4.0928, 3.5195], device='cuda:0'), covar=tensor([0.0666, 0.0454, 0.1020, 0.1336, 0.0265, 0.1181, 0.0436, 0.0672], device='cuda:0'), in_proj_covar=tensor([0.0194, 0.0256, 0.0175, 0.0199, 0.0141, 0.0178, 0.0199, 0.0174], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-15 16:11:27,566 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=2.51 vs. limit=5.0 2023-05-15 16:11:30,764 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.92 vs. limit=2.0 2023-05-15 16:11:41,102 INFO [finetune.py:992] (0/2) Epoch 1, batch 3350, loss[loss=0.1445, simple_loss=0.2236, pruned_loss=0.03266, over 12166.00 frames. ], tot_loss[loss=0.1712, simple_loss=0.259, pruned_loss=0.04164, over 2362154.30 frames. ], batch size: 29, lr: 5.00e-03, grad_scale: 8.0 2023-05-15 16:11:52,675 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=103366.0, num_to_drop=1, layers_to_drop={0} 2023-05-15 16:12:18,310 INFO [finetune.py:992] (0/2) Epoch 1, batch 3400, loss[loss=0.1633, simple_loss=0.2419, pruned_loss=0.04234, over 12127.00 frames. ], tot_loss[loss=0.1708, simple_loss=0.2587, pruned_loss=0.04149, over 2370608.76 frames. ], batch size: 30, lr: 5.00e-03, grad_scale: 8.0 2023-05-15 16:12:19,026 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.040e+02 2.898e+02 3.310e+02 3.849e+02 7.246e+02, threshold=6.620e+02, percent-clipped=1.0 2023-05-15 16:12:33,412 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.31 vs. limit=2.0 2023-05-15 16:12:38,105 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=103427.0, num_to_drop=1, layers_to_drop={2} 2023-05-15 16:12:53,422 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.4301, 3.5381, 3.1846, 3.1804, 2.8401, 2.8186, 3.4360, 2.1778], device='cuda:0'), covar=tensor([0.0312, 0.0101, 0.0134, 0.0152, 0.0320, 0.0260, 0.0108, 0.0410], device='cuda:0'), in_proj_covar=tensor([0.0187, 0.0161, 0.0152, 0.0182, 0.0204, 0.0198, 0.0160, 0.0199], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-15 16:12:55,372 INFO [finetune.py:992] (0/2) Epoch 1, batch 3450, loss[loss=0.148, simple_loss=0.2332, pruned_loss=0.03136, over 12011.00 frames. ], tot_loss[loss=0.1704, simple_loss=0.2583, pruned_loss=0.04124, over 2359799.12 frames. ], batch size: 28, lr: 5.00e-03, grad_scale: 8.0 2023-05-15 16:12:57,620 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=103454.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 16:13:22,118 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=103488.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 16:13:31,325 INFO [finetune.py:992] (0/2) Epoch 1, batch 3500, loss[loss=0.1953, simple_loss=0.2789, pruned_loss=0.0558, over 10284.00 frames. ], tot_loss[loss=0.1697, simple_loss=0.2577, pruned_loss=0.04087, over 2367355.97 frames. ], batch size: 68, lr: 5.00e-03, grad_scale: 8.0 2023-05-15 16:13:32,038 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.959e+02 2.723e+02 3.181e+02 3.899e+02 6.414e+02, threshold=6.363e+02, percent-clipped=0.0 2023-05-15 16:13:35,166 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.3941, 5.1730, 5.2390, 5.3217, 4.9245, 5.0085, 4.7521, 5.2708], device='cuda:0'), covar=tensor([0.0580, 0.0555, 0.0747, 0.0531, 0.1956, 0.1185, 0.0631, 0.0847], device='cuda:0'), in_proj_covar=tensor([0.0512, 0.0654, 0.0553, 0.0624, 0.0826, 0.0729, 0.0533, 0.0478], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0005, 0.0004, 0.0005, 0.0006, 0.0005, 0.0004, 0.0004], device='cuda:0') 2023-05-15 16:13:35,857 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=103507.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 16:13:36,733 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.9062, 3.4536, 5.2320, 2.7347, 2.9027, 4.0038, 3.3768, 4.0779], device='cuda:0'), covar=tensor([0.0367, 0.1023, 0.0209, 0.1186, 0.1899, 0.1194, 0.1279, 0.0912], device='cuda:0'), in_proj_covar=tensor([0.0237, 0.0233, 0.0245, 0.0185, 0.0243, 0.0290, 0.0231, 0.0258], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-15 16:14:02,735 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=103543.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 16:14:09,214 INFO [finetune.py:992] (0/2) Epoch 1, batch 3550, loss[loss=0.1477, simple_loss=0.2279, pruned_loss=0.03374, over 12372.00 frames. ], tot_loss[loss=0.1691, simple_loss=0.2571, pruned_loss=0.04053, over 2369559.79 frames. ], batch size: 30, lr: 5.00e-03, grad_scale: 8.0 2023-05-15 16:14:12,124 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=103555.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 16:14:26,927 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.5731, 5.3747, 5.4398, 5.5093, 5.1008, 5.1150, 4.9755, 5.4343], device='cuda:0'), covar=tensor([0.0573, 0.0507, 0.0591, 0.0532, 0.1574, 0.1283, 0.0537, 0.0759], device='cuda:0'), in_proj_covar=tensor([0.0513, 0.0655, 0.0553, 0.0623, 0.0825, 0.0731, 0.0535, 0.0478], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0005, 0.0004, 0.0005, 0.0006, 0.0005, 0.0004, 0.0004], device='cuda:0') 2023-05-15 16:14:38,325 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=103591.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 16:14:45,737 INFO [finetune.py:992] (0/2) Epoch 1, batch 3600, loss[loss=0.198, simple_loss=0.289, pruned_loss=0.05349, over 12278.00 frames. ], tot_loss[loss=0.1691, simple_loss=0.2571, pruned_loss=0.04055, over 2374760.43 frames. ], batch size: 37, lr: 5.00e-03, grad_scale: 8.0 2023-05-15 16:14:46,424 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.850e+02 2.881e+02 3.349e+02 4.086e+02 6.568e+02, threshold=6.698e+02, percent-clipped=1.0 2023-05-15 16:15:21,469 INFO [finetune.py:992] (0/2) Epoch 1, batch 3650, loss[loss=0.153, simple_loss=0.2448, pruned_loss=0.03062, over 11628.00 frames. ], tot_loss[loss=0.1693, simple_loss=0.2575, pruned_loss=0.04057, over 2378090.08 frames. ], batch size: 48, lr: 5.00e-03, grad_scale: 8.0 2023-05-15 16:15:59,305 INFO [finetune.py:992] (0/2) Epoch 1, batch 3700, loss[loss=0.1602, simple_loss=0.2544, pruned_loss=0.03304, over 12276.00 frames. ], tot_loss[loss=0.1692, simple_loss=0.2572, pruned_loss=0.04055, over 2379610.35 frames. ], batch size: 37, lr: 5.00e-03, grad_scale: 8.0 2023-05-15 16:16:00,019 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.124e+02 2.853e+02 3.301e+02 3.898e+02 6.815e+02, threshold=6.603e+02, percent-clipped=1.0 2023-05-15 16:16:14,390 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=103722.0, num_to_drop=1, layers_to_drop={3} 2023-05-15 16:16:27,698 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=103740.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 16:16:35,398 INFO [finetune.py:992] (0/2) Epoch 1, batch 3750, loss[loss=0.1553, simple_loss=0.2336, pruned_loss=0.03857, over 11821.00 frames. ], tot_loss[loss=0.1691, simple_loss=0.2571, pruned_loss=0.04052, over 2374998.36 frames. ], batch size: 26, lr: 5.00e-03, grad_scale: 8.0 2023-05-15 16:16:37,660 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=103754.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 16:16:45,522 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=103765.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 16:17:02,107 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=103788.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 16:17:12,314 INFO [finetune.py:992] (0/2) Epoch 1, batch 3800, loss[loss=0.1533, simple_loss=0.2492, pruned_loss=0.02871, over 12303.00 frames. ], tot_loss[loss=0.1684, simple_loss=0.2563, pruned_loss=0.04021, over 2378278.42 frames. ], batch size: 34, lr: 5.00e-03, grad_scale: 8.0 2023-05-15 16:17:12,530 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=103801.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 16:17:13,032 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.909e+02 2.959e+02 3.693e+02 4.385e+02 1.112e+03, threshold=7.387e+02, percent-clipped=2.0 2023-05-15 16:17:13,135 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=103802.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 16:17:30,361 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=103826.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 16:17:37,450 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=103836.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 16:17:48,646 INFO [finetune.py:992] (0/2) Epoch 1, batch 3850, loss[loss=0.1886, simple_loss=0.2769, pruned_loss=0.05014, over 11207.00 frames. ], tot_loss[loss=0.1695, simple_loss=0.2574, pruned_loss=0.04074, over 2385543.79 frames. ], batch size: 55, lr: 5.00e-03, grad_scale: 8.0 2023-05-15 16:18:07,107 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=103876.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 16:18:24,765 INFO [finetune.py:992] (0/2) Epoch 1, batch 3900, loss[loss=0.1925, simple_loss=0.2777, pruned_loss=0.05369, over 12122.00 frames. ], tot_loss[loss=0.1687, simple_loss=0.2567, pruned_loss=0.04042, over 2376181.15 frames. ], batch size: 38, lr: 5.00e-03, grad_scale: 8.0 2023-05-15 16:18:25,492 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.999e+02 2.744e+02 3.382e+02 4.503e+02 9.689e+02, threshold=6.763e+02, percent-clipped=2.0 2023-05-15 16:18:31,591 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.52 vs. limit=2.0 2023-05-15 16:18:50,819 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=103937.0, num_to_drop=1, layers_to_drop={1} 2023-05-15 16:19:01,525 INFO [finetune.py:992] (0/2) Epoch 1, batch 3950, loss[loss=0.1416, simple_loss=0.2234, pruned_loss=0.02994, over 11781.00 frames. ], tot_loss[loss=0.1688, simple_loss=0.257, pruned_loss=0.04024, over 2381546.19 frames. ], batch size: 26, lr: 5.00e-03, grad_scale: 8.0 2023-05-15 16:19:29,850 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=103990.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 16:19:35,517 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=103997.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 16:19:37,748 INFO [checkpoint.py:75] (0/2) Saving checkpoint to pruned_transducer_stateless7/exp_giga_finetune/checkpoint-4000.pt 2023-05-15 16:19:41,612 INFO [finetune.py:992] (0/2) Epoch 1, batch 4000, loss[loss=0.1538, simple_loss=0.2353, pruned_loss=0.03617, over 12191.00 frames. ], tot_loss[loss=0.1686, simple_loss=0.2571, pruned_loss=0.04005, over 2385023.49 frames. ], batch size: 29, lr: 5.00e-03, grad_scale: 8.0 2023-05-15 16:19:42,212 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.022e+02 2.918e+02 3.442e+02 3.941e+02 5.823e+02, threshold=6.884e+02, percent-clipped=0.0 2023-05-15 16:19:56,594 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=104022.0, num_to_drop=1, layers_to_drop={1} 2023-05-15 16:19:56,691 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.4908, 4.7627, 2.7857, 2.1231, 4.2823, 2.4415, 4.0945, 3.2221], device='cuda:0'), covar=tensor([0.0480, 0.0409, 0.0971, 0.1766, 0.0239, 0.1219, 0.0323, 0.0796], device='cuda:0'), in_proj_covar=tensor([0.0192, 0.0251, 0.0174, 0.0198, 0.0139, 0.0176, 0.0195, 0.0172], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-15 16:20:16,028 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=104049.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 16:20:17,281 INFO [finetune.py:992] (0/2) Epoch 1, batch 4050, loss[loss=0.1282, simple_loss=0.2158, pruned_loss=0.02029, over 12271.00 frames. ], tot_loss[loss=0.1692, simple_loss=0.2575, pruned_loss=0.04041, over 2391747.41 frames. ], batch size: 32, lr: 5.00e-03, grad_scale: 8.0 2023-05-15 16:20:17,534 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=104051.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 16:20:22,658 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=104058.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 16:20:31,158 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=104070.0, num_to_drop=1, layers_to_drop={0} 2023-05-15 16:20:50,338 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=104096.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 16:20:53,830 INFO [finetune.py:992] (0/2) Epoch 1, batch 4100, loss[loss=0.2084, simple_loss=0.2925, pruned_loss=0.06217, over 12037.00 frames. ], tot_loss[loss=0.17, simple_loss=0.2581, pruned_loss=0.04093, over 2378220.89 frames. ], batch size: 40, lr: 5.00e-03, grad_scale: 8.0 2023-05-15 16:20:54,467 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.086e+02 2.927e+02 3.555e+02 4.282e+02 8.168e+02, threshold=7.110e+02, percent-clipped=2.0 2023-05-15 16:21:00,264 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=104110.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 16:21:08,176 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=104121.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 16:21:14,105 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.2905, 2.3110, 3.6688, 4.2352, 3.9171, 4.2434, 3.8262, 2.7649], device='cuda:0'), covar=tensor([0.0034, 0.0445, 0.0114, 0.0038, 0.0088, 0.0060, 0.0087, 0.0379], device='cuda:0'), in_proj_covar=tensor([0.0081, 0.0124, 0.0108, 0.0078, 0.0104, 0.0113, 0.0089, 0.0141], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-15 16:21:25,079 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.0091, 5.0249, 4.8724, 4.9701, 4.3939, 5.0119, 5.0450, 5.2415], device='cuda:0'), covar=tensor([0.0216, 0.0127, 0.0173, 0.0262, 0.0823, 0.0248, 0.0135, 0.0153], device='cuda:0'), in_proj_covar=tensor([0.0186, 0.0190, 0.0187, 0.0236, 0.0239, 0.0202, 0.0174, 0.0224], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0005, 0.0005, 0.0005, 0.0004, 0.0005], device='cuda:0') 2023-05-15 16:21:27,211 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.4237, 5.2367, 5.3187, 5.4342, 4.9543, 5.1018, 4.8911, 5.3329], device='cuda:0'), covar=tensor([0.0645, 0.0583, 0.0669, 0.0550, 0.2049, 0.1196, 0.0541, 0.0915], device='cuda:0'), in_proj_covar=tensor([0.0519, 0.0658, 0.0562, 0.0628, 0.0829, 0.0742, 0.0539, 0.0483], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0005, 0.0004, 0.0005, 0.0006, 0.0005, 0.0004, 0.0004], device='cuda:0') 2023-05-15 16:21:30,610 INFO [finetune.py:992] (0/2) Epoch 1, batch 4150, loss[loss=0.1635, simple_loss=0.2578, pruned_loss=0.03455, over 12355.00 frames. ], tot_loss[loss=0.1698, simple_loss=0.2578, pruned_loss=0.04093, over 2379613.63 frames. ], batch size: 36, lr: 5.00e-03, grad_scale: 8.0 2023-05-15 16:22:02,429 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.3666, 4.7620, 2.8127, 2.7093, 3.9846, 2.5337, 4.0389, 3.4032], device='cuda:0'), covar=tensor([0.0605, 0.0412, 0.1103, 0.1382, 0.0294, 0.1239, 0.0395, 0.0683], device='cuda:0'), in_proj_covar=tensor([0.0192, 0.0250, 0.0173, 0.0197, 0.0139, 0.0176, 0.0195, 0.0172], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-15 16:22:06,913 INFO [finetune.py:992] (0/2) Epoch 1, batch 4200, loss[loss=0.1679, simple_loss=0.2541, pruned_loss=0.04092, over 12180.00 frames. ], tot_loss[loss=0.1697, simple_loss=0.2581, pruned_loss=0.04065, over 2382863.37 frames. ], batch size: 35, lr: 5.00e-03, grad_scale: 8.0 2023-05-15 16:22:07,612 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.974e+02 2.892e+02 3.381e+02 4.316e+02 9.943e+02, threshold=6.762e+02, percent-clipped=2.0 2023-05-15 16:22:28,986 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=104232.0, num_to_drop=1, layers_to_drop={0} 2023-05-15 16:22:43,272 INFO [finetune.py:992] (0/2) Epoch 1, batch 4250, loss[loss=0.1594, simple_loss=0.2402, pruned_loss=0.03927, over 12332.00 frames. ], tot_loss[loss=0.17, simple_loss=0.2581, pruned_loss=0.04096, over 2369462.98 frames. ], batch size: 30, lr: 5.00e-03, grad_scale: 8.0 2023-05-15 16:23:19,532 INFO [finetune.py:992] (0/2) Epoch 1, batch 4300, loss[loss=0.1872, simple_loss=0.2859, pruned_loss=0.04427, over 12059.00 frames. ], tot_loss[loss=0.17, simple_loss=0.2579, pruned_loss=0.04109, over 2366175.79 frames. ], batch size: 40, lr: 5.00e-03, grad_scale: 8.0 2023-05-15 16:23:20,204 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.944e+02 2.993e+02 3.570e+02 4.476e+02 1.466e+03, threshold=7.140e+02, percent-clipped=3.0 2023-05-15 16:23:43,979 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-05-15 16:23:52,407 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=104346.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 16:23:55,865 INFO [finetune.py:992] (0/2) Epoch 1, batch 4350, loss[loss=0.1887, simple_loss=0.2773, pruned_loss=0.05003, over 11770.00 frames. ], tot_loss[loss=0.1705, simple_loss=0.2584, pruned_loss=0.04127, over 2370677.26 frames. ], batch size: 44, lr: 5.00e-03, grad_scale: 8.0 2023-05-15 16:23:57,354 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=104353.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 16:24:13,267 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.1902, 5.1331, 5.0173, 5.1337, 4.2359, 5.1649, 5.1520, 5.2757], device='cuda:0'), covar=tensor([0.0225, 0.0136, 0.0182, 0.0249, 0.1139, 0.0221, 0.0147, 0.0185], device='cuda:0'), in_proj_covar=tensor([0.0187, 0.0192, 0.0188, 0.0237, 0.0241, 0.0203, 0.0175, 0.0226], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0005, 0.0005, 0.0005, 0.0004, 0.0005], device='cuda:0') 2023-05-15 16:24:17,999 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.43 vs. limit=2.0 2023-05-15 16:24:19,949 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.0110, 4.0156, 4.0036, 4.4978, 2.9317, 4.0237, 2.4020, 4.0594], device='cuda:0'), covar=tensor([0.1559, 0.0631, 0.0915, 0.0564, 0.1088, 0.0522, 0.1846, 0.1375], device='cuda:0'), in_proj_covar=tensor([0.0223, 0.0266, 0.0306, 0.0368, 0.0247, 0.0241, 0.0263, 0.0405], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-15 16:24:24,177 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.9988, 2.3828, 3.6186, 2.9477, 3.4302, 3.0865, 2.4289, 3.5522], device='cuda:0'), covar=tensor([0.0109, 0.0301, 0.0116, 0.0197, 0.0104, 0.0156, 0.0280, 0.0087], device='cuda:0'), in_proj_covar=tensor([0.0158, 0.0190, 0.0168, 0.0172, 0.0192, 0.0151, 0.0179, 0.0160], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-15 16:24:27,021 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.3762, 5.0189, 5.3078, 4.5865, 4.9864, 4.7802, 5.3087, 5.0431], device='cuda:0'), covar=tensor([0.0270, 0.0345, 0.0293, 0.0263, 0.0295, 0.0296, 0.0250, 0.0270], device='cuda:0'), in_proj_covar=tensor([0.0240, 0.0252, 0.0264, 0.0242, 0.0240, 0.0238, 0.0220, 0.0204], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0005, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-15 16:24:29,154 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=104396.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 16:24:32,920 INFO [finetune.py:992] (0/2) Epoch 1, batch 4400, loss[loss=0.1711, simple_loss=0.2558, pruned_loss=0.04318, over 12246.00 frames. ], tot_loss[loss=0.1703, simple_loss=0.2585, pruned_loss=0.04104, over 2373018.41 frames. ], batch size: 32, lr: 5.00e-03, grad_scale: 8.0 2023-05-15 16:24:33,595 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.787e+02 2.824e+02 3.318e+02 3.804e+02 6.419e+02, threshold=6.637e+02, percent-clipped=0.0 2023-05-15 16:24:35,790 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=104405.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 16:24:47,371 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=104421.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 16:25:04,028 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.4909, 3.5566, 3.3154, 3.2307, 2.9004, 2.6856, 3.4989, 2.4734], device='cuda:0'), covar=tensor([0.0294, 0.0125, 0.0123, 0.0136, 0.0290, 0.0265, 0.0092, 0.0344], device='cuda:0'), in_proj_covar=tensor([0.0189, 0.0164, 0.0156, 0.0183, 0.0207, 0.0204, 0.0161, 0.0203], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-15 16:25:04,580 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=104444.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 16:25:09,479 INFO [finetune.py:992] (0/2) Epoch 1, batch 4450, loss[loss=0.1628, simple_loss=0.2498, pruned_loss=0.03788, over 12330.00 frames. ], tot_loss[loss=0.1711, simple_loss=0.2593, pruned_loss=0.04144, over 2370170.24 frames. ], batch size: 31, lr: 5.00e-03, grad_scale: 8.0 2023-05-15 16:25:22,513 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=104469.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 16:25:45,043 INFO [finetune.py:992] (0/2) Epoch 1, batch 4500, loss[loss=0.1547, simple_loss=0.2494, pruned_loss=0.03004, over 12144.00 frames. ], tot_loss[loss=0.1723, simple_loss=0.2605, pruned_loss=0.04204, over 2371763.20 frames. ], batch size: 36, lr: 5.00e-03, grad_scale: 8.0 2023-05-15 16:25:45,760 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.791e+02 2.830e+02 3.188e+02 3.845e+02 1.100e+03, threshold=6.377e+02, percent-clipped=3.0 2023-05-15 16:26:07,359 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=104532.0, num_to_drop=1, layers_to_drop={2} 2023-05-15 16:26:21,438 INFO [finetune.py:992] (0/2) Epoch 1, batch 4550, loss[loss=0.222, simple_loss=0.292, pruned_loss=0.07605, over 8179.00 frames. ], tot_loss[loss=0.1723, simple_loss=0.2603, pruned_loss=0.04216, over 2369712.83 frames. ], batch size: 98, lr: 5.00e-03, grad_scale: 8.0 2023-05-15 16:26:42,705 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=104580.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 16:26:58,131 INFO [finetune.py:992] (0/2) Epoch 1, batch 4600, loss[loss=0.1696, simple_loss=0.251, pruned_loss=0.04405, over 12335.00 frames. ], tot_loss[loss=0.1723, simple_loss=0.2603, pruned_loss=0.04218, over 2367275.26 frames. ], batch size: 30, lr: 5.00e-03, grad_scale: 8.0 2023-05-15 16:26:58,776 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.763e+02 2.795e+02 3.549e+02 4.331e+02 1.029e+03, threshold=7.098e+02, percent-clipped=4.0 2023-05-15 16:27:30,549 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=104646.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 16:27:34,055 INFO [finetune.py:992] (0/2) Epoch 1, batch 4650, loss[loss=0.1741, simple_loss=0.2683, pruned_loss=0.03997, over 12349.00 frames. ], tot_loss[loss=0.1719, simple_loss=0.2598, pruned_loss=0.04202, over 2367833.06 frames. ], batch size: 36, lr: 5.00e-03, grad_scale: 8.0 2023-05-15 16:27:35,669 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=104653.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 16:27:41,082 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.87 vs. limit=2.0 2023-05-15 16:28:05,976 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=104694.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 16:28:10,894 INFO [finetune.py:992] (0/2) Epoch 1, batch 4700, loss[loss=0.1733, simple_loss=0.2678, pruned_loss=0.03938, over 12283.00 frames. ], tot_loss[loss=0.1722, simple_loss=0.2603, pruned_loss=0.04202, over 2369163.11 frames. ], batch size: 34, lr: 5.00e-03, grad_scale: 8.0 2023-05-15 16:28:10,954 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=104701.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 16:28:11,571 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.859e+02 2.857e+02 3.257e+02 4.088e+02 7.241e+02, threshold=6.513e+02, percent-clipped=1.0 2023-05-15 16:28:13,612 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.78 vs. limit=2.0 2023-05-15 16:28:13,944 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=104705.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 16:28:30,331 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.0184, 5.9018, 5.4071, 5.3486, 5.9490, 5.2391, 5.6241, 5.5059], device='cuda:0'), covar=tensor([0.1469, 0.0917, 0.0953, 0.1897, 0.1006, 0.2280, 0.1591, 0.0987], device='cuda:0'), in_proj_covar=tensor([0.0338, 0.0477, 0.0369, 0.0421, 0.0451, 0.0418, 0.0387, 0.0369], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-15 16:28:47,412 INFO [finetune.py:992] (0/2) Epoch 1, batch 4750, loss[loss=0.1484, simple_loss=0.2344, pruned_loss=0.03119, over 12270.00 frames. ], tot_loss[loss=0.172, simple_loss=0.26, pruned_loss=0.04194, over 2364849.35 frames. ], batch size: 28, lr: 5.00e-03, grad_scale: 8.0 2023-05-15 16:28:48,913 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=104753.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 16:29:03,473 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.6861, 4.5221, 4.6060, 4.5561, 4.4867, 4.5975, 4.5335, 2.3870], device='cuda:0'), covar=tensor([0.0184, 0.0099, 0.0129, 0.0127, 0.0084, 0.0173, 0.0142, 0.0928], device='cuda:0'), in_proj_covar=tensor([0.0065, 0.0075, 0.0078, 0.0071, 0.0059, 0.0088, 0.0077, 0.0090], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0004], device='cuda:0') 2023-05-15 16:29:06,421 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.15 vs. limit=2.0 2023-05-15 16:29:20,067 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=4.69 vs. limit=5.0 2023-05-15 16:29:23,543 INFO [finetune.py:992] (0/2) Epoch 1, batch 4800, loss[loss=0.1817, simple_loss=0.2668, pruned_loss=0.04828, over 12020.00 frames. ], tot_loss[loss=0.1718, simple_loss=0.2598, pruned_loss=0.04192, over 2373868.12 frames. ], batch size: 40, lr: 5.00e-03, grad_scale: 8.0 2023-05-15 16:29:24,200 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.815e+02 2.849e+02 3.513e+02 4.430e+02 7.695e+02, threshold=7.026e+02, percent-clipped=3.0 2023-05-15 16:29:59,445 INFO [finetune.py:992] (0/2) Epoch 1, batch 4850, loss[loss=0.1963, simple_loss=0.2862, pruned_loss=0.05326, over 11319.00 frames. ], tot_loss[loss=0.172, simple_loss=0.2599, pruned_loss=0.04202, over 2381443.13 frames. ], batch size: 55, lr: 5.00e-03, grad_scale: 8.0 2023-05-15 16:30:35,656 INFO [finetune.py:992] (0/2) Epoch 1, batch 4900, loss[loss=0.1582, simple_loss=0.2537, pruned_loss=0.03139, over 12150.00 frames. ], tot_loss[loss=0.1714, simple_loss=0.2594, pruned_loss=0.04167, over 2392219.31 frames. ], batch size: 36, lr: 5.00e-03, grad_scale: 8.0 2023-05-15 16:30:36,269 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.977e+02 3.072e+02 3.542e+02 4.222e+02 6.583e+02, threshold=7.085e+02, percent-clipped=0.0 2023-05-15 16:31:11,171 INFO [finetune.py:992] (0/2) Epoch 1, batch 4950, loss[loss=0.1632, simple_loss=0.2525, pruned_loss=0.03702, over 12111.00 frames. ], tot_loss[loss=0.1719, simple_loss=0.2597, pruned_loss=0.04206, over 2383496.97 frames. ], batch size: 33, lr: 5.00e-03, grad_scale: 8.0 2023-05-15 16:31:27,254 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.4491, 2.3458, 3.2102, 4.2459, 2.2650, 4.4808, 4.4353, 4.5454], device='cuda:0'), covar=tensor([0.0111, 0.1201, 0.0446, 0.0136, 0.1169, 0.0171, 0.0118, 0.0078], device='cuda:0'), in_proj_covar=tensor([0.0121, 0.0205, 0.0194, 0.0117, 0.0188, 0.0182, 0.0173, 0.0122], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-15 16:31:45,492 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=4.48 vs. limit=5.0 2023-05-15 16:31:48,106 INFO [finetune.py:992] (0/2) Epoch 1, batch 5000, loss[loss=0.1698, simple_loss=0.2558, pruned_loss=0.04189, over 12069.00 frames. ], tot_loss[loss=0.1718, simple_loss=0.2598, pruned_loss=0.04187, over 2382813.40 frames. ], batch size: 32, lr: 5.00e-03, grad_scale: 8.0 2023-05-15 16:31:48,787 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.718e+02 2.772e+02 3.423e+02 4.530e+02 7.566e+02, threshold=6.846e+02, percent-clipped=4.0 2023-05-15 16:31:52,609 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.9330, 4.7695, 4.8543, 4.7949, 4.6802, 4.8615, 4.7795, 2.8029], device='cuda:0'), covar=tensor([0.0161, 0.0075, 0.0105, 0.0100, 0.0066, 0.0128, 0.0103, 0.0767], device='cuda:0'), in_proj_covar=tensor([0.0064, 0.0074, 0.0077, 0.0071, 0.0059, 0.0087, 0.0076, 0.0090], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0004], device='cuda:0') 2023-05-15 16:31:53,317 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.2398, 5.2073, 5.0648, 5.0929, 4.7417, 5.1983, 5.2172, 5.4264], device='cuda:0'), covar=tensor([0.0139, 0.0108, 0.0154, 0.0256, 0.0598, 0.0197, 0.0110, 0.0134], device='cuda:0'), in_proj_covar=tensor([0.0181, 0.0186, 0.0182, 0.0229, 0.0232, 0.0197, 0.0170, 0.0218], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0005, 0.0005, 0.0005, 0.0004, 0.0005], device='cuda:0') 2023-05-15 16:32:24,729 INFO [finetune.py:992] (0/2) Epoch 1, batch 5050, loss[loss=0.1796, simple_loss=0.2695, pruned_loss=0.04485, over 12126.00 frames. ], tot_loss[loss=0.1716, simple_loss=0.2594, pruned_loss=0.0419, over 2377960.28 frames. ], batch size: 39, lr: 5.00e-03, grad_scale: 8.0 2023-05-15 16:32:34,610 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.36 vs. limit=2.0 2023-05-15 16:33:00,457 INFO [finetune.py:992] (0/2) Epoch 1, batch 5100, loss[loss=0.1532, simple_loss=0.2357, pruned_loss=0.03537, over 12089.00 frames. ], tot_loss[loss=0.1717, simple_loss=0.2592, pruned_loss=0.0421, over 2381237.80 frames. ], batch size: 32, lr: 5.00e-03, grad_scale: 8.0 2023-05-15 16:33:01,108 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.885e+02 2.904e+02 3.598e+02 4.298e+02 1.348e+03, threshold=7.197e+02, percent-clipped=3.0 2023-05-15 16:33:36,748 INFO [finetune.py:992] (0/2) Epoch 1, batch 5150, loss[loss=0.1694, simple_loss=0.2599, pruned_loss=0.03945, over 12092.00 frames. ], tot_loss[loss=0.1721, simple_loss=0.2597, pruned_loss=0.04226, over 2377502.87 frames. ], batch size: 32, lr: 5.00e-03, grad_scale: 8.0 2023-05-15 16:33:43,428 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.3534, 4.7439, 3.0427, 2.7859, 3.9746, 2.5984, 4.0653, 3.3614], device='cuda:0'), covar=tensor([0.0642, 0.0506, 0.0934, 0.1331, 0.0291, 0.1204, 0.0430, 0.0724], device='cuda:0'), in_proj_covar=tensor([0.0191, 0.0251, 0.0174, 0.0196, 0.0139, 0.0178, 0.0195, 0.0172], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-15 16:33:56,691 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=4.89 vs. limit=5.0 2023-05-15 16:34:13,371 INFO [finetune.py:992] (0/2) Epoch 1, batch 5200, loss[loss=0.1686, simple_loss=0.2642, pruned_loss=0.03644, over 12339.00 frames. ], tot_loss[loss=0.1731, simple_loss=0.2606, pruned_loss=0.04276, over 2369465.27 frames. ], batch size: 36, lr: 5.00e-03, grad_scale: 8.0 2023-05-15 16:34:14,040 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.131e+02 2.916e+02 3.523e+02 4.295e+02 9.519e+02, threshold=7.045e+02, percent-clipped=2.0 2023-05-15 16:34:23,035 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.96 vs. limit=2.0 2023-05-15 16:34:30,880 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.42 vs. limit=2.0 2023-05-15 16:34:48,634 INFO [finetune.py:992] (0/2) Epoch 1, batch 5250, loss[loss=0.1539, simple_loss=0.2458, pruned_loss=0.03104, over 12274.00 frames. ], tot_loss[loss=0.1731, simple_loss=0.2608, pruned_loss=0.04273, over 2374498.10 frames. ], batch size: 33, lr: 5.00e-03, grad_scale: 8.0 2023-05-15 16:35:04,614 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.9033, 5.8803, 5.6834, 5.1652, 5.0937, 5.7720, 5.3093, 5.1807], device='cuda:0'), covar=tensor([0.0655, 0.0813, 0.0545, 0.1310, 0.0695, 0.0674, 0.1596, 0.0911], device='cuda:0'), in_proj_covar=tensor([0.0583, 0.0516, 0.0471, 0.0589, 0.0377, 0.0671, 0.0732, 0.0530], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0004, 0.0004, 0.0003], device='cuda:0') 2023-05-15 16:35:24,857 INFO [finetune.py:992] (0/2) Epoch 1, batch 5300, loss[loss=0.1755, simple_loss=0.2751, pruned_loss=0.03799, over 12148.00 frames. ], tot_loss[loss=0.1733, simple_loss=0.2609, pruned_loss=0.04282, over 2365768.64 frames. ], batch size: 36, lr: 5.00e-03, grad_scale: 8.0 2023-05-15 16:35:25,568 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.212e+02 2.987e+02 3.418e+02 4.113e+02 6.916e+02, threshold=6.836e+02, percent-clipped=0.0 2023-05-15 16:36:01,029 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.3924, 5.1851, 5.2091, 5.3550, 4.9141, 4.9631, 4.8342, 5.2214], device='cuda:0'), covar=tensor([0.0499, 0.0485, 0.0711, 0.0482, 0.1804, 0.1274, 0.0499, 0.0814], device='cuda:0'), in_proj_covar=tensor([0.0502, 0.0646, 0.0554, 0.0614, 0.0807, 0.0731, 0.0529, 0.0471], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0005, 0.0004, 0.0005, 0.0006, 0.0005, 0.0004, 0.0004], device='cuda:0') 2023-05-15 16:36:01,586 INFO [finetune.py:992] (0/2) Epoch 1, batch 5350, loss[loss=0.1831, simple_loss=0.2705, pruned_loss=0.04782, over 12134.00 frames. ], tot_loss[loss=0.1734, simple_loss=0.2615, pruned_loss=0.04268, over 2363616.29 frames. ], batch size: 36, lr: 5.00e-03, grad_scale: 16.0 2023-05-15 16:36:06,151 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([6.0417, 6.0139, 5.7766, 5.4054, 5.2386, 5.9667, 5.5027, 5.2903], device='cuda:0'), covar=tensor([0.0680, 0.0879, 0.0572, 0.1200, 0.0588, 0.0627, 0.1412, 0.0868], device='cuda:0'), in_proj_covar=tensor([0.0581, 0.0514, 0.0469, 0.0587, 0.0374, 0.0667, 0.0728, 0.0528], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0004, 0.0004, 0.0003], device='cuda:0') 2023-05-15 16:36:28,423 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=105388.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 16:36:38,046 INFO [finetune.py:992] (0/2) Epoch 1, batch 5400, loss[loss=0.1745, simple_loss=0.2665, pruned_loss=0.04127, over 12394.00 frames. ], tot_loss[loss=0.1724, simple_loss=0.2603, pruned_loss=0.0423, over 2361547.44 frames. ], batch size: 38, lr: 5.00e-03, grad_scale: 16.0 2023-05-15 16:36:38,763 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.840e+02 2.689e+02 3.380e+02 4.059e+02 1.017e+03, threshold=6.760e+02, percent-clipped=1.0 2023-05-15 16:37:13,350 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=105449.0, num_to_drop=1, layers_to_drop={2} 2023-05-15 16:37:14,586 INFO [finetune.py:992] (0/2) Epoch 1, batch 5450, loss[loss=0.1491, simple_loss=0.2435, pruned_loss=0.02737, over 12095.00 frames. ], tot_loss[loss=0.1716, simple_loss=0.2596, pruned_loss=0.04177, over 2369507.47 frames. ], batch size: 32, lr: 5.00e-03, grad_scale: 16.0 2023-05-15 16:37:40,629 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=105486.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 16:37:51,360 INFO [finetune.py:992] (0/2) Epoch 1, batch 5500, loss[loss=0.1671, simple_loss=0.2659, pruned_loss=0.03412, over 12349.00 frames. ], tot_loss[loss=0.1718, simple_loss=0.2596, pruned_loss=0.04197, over 2359202.32 frames. ], batch size: 36, lr: 5.00e-03, grad_scale: 16.0 2023-05-15 16:37:51,979 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.936e+02 2.780e+02 3.412e+02 4.141e+02 8.244e+02, threshold=6.823e+02, percent-clipped=2.0 2023-05-15 16:37:52,153 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=105502.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 16:37:59,897 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.9692, 5.7968, 5.3091, 5.3069, 5.9021, 5.1923, 5.5293, 5.4590], device='cuda:0'), covar=tensor([0.1387, 0.0973, 0.0950, 0.1927, 0.0850, 0.2101, 0.1478, 0.0888], device='cuda:0'), in_proj_covar=tensor([0.0338, 0.0475, 0.0367, 0.0420, 0.0445, 0.0416, 0.0385, 0.0367], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-15 16:38:09,994 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.9086, 3.5095, 5.2819, 2.8464, 2.9420, 4.1785, 3.2863, 4.2669], device='cuda:0'), covar=tensor([0.0371, 0.0988, 0.0222, 0.1028, 0.1716, 0.1047, 0.1256, 0.0795], device='cuda:0'), in_proj_covar=tensor([0.0237, 0.0235, 0.0243, 0.0184, 0.0242, 0.0290, 0.0231, 0.0256], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-15 16:38:24,669 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=105547.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 16:38:27,278 INFO [finetune.py:992] (0/2) Epoch 1, batch 5550, loss[loss=0.1672, simple_loss=0.2555, pruned_loss=0.03941, over 12289.00 frames. ], tot_loss[loss=0.1727, simple_loss=0.2603, pruned_loss=0.04259, over 2356827.80 frames. ], batch size: 34, lr: 5.00e-03, grad_scale: 16.0 2023-05-15 16:38:31,716 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.6165, 2.3804, 3.7755, 4.5564, 4.1637, 4.5679, 4.0351, 3.2474], device='cuda:0'), covar=tensor([0.0024, 0.0366, 0.0105, 0.0035, 0.0074, 0.0059, 0.0082, 0.0290], device='cuda:0'), in_proj_covar=tensor([0.0079, 0.0121, 0.0106, 0.0077, 0.0100, 0.0109, 0.0085, 0.0137], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-15 16:38:35,982 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=105563.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 16:39:04,069 INFO [finetune.py:992] (0/2) Epoch 1, batch 5600, loss[loss=0.1538, simple_loss=0.2439, pruned_loss=0.03189, over 12409.00 frames. ], tot_loss[loss=0.173, simple_loss=0.2607, pruned_loss=0.04262, over 2358370.96 frames. ], batch size: 32, lr: 5.00e-03, grad_scale: 16.0 2023-05-15 16:39:04,721 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.950e+02 2.813e+02 3.398e+02 3.932e+02 6.738e+02, threshold=6.795e+02, percent-clipped=0.0 2023-05-15 16:39:39,746 INFO [finetune.py:992] (0/2) Epoch 1, batch 5650, loss[loss=0.1369, simple_loss=0.2188, pruned_loss=0.02753, over 12259.00 frames. ], tot_loss[loss=0.1731, simple_loss=0.261, pruned_loss=0.04257, over 2361608.48 frames. ], batch size: 28, lr: 5.00e-03, grad_scale: 16.0 2023-05-15 16:39:57,272 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.3748, 2.7188, 3.9872, 3.2801, 3.6850, 3.3535, 2.8263, 3.8014], device='cuda:0'), covar=tensor([0.0094, 0.0258, 0.0081, 0.0185, 0.0120, 0.0142, 0.0253, 0.0094], device='cuda:0'), in_proj_covar=tensor([0.0158, 0.0190, 0.0167, 0.0172, 0.0192, 0.0150, 0.0180, 0.0160], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-15 16:40:08,738 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.2247, 4.9041, 5.0786, 5.0690, 4.8799, 5.0685, 5.0255, 2.9440], device='cuda:0'), covar=tensor([0.0061, 0.0052, 0.0056, 0.0042, 0.0044, 0.0064, 0.0071, 0.0529], device='cuda:0'), in_proj_covar=tensor([0.0065, 0.0075, 0.0078, 0.0072, 0.0059, 0.0088, 0.0078, 0.0091], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0004], device='cuda:0') 2023-05-15 16:40:16,551 INFO [finetune.py:992] (0/2) Epoch 1, batch 5700, loss[loss=0.1766, simple_loss=0.2744, pruned_loss=0.03937, over 12113.00 frames. ], tot_loss[loss=0.173, simple_loss=0.2608, pruned_loss=0.04261, over 2359581.96 frames. ], batch size: 38, lr: 5.00e-03, grad_scale: 16.0 2023-05-15 16:40:17,229 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.090e+02 2.924e+02 3.415e+02 4.302e+02 8.366e+02, threshold=6.830e+02, percent-clipped=2.0 2023-05-15 16:40:31,862 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.1189, 4.7626, 4.8906, 4.9056, 4.6860, 4.9424, 4.9191, 2.6363], device='cuda:0'), covar=tensor([0.0071, 0.0053, 0.0066, 0.0047, 0.0050, 0.0070, 0.0053, 0.0670], device='cuda:0'), in_proj_covar=tensor([0.0065, 0.0076, 0.0079, 0.0072, 0.0060, 0.0088, 0.0078, 0.0092], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0004], device='cuda:0') 2023-05-15 16:40:48,251 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=105744.0, num_to_drop=1, layers_to_drop={2} 2023-05-15 16:40:53,141 INFO [finetune.py:992] (0/2) Epoch 1, batch 5750, loss[loss=0.291, simple_loss=0.3342, pruned_loss=0.1239, over 8120.00 frames. ], tot_loss[loss=0.1726, simple_loss=0.2604, pruned_loss=0.04239, over 2357942.70 frames. ], batch size: 98, lr: 5.00e-03, grad_scale: 16.0 2023-05-15 16:40:56,798 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.8561, 3.4364, 5.2525, 2.7162, 2.7653, 4.0104, 3.3579, 4.0681], device='cuda:0'), covar=tensor([0.0387, 0.1013, 0.0225, 0.1213, 0.1885, 0.1137, 0.1248, 0.0962], device='cuda:0'), in_proj_covar=tensor([0.0234, 0.0230, 0.0239, 0.0181, 0.0238, 0.0285, 0.0227, 0.0252], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-15 16:41:22,347 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=105792.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 16:41:27,449 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=105799.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 16:41:28,901 INFO [finetune.py:992] (0/2) Epoch 1, batch 5800, loss[loss=0.1806, simple_loss=0.267, pruned_loss=0.04715, over 12107.00 frames. ], tot_loss[loss=0.1727, simple_loss=0.2604, pruned_loss=0.04256, over 2364404.69 frames. ], batch size: 33, lr: 5.00e-03, grad_scale: 16.0 2023-05-15 16:41:29,278 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 2023-05-15 16:41:29,592 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.183e+02 2.921e+02 3.626e+02 4.206e+02 7.946e+02, threshold=7.252e+02, percent-clipped=5.0 2023-05-15 16:41:36,912 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.4015, 5.2360, 5.3358, 5.3995, 5.0094, 5.0274, 4.8761, 5.2950], device='cuda:0'), covar=tensor([0.0591, 0.0512, 0.0636, 0.0504, 0.1590, 0.1143, 0.0526, 0.0845], device='cuda:0'), in_proj_covar=tensor([0.0501, 0.0638, 0.0546, 0.0607, 0.0796, 0.0722, 0.0528, 0.0468], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0005, 0.0004, 0.0005, 0.0006, 0.0005, 0.0004, 0.0004], device='cuda:0') 2023-05-15 16:41:42,152 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 2023-05-15 16:41:59,176 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=105842.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 16:41:59,929 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.1907, 4.8215, 5.1216, 4.5296, 4.8680, 4.6009, 5.1464, 4.8369], device='cuda:0'), covar=tensor([0.0219, 0.0313, 0.0274, 0.0222, 0.0256, 0.0280, 0.0214, 0.0275], device='cuda:0'), in_proj_covar=tensor([0.0238, 0.0250, 0.0265, 0.0241, 0.0239, 0.0240, 0.0220, 0.0202], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0005, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-15 16:42:05,523 INFO [finetune.py:992] (0/2) Epoch 1, batch 5850, loss[loss=0.1776, simple_loss=0.2694, pruned_loss=0.04291, over 12105.00 frames. ], tot_loss[loss=0.1733, simple_loss=0.2608, pruned_loss=0.04287, over 2368202.12 frames. ], batch size: 38, lr: 5.00e-03, grad_scale: 16.0 2023-05-15 16:42:07,166 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=105853.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 16:42:10,749 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=105858.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 16:42:12,311 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=105860.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 16:42:16,314 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.90 vs. limit=2.0 2023-05-15 16:42:42,321 INFO [finetune.py:992] (0/2) Epoch 1, batch 5900, loss[loss=0.1608, simple_loss=0.2484, pruned_loss=0.03662, over 12247.00 frames. ], tot_loss[loss=0.1737, simple_loss=0.2607, pruned_loss=0.04337, over 2365211.68 frames. ], batch size: 32, lr: 5.00e-03, grad_scale: 16.0 2023-05-15 16:42:43,030 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.033e+02 2.888e+02 3.561e+02 4.128e+02 1.057e+03, threshold=7.123e+02, percent-clipped=1.0 2023-05-15 16:43:17,592 INFO [finetune.py:992] (0/2) Epoch 1, batch 5950, loss[loss=0.1882, simple_loss=0.2754, pruned_loss=0.05048, over 12143.00 frames. ], tot_loss[loss=0.1747, simple_loss=0.2614, pruned_loss=0.04396, over 2358343.45 frames. ], batch size: 38, lr: 5.00e-03, grad_scale: 16.0 2023-05-15 16:43:28,114 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.96 vs. limit=2.0 2023-05-15 16:43:41,034 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=4.76 vs. limit=5.0 2023-05-15 16:43:53,578 INFO [checkpoint.py:75] (0/2) Saving checkpoint to pruned_transducer_stateless7/exp_giga_finetune/checkpoint-6000.pt 2023-05-15 16:43:57,610 INFO [finetune.py:992] (0/2) Epoch 1, batch 6000, loss[loss=0.1656, simple_loss=0.2472, pruned_loss=0.04204, over 12174.00 frames. ], tot_loss[loss=0.1748, simple_loss=0.2618, pruned_loss=0.04384, over 2366099.26 frames. ], batch size: 29, lr: 5.00e-03, grad_scale: 16.0 2023-05-15 16:43:57,611 INFO [finetune.py:1017] (0/2) Computing validation loss 2023-05-15 16:44:17,041 INFO [finetune.py:1026] (0/2) Epoch 1, validation: loss=0.3651, simple_loss=0.4266, pruned_loss=0.1518, over 1020973.00 frames. 2023-05-15 16:44:17,042 INFO [finetune.py:1027] (0/2) Maximum memory allocated so far is 12608MB 2023-05-15 16:44:17,700 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.761e+02 2.870e+02 3.370e+02 4.278e+02 8.951e+02, threshold=6.741e+02, percent-clipped=4.0 2023-05-15 16:44:35,636 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.41 vs. limit=2.0 2023-05-15 16:44:48,416 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=106044.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 16:44:50,523 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.0997, 5.9982, 5.4385, 5.5670, 5.9936, 5.2615, 5.7079, 5.6132], device='cuda:0'), covar=tensor([0.1348, 0.0815, 0.0936, 0.1545, 0.0807, 0.2011, 0.1243, 0.0875], device='cuda:0'), in_proj_covar=tensor([0.0335, 0.0472, 0.0367, 0.0416, 0.0442, 0.0416, 0.0381, 0.0365], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-15 16:44:53,230 INFO [finetune.py:992] (0/2) Epoch 1, batch 6050, loss[loss=0.1787, simple_loss=0.2713, pruned_loss=0.04304, over 12192.00 frames. ], tot_loss[loss=0.1749, simple_loss=0.2623, pruned_loss=0.04371, over 2371385.13 frames. ], batch size: 35, lr: 5.00e-03, grad_scale: 16.0 2023-05-15 16:45:12,266 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=106077.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 16:45:17,119 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.2742, 4.6848, 2.8254, 2.4571, 3.9541, 2.3607, 4.0080, 3.2005], device='cuda:0'), covar=tensor([0.0613, 0.0535, 0.1072, 0.1548, 0.0295, 0.1384, 0.0389, 0.0790], device='cuda:0'), in_proj_covar=tensor([0.0193, 0.0252, 0.0175, 0.0198, 0.0139, 0.0180, 0.0195, 0.0173], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-15 16:45:23,580 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=106092.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 16:45:30,118 INFO [finetune.py:992] (0/2) Epoch 1, batch 6100, loss[loss=0.182, simple_loss=0.2653, pruned_loss=0.04931, over 12370.00 frames. ], tot_loss[loss=0.1749, simple_loss=0.2626, pruned_loss=0.04358, over 2374421.88 frames. ], batch size: 38, lr: 5.00e-03, grad_scale: 8.0 2023-05-15 16:45:31,570 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.855e+02 2.834e+02 3.255e+02 3.982e+02 7.817e+02, threshold=6.509e+02, percent-clipped=2.0 2023-05-15 16:45:54,201 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.2898, 4.9153, 5.1169, 4.4641, 4.9145, 4.5190, 5.1246, 5.0119], device='cuda:0'), covar=tensor([0.0308, 0.0551, 0.0530, 0.0336, 0.0327, 0.0363, 0.0402, 0.0305], device='cuda:0'), in_proj_covar=tensor([0.0237, 0.0249, 0.0264, 0.0240, 0.0239, 0.0239, 0.0220, 0.0202], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0005, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003], device='cuda:0') 2023-05-15 16:45:57,142 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=106138.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 16:46:00,634 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=106142.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 16:46:04,912 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=106148.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 16:46:06,972 INFO [finetune.py:992] (0/2) Epoch 1, batch 6150, loss[loss=0.1808, simple_loss=0.2698, pruned_loss=0.04588, over 12143.00 frames. ], tot_loss[loss=0.1751, simple_loss=0.2629, pruned_loss=0.04368, over 2373910.32 frames. ], batch size: 38, lr: 5.00e-03, grad_scale: 8.0 2023-05-15 16:46:09,902 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=106155.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 16:46:12,176 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=106158.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 16:46:34,897 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=106190.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 16:46:43,150 INFO [finetune.py:992] (0/2) Epoch 1, batch 6200, loss[loss=0.1741, simple_loss=0.2591, pruned_loss=0.04448, over 12348.00 frames. ], tot_loss[loss=0.1757, simple_loss=0.2633, pruned_loss=0.04408, over 2366369.75 frames. ], batch size: 31, lr: 5.00e-03, grad_scale: 8.0 2023-05-15 16:46:44,522 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.121e+02 2.817e+02 3.317e+02 3.981e+02 7.574e+02, threshold=6.635e+02, percent-clipped=2.0 2023-05-15 16:46:45,498 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.2277, 2.0770, 2.5044, 2.2801, 2.4387, 2.3445, 1.9756, 2.4912], device='cuda:0'), covar=tensor([0.0099, 0.0187, 0.0119, 0.0142, 0.0117, 0.0124, 0.0190, 0.0099], device='cuda:0'), in_proj_covar=tensor([0.0158, 0.0190, 0.0167, 0.0172, 0.0192, 0.0152, 0.0181, 0.0160], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-15 16:46:46,766 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=106206.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 16:47:09,064 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.2387, 4.8566, 5.1893, 4.5663, 4.8659, 4.6740, 5.2544, 4.9401], device='cuda:0'), covar=tensor([0.0227, 0.0316, 0.0259, 0.0234, 0.0270, 0.0280, 0.0169, 0.0225], device='cuda:0'), in_proj_covar=tensor([0.0235, 0.0248, 0.0264, 0.0239, 0.0238, 0.0239, 0.0219, 0.0200], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0005, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003], device='cuda:0') 2023-05-15 16:47:19,652 INFO [finetune.py:992] (0/2) Epoch 1, batch 6250, loss[loss=0.1771, simple_loss=0.2672, pruned_loss=0.04351, over 12282.00 frames. ], tot_loss[loss=0.1763, simple_loss=0.2636, pruned_loss=0.04444, over 2366515.68 frames. ], batch size: 37, lr: 5.00e-03, grad_scale: 8.0 2023-05-15 16:47:31,798 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=106268.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 16:47:38,764 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.4810, 5.3132, 5.4044, 5.4528, 5.0597, 5.1237, 4.9527, 5.4503], device='cuda:0'), covar=tensor([0.0589, 0.0517, 0.0601, 0.0564, 0.1789, 0.1168, 0.0486, 0.0725], device='cuda:0'), in_proj_covar=tensor([0.0500, 0.0634, 0.0545, 0.0603, 0.0796, 0.0717, 0.0526, 0.0470], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0005, 0.0004, 0.0005, 0.0006, 0.0005, 0.0004, 0.0004], device='cuda:0') 2023-05-15 16:47:55,373 INFO [finetune.py:992] (0/2) Epoch 1, batch 6300, loss[loss=0.1965, simple_loss=0.2844, pruned_loss=0.05432, over 11232.00 frames. ], tot_loss[loss=0.1765, simple_loss=0.2639, pruned_loss=0.04452, over 2368930.58 frames. ], batch size: 55, lr: 5.00e-03, grad_scale: 8.0 2023-05-15 16:47:56,623 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.327e+02 3.079e+02 3.553e+02 4.255e+02 1.115e+03, threshold=7.105e+02, percent-clipped=3.0 2023-05-15 16:48:13,679 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.41 vs. limit=2.0 2023-05-15 16:48:15,380 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=106329.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 16:48:20,426 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=106336.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 16:48:25,736 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.94 vs. limit=2.0 2023-05-15 16:48:30,961 INFO [finetune.py:992] (0/2) Epoch 1, batch 6350, loss[loss=0.1849, simple_loss=0.2601, pruned_loss=0.05481, over 12018.00 frames. ], tot_loss[loss=0.1758, simple_loss=0.2633, pruned_loss=0.04415, over 2371157.92 frames. ], batch size: 28, lr: 4.99e-03, grad_scale: 8.0 2023-05-15 16:48:44,945 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.35 vs. limit=2.0 2023-05-15 16:48:45,462 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=2.58 vs. limit=5.0 2023-05-15 16:48:53,246 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.5321, 2.3422, 3.8159, 4.5867, 4.0809, 4.4679, 3.9817, 3.3448], device='cuda:0'), covar=tensor([0.0030, 0.0381, 0.0106, 0.0029, 0.0088, 0.0061, 0.0076, 0.0274], device='cuda:0'), in_proj_covar=tensor([0.0080, 0.0122, 0.0106, 0.0077, 0.0100, 0.0111, 0.0086, 0.0138], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-15 16:49:04,713 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=106397.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 16:49:07,699 INFO [finetune.py:992] (0/2) Epoch 1, batch 6400, loss[loss=0.1564, simple_loss=0.2462, pruned_loss=0.03327, over 12106.00 frames. ], tot_loss[loss=0.1748, simple_loss=0.2625, pruned_loss=0.04353, over 2380904.44 frames. ], batch size: 33, lr: 4.99e-03, grad_scale: 8.0 2023-05-15 16:49:09,153 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.991e+02 2.851e+02 3.219e+02 3.942e+02 7.404e+02, threshold=6.437e+02, percent-clipped=1.0 2023-05-15 16:49:30,904 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=106433.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 16:49:35,152 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.4541, 5.0415, 5.3815, 4.7801, 5.0512, 4.8868, 5.4665, 5.0075], device='cuda:0'), covar=tensor([0.0192, 0.0326, 0.0250, 0.0225, 0.0288, 0.0247, 0.0166, 0.0271], device='cuda:0'), in_proj_covar=tensor([0.0237, 0.0248, 0.0263, 0.0238, 0.0237, 0.0238, 0.0218, 0.0202], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0005, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003], device='cuda:0') 2023-05-15 16:49:42,321 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=106448.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 16:49:44,367 INFO [finetune.py:992] (0/2) Epoch 1, batch 6450, loss[loss=0.1636, simple_loss=0.2577, pruned_loss=0.03476, over 12352.00 frames. ], tot_loss[loss=0.1743, simple_loss=0.262, pruned_loss=0.04325, over 2384902.56 frames. ], batch size: 36, lr: 4.99e-03, grad_scale: 8.0 2023-05-15 16:49:46,687 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=106454.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 16:49:47,365 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=106455.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 16:50:16,410 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=106496.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 16:50:19,977 INFO [finetune.py:992] (0/2) Epoch 1, batch 6500, loss[loss=0.1491, simple_loss=0.233, pruned_loss=0.03263, over 12180.00 frames. ], tot_loss[loss=0.1747, simple_loss=0.2623, pruned_loss=0.04355, over 2381183.69 frames. ], batch size: 29, lr: 4.99e-03, grad_scale: 8.0 2023-05-15 16:50:21,365 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.022e+02 2.770e+02 3.404e+02 3.913e+02 8.890e+02, threshold=6.807e+02, percent-clipped=2.0 2023-05-15 16:50:21,446 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=106503.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 16:50:30,203 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=106515.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 16:50:56,136 INFO [finetune.py:992] (0/2) Epoch 1, batch 6550, loss[loss=0.1891, simple_loss=0.2805, pruned_loss=0.04889, over 12056.00 frames. ], tot_loss[loss=0.1743, simple_loss=0.262, pruned_loss=0.04326, over 2389461.33 frames. ], batch size: 40, lr: 4.99e-03, grad_scale: 8.0 2023-05-15 16:51:32,699 INFO [finetune.py:992] (0/2) Epoch 1, batch 6600, loss[loss=0.2844, simple_loss=0.351, pruned_loss=0.1089, over 8402.00 frames. ], tot_loss[loss=0.1747, simple_loss=0.2622, pruned_loss=0.04359, over 2383194.84 frames. ], batch size: 98, lr: 4.99e-03, grad_scale: 8.0 2023-05-15 16:51:34,107 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.440e+02 3.236e+02 3.772e+02 4.434e+02 3.589e+03, threshold=7.544e+02, percent-clipped=5.0 2023-05-15 16:51:49,168 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=106624.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 16:51:50,842 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2023-05-15 16:51:59,133 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([6.1422, 6.0600, 5.8723, 5.4705, 5.2562, 6.0515, 5.6629, 5.3885], device='cuda:0'), covar=tensor([0.0568, 0.0897, 0.0582, 0.1314, 0.0619, 0.0658, 0.1255, 0.0898], device='cuda:0'), in_proj_covar=tensor([0.0584, 0.0514, 0.0467, 0.0585, 0.0375, 0.0668, 0.0723, 0.0527], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0004, 0.0004, 0.0003], device='cuda:0') 2023-05-15 16:52:03,570 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.7720, 4.9016, 4.3786, 5.2226, 4.8971, 3.1616, 4.5947, 3.3755], device='cuda:0'), covar=tensor([0.0536, 0.0609, 0.1133, 0.0285, 0.0798, 0.1328, 0.0892, 0.2728], device='cuda:0'), in_proj_covar=tensor([0.0297, 0.0378, 0.0353, 0.0264, 0.0364, 0.0265, 0.0335, 0.0363], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-15 16:52:07,802 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.3537, 2.5944, 3.8336, 3.2205, 3.5898, 3.3590, 2.7836, 3.7775], device='cuda:0'), covar=tensor([0.0094, 0.0311, 0.0125, 0.0199, 0.0126, 0.0156, 0.0265, 0.0091], device='cuda:0'), in_proj_covar=tensor([0.0158, 0.0194, 0.0169, 0.0174, 0.0193, 0.0152, 0.0182, 0.0161], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-15 16:52:08,320 INFO [finetune.py:992] (0/2) Epoch 1, batch 6650, loss[loss=0.1376, simple_loss=0.2224, pruned_loss=0.02643, over 12265.00 frames. ], tot_loss[loss=0.176, simple_loss=0.2632, pruned_loss=0.04437, over 2376041.03 frames. ], batch size: 28, lr: 4.99e-03, grad_scale: 8.0 2023-05-15 16:52:14,282 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.5740, 2.8547, 3.8375, 2.4726, 2.6190, 3.1329, 2.8877, 3.2538], device='cuda:0'), covar=tensor([0.0554, 0.1111, 0.0462, 0.1057, 0.1615, 0.1319, 0.1128, 0.1059], device='cuda:0'), in_proj_covar=tensor([0.0235, 0.0232, 0.0241, 0.0181, 0.0239, 0.0285, 0.0228, 0.0256], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-15 16:52:38,327 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=106692.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 16:52:44,671 INFO [finetune.py:992] (0/2) Epoch 1, batch 6700, loss[loss=0.1546, simple_loss=0.232, pruned_loss=0.03862, over 12127.00 frames. ], tot_loss[loss=0.1751, simple_loss=0.2622, pruned_loss=0.04401, over 2374544.12 frames. ], batch size: 30, lr: 4.99e-03, grad_scale: 8.0 2023-05-15 16:52:45,606 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.5552, 2.5281, 3.8173, 4.5467, 4.0728, 4.4990, 3.9170, 3.2107], device='cuda:0'), covar=tensor([0.0023, 0.0347, 0.0109, 0.0029, 0.0091, 0.0047, 0.0076, 0.0295], device='cuda:0'), in_proj_covar=tensor([0.0078, 0.0120, 0.0105, 0.0076, 0.0100, 0.0109, 0.0085, 0.0136], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-15 16:52:46,068 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.078e+02 3.048e+02 3.636e+02 4.318e+02 9.262e+02, threshold=7.272e+02, percent-clipped=2.0 2023-05-15 16:53:08,303 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=106733.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 16:53:21,025 INFO [finetune.py:992] (0/2) Epoch 1, batch 6750, loss[loss=0.1402, simple_loss=0.2226, pruned_loss=0.02895, over 11994.00 frames. ], tot_loss[loss=0.1741, simple_loss=0.2611, pruned_loss=0.04354, over 2371192.76 frames. ], batch size: 28, lr: 4.99e-03, grad_scale: 8.0 2023-05-15 16:53:35,873 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2023-05-15 16:53:41,351 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.1299, 2.4250, 3.2656, 4.0683, 2.1250, 4.2013, 4.1231, 4.2799], device='cuda:0'), covar=tensor([0.0120, 0.1077, 0.0363, 0.0121, 0.1201, 0.0168, 0.0139, 0.0076], device='cuda:0'), in_proj_covar=tensor([0.0121, 0.0201, 0.0189, 0.0115, 0.0187, 0.0181, 0.0170, 0.0120], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-15 16:53:42,611 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=106781.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 16:53:45,131 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.42 vs. limit=2.0 2023-05-15 16:53:56,873 INFO [finetune.py:992] (0/2) Epoch 1, batch 6800, loss[loss=0.1982, simple_loss=0.2846, pruned_loss=0.05588, over 12119.00 frames. ], tot_loss[loss=0.1734, simple_loss=0.2607, pruned_loss=0.04307, over 2378358.77 frames. ], batch size: 38, lr: 4.99e-03, grad_scale: 8.0 2023-05-15 16:53:58,275 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.844e+02 2.775e+02 3.272e+02 4.008e+02 8.494e+02, threshold=6.545e+02, percent-clipped=1.0 2023-05-15 16:54:03,952 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=106810.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 16:54:04,142 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.6593, 2.8095, 4.5106, 4.6601, 2.9306, 2.6887, 3.1088, 2.0982], device='cuda:0'), covar=tensor([0.1407, 0.2958, 0.0410, 0.0405, 0.1129, 0.1916, 0.2238, 0.3622], device='cuda:0'), in_proj_covar=tensor([0.0279, 0.0373, 0.0266, 0.0291, 0.0251, 0.0279, 0.0354, 0.0350], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-15 16:54:33,084 INFO [finetune.py:992] (0/2) Epoch 1, batch 6850, loss[loss=0.1867, simple_loss=0.2784, pruned_loss=0.04747, over 12023.00 frames. ], tot_loss[loss=0.1742, simple_loss=0.262, pruned_loss=0.04323, over 2381166.88 frames. ], batch size: 40, lr: 4.99e-03, grad_scale: 8.0 2023-05-15 16:55:09,344 INFO [finetune.py:992] (0/2) Epoch 1, batch 6900, loss[loss=0.2129, simple_loss=0.2895, pruned_loss=0.06815, over 8453.00 frames. ], tot_loss[loss=0.1742, simple_loss=0.2615, pruned_loss=0.04341, over 2376862.63 frames. ], batch size: 99, lr: 4.99e-03, grad_scale: 8.0 2023-05-15 16:55:10,779 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.008e+02 2.903e+02 3.472e+02 4.212e+02 6.915e+02, threshold=6.944e+02, percent-clipped=1.0 2023-05-15 16:55:25,819 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=106924.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 16:55:28,728 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.6566, 2.8562, 3.6302, 4.5791, 2.7699, 4.6963, 4.5920, 4.8624], device='cuda:0'), covar=tensor([0.0097, 0.1066, 0.0331, 0.0133, 0.1013, 0.0158, 0.0134, 0.0066], device='cuda:0'), in_proj_covar=tensor([0.0123, 0.0203, 0.0191, 0.0116, 0.0189, 0.0183, 0.0172, 0.0121], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002], device='cuda:0') 2023-05-15 16:55:44,749 INFO [finetune.py:992] (0/2) Epoch 1, batch 6950, loss[loss=0.1821, simple_loss=0.2727, pruned_loss=0.04577, over 12053.00 frames. ], tot_loss[loss=0.1742, simple_loss=0.2617, pruned_loss=0.0434, over 2374004.69 frames. ], batch size: 40, lr: 4.99e-03, grad_scale: 8.0 2023-05-15 16:56:00,677 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=106972.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 16:56:12,205 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.5091, 4.8211, 2.8501, 2.0479, 4.2636, 2.4107, 4.0896, 3.1126], device='cuda:0'), covar=tensor([0.0514, 0.0500, 0.1171, 0.2070, 0.0250, 0.1540, 0.0442, 0.0919], device='cuda:0'), in_proj_covar=tensor([0.0191, 0.0250, 0.0174, 0.0195, 0.0138, 0.0179, 0.0193, 0.0173], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-15 16:56:14,988 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=106992.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 16:56:21,512 INFO [finetune.py:992] (0/2) Epoch 1, batch 7000, loss[loss=0.2542, simple_loss=0.3261, pruned_loss=0.09113, over 8330.00 frames. ], tot_loss[loss=0.1752, simple_loss=0.2625, pruned_loss=0.04397, over 2361190.72 frames. ], batch size: 98, lr: 4.99e-03, grad_scale: 8.0 2023-05-15 16:56:22,940 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.089e+02 2.961e+02 3.430e+02 4.386e+02 1.005e+03, threshold=6.860e+02, percent-clipped=4.0 2023-05-15 16:56:42,153 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.4458, 4.3621, 4.2608, 4.3556, 3.9918, 4.4306, 4.3867, 4.5930], device='cuda:0'), covar=tensor([0.0191, 0.0145, 0.0193, 0.0279, 0.0713, 0.0293, 0.0147, 0.0174], device='cuda:0'), in_proj_covar=tensor([0.0181, 0.0185, 0.0183, 0.0228, 0.0232, 0.0197, 0.0169, 0.0219], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0005, 0.0005, 0.0005, 0.0004, 0.0005], device='cuda:0') 2023-05-15 16:56:49,834 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=107040.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 16:56:57,650 INFO [finetune.py:992] (0/2) Epoch 1, batch 7050, loss[loss=0.1574, simple_loss=0.2532, pruned_loss=0.03084, over 12305.00 frames. ], tot_loss[loss=0.1753, simple_loss=0.2626, pruned_loss=0.04399, over 2354351.79 frames. ], batch size: 34, lr: 4.99e-03, grad_scale: 8.0 2023-05-15 16:57:23,332 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.97 vs. limit=2.0 2023-05-15 16:57:33,392 INFO [finetune.py:992] (0/2) Epoch 1, batch 7100, loss[loss=0.2734, simple_loss=0.3335, pruned_loss=0.1067, over 8294.00 frames. ], tot_loss[loss=0.1754, simple_loss=0.2626, pruned_loss=0.04408, over 2358402.73 frames. ], batch size: 97, lr: 4.99e-03, grad_scale: 8.0 2023-05-15 16:57:34,805 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.903e+02 2.881e+02 3.375e+02 4.211e+02 1.027e+03, threshold=6.751e+02, percent-clipped=1.0 2023-05-15 16:57:40,062 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=107110.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 16:58:01,635 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=107140.0, num_to_drop=1, layers_to_drop={1} 2023-05-15 16:58:06,121 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.0471, 3.5216, 5.2662, 2.6600, 2.9826, 4.1088, 3.6161, 4.0077], device='cuda:0'), covar=tensor([0.0336, 0.1012, 0.0272, 0.1307, 0.1838, 0.1139, 0.1118, 0.1095], device='cuda:0'), in_proj_covar=tensor([0.0235, 0.0232, 0.0242, 0.0181, 0.0239, 0.0285, 0.0229, 0.0255], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-15 16:58:09,323 INFO [finetune.py:992] (0/2) Epoch 1, batch 7150, loss[loss=0.1823, simple_loss=0.2801, pruned_loss=0.04227, over 12124.00 frames. ], tot_loss[loss=0.1748, simple_loss=0.2623, pruned_loss=0.04368, over 2370485.39 frames. ], batch size: 39, lr: 4.99e-03, grad_scale: 8.0 2023-05-15 16:58:10,266 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=107152.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 16:58:14,427 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=107158.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 16:58:46,317 INFO [finetune.py:992] (0/2) Epoch 1, batch 7200, loss[loss=0.1556, simple_loss=0.2442, pruned_loss=0.03349, over 12174.00 frames. ], tot_loss[loss=0.1743, simple_loss=0.2617, pruned_loss=0.0434, over 2374877.21 frames. ], batch size: 31, lr: 4.99e-03, grad_scale: 8.0 2023-05-15 16:58:46,529 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=107201.0, num_to_drop=1, layers_to_drop={1} 2023-05-15 16:58:47,846 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.124e+02 2.811e+02 3.297e+02 3.942e+02 8.760e+02, threshold=6.593e+02, percent-clipped=1.0 2023-05-15 16:58:55,088 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=107213.0, num_to_drop=1, layers_to_drop={0} 2023-05-15 16:59:21,832 INFO [finetune.py:992] (0/2) Epoch 1, batch 7250, loss[loss=0.1686, simple_loss=0.2568, pruned_loss=0.04022, over 12136.00 frames. ], tot_loss[loss=0.1739, simple_loss=0.2609, pruned_loss=0.04339, over 2381461.42 frames. ], batch size: 39, lr: 4.99e-03, grad_scale: 8.0 2023-05-15 16:59:57,168 INFO [finetune.py:992] (0/2) Epoch 1, batch 7300, loss[loss=0.1789, simple_loss=0.2727, pruned_loss=0.04256, over 12152.00 frames. ], tot_loss[loss=0.1749, simple_loss=0.262, pruned_loss=0.04385, over 2380994.83 frames. ], batch size: 34, lr: 4.99e-03, grad_scale: 8.0 2023-05-15 16:59:58,595 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.952e+02 2.821e+02 3.345e+02 4.009e+02 7.143e+02, threshold=6.691e+02, percent-clipped=2.0 2023-05-15 17:00:33,503 INFO [finetune.py:992] (0/2) Epoch 1, batch 7350, loss[loss=0.1753, simple_loss=0.2699, pruned_loss=0.04034, over 12036.00 frames. ], tot_loss[loss=0.1745, simple_loss=0.2618, pruned_loss=0.04358, over 2380138.47 frames. ], batch size: 37, lr: 4.99e-03, grad_scale: 8.0 2023-05-15 17:00:44,431 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.4640, 5.1061, 5.4927, 4.8410, 5.1278, 4.9117, 5.4961, 5.1274], device='cuda:0'), covar=tensor([0.0212, 0.0315, 0.0219, 0.0200, 0.0220, 0.0232, 0.0167, 0.0198], device='cuda:0'), in_proj_covar=tensor([0.0231, 0.0242, 0.0253, 0.0231, 0.0230, 0.0231, 0.0212, 0.0195], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-05-15 17:01:09,729 INFO [finetune.py:992] (0/2) Epoch 1, batch 7400, loss[loss=0.1713, simple_loss=0.2552, pruned_loss=0.04369, over 12277.00 frames. ], tot_loss[loss=0.1745, simple_loss=0.2619, pruned_loss=0.04356, over 2376090.42 frames. ], batch size: 33, lr: 4.99e-03, grad_scale: 8.0 2023-05-15 17:01:11,046 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.804e+02 2.953e+02 3.402e+02 4.337e+02 2.849e+03, threshold=6.805e+02, percent-clipped=4.0 2023-05-15 17:01:42,648 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([6.0732, 6.0334, 5.7957, 5.4081, 5.2458, 5.9773, 5.5872, 5.3012], device='cuda:0'), covar=tensor([0.0618, 0.0794, 0.0622, 0.1361, 0.0563, 0.0706, 0.1505, 0.1000], device='cuda:0'), in_proj_covar=tensor([0.0591, 0.0518, 0.0471, 0.0594, 0.0378, 0.0678, 0.0740, 0.0536], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0004, 0.0004, 0.0003], device='cuda:0') 2023-05-15 17:01:45,478 INFO [finetune.py:992] (0/2) Epoch 1, batch 7450, loss[loss=0.1868, simple_loss=0.2787, pruned_loss=0.04746, over 12090.00 frames. ], tot_loss[loss=0.1738, simple_loss=0.261, pruned_loss=0.04327, over 2382119.79 frames. ], batch size: 32, lr: 4.99e-03, grad_scale: 8.0 2023-05-15 17:02:18,619 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=107496.0, num_to_drop=1, layers_to_drop={0} 2023-05-15 17:02:21,937 INFO [finetune.py:992] (0/2) Epoch 1, batch 7500, loss[loss=0.2048, simple_loss=0.2996, pruned_loss=0.05497, over 12144.00 frames. ], tot_loss[loss=0.1742, simple_loss=0.2615, pruned_loss=0.0434, over 2387554.24 frames. ], batch size: 34, lr: 4.99e-03, grad_scale: 8.0 2023-05-15 17:02:23,395 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.422e+02 3.043e+02 3.581e+02 4.065e+02 1.110e+03, threshold=7.161e+02, percent-clipped=3.0 2023-05-15 17:02:27,117 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=107508.0, num_to_drop=1, layers_to_drop={1} 2023-05-15 17:02:27,890 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.3840, 5.2230, 5.2720, 5.3186, 4.9994, 4.9688, 4.8237, 5.2346], device='cuda:0'), covar=tensor([0.0567, 0.0491, 0.0622, 0.0574, 0.1604, 0.1133, 0.0467, 0.1036], device='cuda:0'), in_proj_covar=tensor([0.0496, 0.0635, 0.0544, 0.0600, 0.0784, 0.0713, 0.0523, 0.0466], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0005, 0.0004, 0.0005, 0.0006, 0.0005, 0.0004, 0.0004], device='cuda:0') 2023-05-15 17:02:28,630 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=107510.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 17:02:32,928 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.6545, 3.7505, 3.3881, 3.2520, 2.9033, 2.8430, 3.7645, 2.4934], device='cuda:0'), covar=tensor([0.0265, 0.0098, 0.0137, 0.0158, 0.0332, 0.0319, 0.0095, 0.0362], device='cuda:0'), in_proj_covar=tensor([0.0182, 0.0158, 0.0152, 0.0179, 0.0199, 0.0196, 0.0158, 0.0198], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-15 17:02:58,035 INFO [finetune.py:992] (0/2) Epoch 1, batch 7550, loss[loss=0.2051, simple_loss=0.2844, pruned_loss=0.06288, over 8485.00 frames. ], tot_loss[loss=0.1752, simple_loss=0.2625, pruned_loss=0.04391, over 2385278.26 frames. ], batch size: 100, lr: 4.99e-03, grad_scale: 8.0 2023-05-15 17:03:12,550 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=107571.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 17:03:34,049 INFO [finetune.py:992] (0/2) Epoch 1, batch 7600, loss[loss=0.1488, simple_loss=0.2275, pruned_loss=0.03502, over 12342.00 frames. ], tot_loss[loss=0.1746, simple_loss=0.2617, pruned_loss=0.04374, over 2376815.78 frames. ], batch size: 30, lr: 4.99e-03, grad_scale: 8.0 2023-05-15 17:03:35,433 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.560e+02 2.960e+02 3.577e+02 4.566e+02 1.199e+03, threshold=7.155e+02, percent-clipped=4.0 2023-05-15 17:04:03,037 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=4.62 vs. limit=5.0 2023-05-15 17:04:10,195 INFO [finetune.py:992] (0/2) Epoch 1, batch 7650, loss[loss=0.1571, simple_loss=0.2395, pruned_loss=0.03738, over 12010.00 frames. ], tot_loss[loss=0.176, simple_loss=0.2631, pruned_loss=0.0444, over 2372792.61 frames. ], batch size: 28, lr: 4.99e-03, grad_scale: 8.0 2023-05-15 17:04:16,139 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=107659.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 17:04:20,369 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.1666, 2.3037, 3.0788, 3.9972, 2.2326, 4.1837, 4.0965, 4.2331], device='cuda:0'), covar=tensor([0.0080, 0.1020, 0.0398, 0.0125, 0.1054, 0.0174, 0.0133, 0.0075], device='cuda:0'), in_proj_covar=tensor([0.0119, 0.0197, 0.0186, 0.0113, 0.0183, 0.0177, 0.0168, 0.0117], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-15 17:04:46,002 INFO [finetune.py:992] (0/2) Epoch 1, batch 7700, loss[loss=0.1631, simple_loss=0.2433, pruned_loss=0.04146, over 12177.00 frames. ], tot_loss[loss=0.1767, simple_loss=0.2639, pruned_loss=0.04477, over 2372722.39 frames. ], batch size: 29, lr: 4.99e-03, grad_scale: 8.0 2023-05-15 17:04:47,418 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.809e+02 3.089e+02 3.767e+02 4.573e+02 1.081e+03, threshold=7.534e+02, percent-clipped=2.0 2023-05-15 17:04:49,104 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.8606, 4.8132, 4.6997, 4.6864, 4.3945, 4.7928, 4.7864, 5.0286], device='cuda:0'), covar=tensor([0.0192, 0.0135, 0.0183, 0.0272, 0.0720, 0.0255, 0.0141, 0.0155], device='cuda:0'), in_proj_covar=tensor([0.0183, 0.0186, 0.0185, 0.0231, 0.0233, 0.0198, 0.0170, 0.0221], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0005, 0.0005, 0.0005, 0.0004, 0.0005], device='cuda:0') 2023-05-15 17:04:52,705 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.1387, 6.0252, 5.6544, 5.6103, 6.0778, 5.2957, 5.7288, 5.6508], device='cuda:0'), covar=tensor([0.1419, 0.0849, 0.0694, 0.1689, 0.0923, 0.2257, 0.1414, 0.0830], device='cuda:0'), in_proj_covar=tensor([0.0334, 0.0475, 0.0368, 0.0420, 0.0449, 0.0422, 0.0381, 0.0369], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-15 17:04:59,960 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=107720.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 17:05:12,083 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=107737.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 17:05:21,979 INFO [finetune.py:992] (0/2) Epoch 1, batch 7750, loss[loss=0.1726, simple_loss=0.2588, pruned_loss=0.04314, over 11257.00 frames. ], tot_loss[loss=0.177, simple_loss=0.2641, pruned_loss=0.04497, over 2370431.77 frames. ], batch size: 55, lr: 4.99e-03, grad_scale: 8.0 2023-05-15 17:05:54,705 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=107796.0, num_to_drop=1, layers_to_drop={0} 2023-05-15 17:05:55,611 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.4868, 4.7111, 4.0705, 5.1910, 4.6593, 3.0251, 4.3849, 3.2678], device='cuda:0'), covar=tensor([0.0674, 0.0752, 0.1341, 0.0259, 0.1035, 0.1505, 0.0874, 0.2872], device='cuda:0'), in_proj_covar=tensor([0.0293, 0.0377, 0.0351, 0.0263, 0.0362, 0.0264, 0.0332, 0.0363], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-15 17:05:56,270 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=107798.0, num_to_drop=1, layers_to_drop={2} 2023-05-15 17:05:58,493 INFO [finetune.py:992] (0/2) Epoch 1, batch 7800, loss[loss=0.1634, simple_loss=0.2527, pruned_loss=0.0371, over 12293.00 frames. ], tot_loss[loss=0.1764, simple_loss=0.2633, pruned_loss=0.04477, over 2372271.40 frames. ], batch size: 33, lr: 4.99e-03, grad_scale: 8.0 2023-05-15 17:05:59,951 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.925e+02 2.891e+02 3.488e+02 4.279e+02 1.043e+03, threshold=6.975e+02, percent-clipped=3.0 2023-05-15 17:06:03,627 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=107808.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 17:06:29,477 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=107844.0, num_to_drop=1, layers_to_drop={1} 2023-05-15 17:06:34,422 INFO [finetune.py:992] (0/2) Epoch 1, batch 7850, loss[loss=0.1604, simple_loss=0.2594, pruned_loss=0.03076, over 12178.00 frames. ], tot_loss[loss=0.1767, simple_loss=0.2637, pruned_loss=0.04482, over 2369224.50 frames. ], batch size: 35, lr: 4.99e-03, grad_scale: 8.0 2023-05-15 17:06:37,927 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=107856.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 17:06:41,619 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.5816, 2.7184, 3.7964, 4.6124, 4.1963, 4.6065, 4.0079, 3.3479], device='cuda:0'), covar=tensor([0.0030, 0.0345, 0.0111, 0.0032, 0.0065, 0.0049, 0.0089, 0.0277], device='cuda:0'), in_proj_covar=tensor([0.0081, 0.0123, 0.0108, 0.0078, 0.0102, 0.0113, 0.0088, 0.0139], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-15 17:06:45,075 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=107866.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 17:07:10,027 INFO [finetune.py:992] (0/2) Epoch 1, batch 7900, loss[loss=0.1789, simple_loss=0.2636, pruned_loss=0.04712, over 12194.00 frames. ], tot_loss[loss=0.1767, simple_loss=0.2636, pruned_loss=0.04485, over 2366356.99 frames. ], batch size: 35, lr: 4.99e-03, grad_scale: 8.0 2023-05-15 17:07:11,498 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.923e+02 2.826e+02 3.338e+02 3.958e+02 7.049e+02, threshold=6.676e+02, percent-clipped=1.0 2023-05-15 17:07:13,883 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.4983, 4.8006, 2.9357, 2.8886, 4.1499, 2.5799, 4.0779, 3.1910], device='cuda:0'), covar=tensor([0.0539, 0.0470, 0.0953, 0.1268, 0.0216, 0.1228, 0.0393, 0.0776], device='cuda:0'), in_proj_covar=tensor([0.0192, 0.0253, 0.0176, 0.0198, 0.0140, 0.0181, 0.0194, 0.0174], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-15 17:07:46,142 INFO [finetune.py:992] (0/2) Epoch 1, batch 7950, loss[loss=0.3039, simple_loss=0.3499, pruned_loss=0.1289, over 7776.00 frames. ], tot_loss[loss=0.1777, simple_loss=0.2645, pruned_loss=0.04544, over 2359663.06 frames. ], batch size: 98, lr: 4.99e-03, grad_scale: 8.0 2023-05-15 17:07:48,540 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.6765, 2.5858, 4.5001, 4.8012, 3.0735, 2.6370, 2.8500, 1.9908], device='cuda:0'), covar=tensor([0.1285, 0.2929, 0.0409, 0.0286, 0.0983, 0.1795, 0.2406, 0.3585], device='cuda:0'), in_proj_covar=tensor([0.0281, 0.0377, 0.0272, 0.0293, 0.0254, 0.0281, 0.0357, 0.0355], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-15 17:08:21,801 INFO [checkpoint.py:75] (0/2) Saving checkpoint to pruned_transducer_stateless7/exp_giga_finetune/checkpoint-8000.pt 2023-05-15 17:08:25,919 INFO [finetune.py:992] (0/2) Epoch 1, batch 8000, loss[loss=0.1567, simple_loss=0.2432, pruned_loss=0.0351, over 12117.00 frames. ], tot_loss[loss=0.1766, simple_loss=0.2635, pruned_loss=0.04484, over 2371905.20 frames. ], batch size: 30, lr: 4.99e-03, grad_scale: 8.0 2023-05-15 17:08:27,373 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.949e+02 2.835e+02 3.457e+02 4.309e+02 8.802e+02, threshold=6.915e+02, percent-clipped=2.0 2023-05-15 17:08:35,757 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=108015.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 17:09:01,596 INFO [finetune.py:992] (0/2) Epoch 1, batch 8050, loss[loss=0.1889, simple_loss=0.2871, pruned_loss=0.0453, over 12280.00 frames. ], tot_loss[loss=0.1765, simple_loss=0.2633, pruned_loss=0.04483, over 2366875.77 frames. ], batch size: 37, lr: 4.99e-03, grad_scale: 8.0 2023-05-15 17:09:19,892 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.37 vs. limit=2.0 2023-05-15 17:09:32,380 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=108093.0, num_to_drop=1, layers_to_drop={0} 2023-05-15 17:09:37,835 INFO [finetune.py:992] (0/2) Epoch 1, batch 8100, loss[loss=0.1813, simple_loss=0.2774, pruned_loss=0.04266, over 12131.00 frames. ], tot_loss[loss=0.1762, simple_loss=0.2634, pruned_loss=0.04445, over 2372352.30 frames. ], batch size: 39, lr: 4.99e-03, grad_scale: 16.0 2023-05-15 17:09:39,365 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.789e+02 2.748e+02 3.364e+02 4.081e+02 7.047e+02, threshold=6.728e+02, percent-clipped=1.0 2023-05-15 17:10:14,071 INFO [finetune.py:992] (0/2) Epoch 1, batch 8150, loss[loss=0.307, simple_loss=0.3568, pruned_loss=0.1286, over 8180.00 frames. ], tot_loss[loss=0.1771, simple_loss=0.2644, pruned_loss=0.0449, over 2371945.54 frames. ], batch size: 98, lr: 4.99e-03, grad_scale: 16.0 2023-05-15 17:10:24,895 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=108166.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 17:10:50,001 INFO [finetune.py:992] (0/2) Epoch 1, batch 8200, loss[loss=0.1864, simple_loss=0.2716, pruned_loss=0.05064, over 12073.00 frames. ], tot_loss[loss=0.177, simple_loss=0.2644, pruned_loss=0.04483, over 2375646.29 frames. ], batch size: 42, lr: 4.99e-03, grad_scale: 16.0 2023-05-15 17:10:51,342 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.968e+02 2.756e+02 3.279e+02 4.378e+02 1.560e+03, threshold=6.558e+02, percent-clipped=3.0 2023-05-15 17:10:54,626 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.43 vs. limit=2.0 2023-05-15 17:10:59,224 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.1995, 4.3285, 2.5649, 2.2510, 3.7337, 2.2263, 3.7106, 2.8388], device='cuda:0'), covar=tensor([0.0658, 0.0530, 0.1107, 0.1654, 0.0347, 0.1448, 0.0440, 0.0852], device='cuda:0'), in_proj_covar=tensor([0.0190, 0.0252, 0.0175, 0.0197, 0.0141, 0.0180, 0.0193, 0.0173], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-15 17:10:59,734 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=108214.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 17:11:04,387 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-05-15 17:11:26,357 INFO [finetune.py:992] (0/2) Epoch 1, batch 8250, loss[loss=0.2836, simple_loss=0.336, pruned_loss=0.1156, over 7917.00 frames. ], tot_loss[loss=0.1781, simple_loss=0.2653, pruned_loss=0.04542, over 2368574.94 frames. ], batch size: 98, lr: 4.99e-03, grad_scale: 16.0 2023-05-15 17:11:32,801 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.90 vs. limit=2.0 2023-05-15 17:11:40,249 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.4514, 5.2751, 5.3723, 5.4302, 5.0145, 5.0365, 4.9187, 5.3581], device='cuda:0'), covar=tensor([0.0550, 0.0474, 0.0571, 0.0441, 0.1657, 0.1125, 0.0501, 0.0820], device='cuda:0'), in_proj_covar=tensor([0.0496, 0.0632, 0.0543, 0.0599, 0.0788, 0.0714, 0.0524, 0.0467], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0005, 0.0004, 0.0005, 0.0006, 0.0005, 0.0004, 0.0004], device='cuda:0') 2023-05-15 17:12:02,238 INFO [finetune.py:992] (0/2) Epoch 1, batch 8300, loss[loss=0.15, simple_loss=0.2277, pruned_loss=0.03609, over 12273.00 frames. ], tot_loss[loss=0.1766, simple_loss=0.2641, pruned_loss=0.04457, over 2374220.80 frames. ], batch size: 28, lr: 4.99e-03, grad_scale: 16.0 2023-05-15 17:12:03,701 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.245e+02 2.893e+02 3.434e+02 4.076e+02 8.565e+02, threshold=6.869e+02, percent-clipped=3.0 2023-05-15 17:12:12,273 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=108315.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 17:12:19,409 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=108325.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 17:12:37,669 INFO [finetune.py:992] (0/2) Epoch 1, batch 8350, loss[loss=0.1458, simple_loss=0.2374, pruned_loss=0.02705, over 12187.00 frames. ], tot_loss[loss=0.176, simple_loss=0.2637, pruned_loss=0.0442, over 2382629.01 frames. ], batch size: 31, lr: 4.99e-03, grad_scale: 16.0 2023-05-15 17:12:46,218 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=108363.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 17:13:03,368 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=108386.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 17:13:08,360 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=108393.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 17:13:14,220 INFO [finetune.py:992] (0/2) Epoch 1, batch 8400, loss[loss=0.1398, simple_loss=0.2279, pruned_loss=0.02587, over 12185.00 frames. ], tot_loss[loss=0.175, simple_loss=0.2627, pruned_loss=0.04368, over 2384959.52 frames. ], batch size: 29, lr: 4.99e-03, grad_scale: 16.0 2023-05-15 17:13:16,186 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.850e+02 2.951e+02 3.305e+02 4.157e+02 8.866e+02, threshold=6.611e+02, percent-clipped=2.0 2023-05-15 17:13:36,786 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=108432.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 17:13:42,963 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=108441.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 17:13:49,914 INFO [finetune.py:992] (0/2) Epoch 1, batch 8450, loss[loss=0.1773, simple_loss=0.2661, pruned_loss=0.04422, over 12043.00 frames. ], tot_loss[loss=0.1748, simple_loss=0.2625, pruned_loss=0.04352, over 2381480.92 frames. ], batch size: 40, lr: 4.99e-03, grad_scale: 16.0 2023-05-15 17:13:59,567 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.5609, 2.5443, 3.3533, 4.3424, 2.3289, 4.4974, 4.4564, 4.6837], device='cuda:0'), covar=tensor([0.0088, 0.1084, 0.0382, 0.0166, 0.1116, 0.0179, 0.0128, 0.0060], device='cuda:0'), in_proj_covar=tensor([0.0120, 0.0200, 0.0187, 0.0114, 0.0186, 0.0179, 0.0170, 0.0118], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-15 17:14:14,583 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.71 vs. limit=2.0 2023-05-15 17:14:20,055 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=108493.0, num_to_drop=1, layers_to_drop={1} 2023-05-15 17:14:20,928 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-05-15 17:14:25,556 INFO [finetune.py:992] (0/2) Epoch 1, batch 8500, loss[loss=0.1851, simple_loss=0.2786, pruned_loss=0.04583, over 12033.00 frames. ], tot_loss[loss=0.1762, simple_loss=0.2639, pruned_loss=0.04421, over 2370485.94 frames. ], batch size: 40, lr: 4.99e-03, grad_scale: 16.0 2023-05-15 17:14:27,005 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.852e+02 3.104e+02 3.426e+02 4.387e+02 8.910e+02, threshold=6.853e+02, percent-clipped=3.0 2023-05-15 17:14:44,317 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.5908, 2.6295, 3.4251, 4.3959, 2.3871, 4.5168, 4.5231, 4.7353], device='cuda:0'), covar=tensor([0.0078, 0.1076, 0.0370, 0.0115, 0.1115, 0.0184, 0.0128, 0.0057], device='cuda:0'), in_proj_covar=tensor([0.0119, 0.0197, 0.0185, 0.0113, 0.0184, 0.0176, 0.0167, 0.0116], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-15 17:15:02,690 INFO [finetune.py:992] (0/2) Epoch 1, batch 8550, loss[loss=0.1388, simple_loss=0.2146, pruned_loss=0.03148, over 12306.00 frames. ], tot_loss[loss=0.1761, simple_loss=0.2632, pruned_loss=0.04447, over 2357298.44 frames. ], batch size: 28, lr: 4.99e-03, grad_scale: 16.0 2023-05-15 17:15:21,563 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=108577.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 17:15:38,757 INFO [finetune.py:992] (0/2) Epoch 1, batch 8600, loss[loss=0.173, simple_loss=0.2538, pruned_loss=0.04605, over 12025.00 frames. ], tot_loss[loss=0.1772, simple_loss=0.2641, pruned_loss=0.04516, over 2352510.52 frames. ], batch size: 31, lr: 4.99e-03, grad_scale: 16.0 2023-05-15 17:15:40,128 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.899e+02 2.862e+02 3.397e+02 4.304e+02 9.483e+02, threshold=6.794e+02, percent-clipped=4.0 2023-05-15 17:15:44,577 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.3330, 5.1318, 5.2564, 5.3129, 4.9392, 4.9285, 4.8104, 5.2386], device='cuda:0'), covar=tensor([0.0565, 0.0498, 0.0621, 0.0411, 0.1640, 0.1048, 0.0474, 0.0870], device='cuda:0'), in_proj_covar=tensor([0.0491, 0.0628, 0.0533, 0.0588, 0.0776, 0.0705, 0.0516, 0.0461], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0005, 0.0004, 0.0005, 0.0005, 0.0005, 0.0004, 0.0004], device='cuda:0') 2023-05-15 17:16:05,516 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=108638.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 17:16:14,947 INFO [finetune.py:992] (0/2) Epoch 1, batch 8650, loss[loss=0.2004, simple_loss=0.291, pruned_loss=0.05485, over 12012.00 frames. ], tot_loss[loss=0.1765, simple_loss=0.2635, pruned_loss=0.04473, over 2362256.19 frames. ], batch size: 40, lr: 4.99e-03, grad_scale: 16.0 2023-05-15 17:16:21,761 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.42 vs. limit=2.0 2023-05-15 17:16:22,471 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.61 vs. limit=2.0 2023-05-15 17:16:36,707 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=108681.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 17:16:51,187 INFO [finetune.py:992] (0/2) Epoch 1, batch 8700, loss[loss=0.1721, simple_loss=0.2661, pruned_loss=0.03908, over 12364.00 frames. ], tot_loss[loss=0.1763, simple_loss=0.2636, pruned_loss=0.04445, over 2359462.64 frames. ], batch size: 36, lr: 4.99e-03, grad_scale: 16.0 2023-05-15 17:16:52,534 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.655e+02 2.611e+02 3.200e+02 3.810e+02 1.059e+03, threshold=6.400e+02, percent-clipped=5.0 2023-05-15 17:17:08,019 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.2172, 4.9819, 5.1594, 5.0520, 4.9979, 5.1500, 5.0074, 3.2306], device='cuda:0'), covar=tensor([0.0095, 0.0052, 0.0052, 0.0052, 0.0037, 0.0066, 0.0062, 0.0452], device='cuda:0'), in_proj_covar=tensor([0.0066, 0.0076, 0.0079, 0.0071, 0.0059, 0.0087, 0.0077, 0.0091], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0004], device='cuda:0') 2023-05-15 17:17:17,169 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.4227, 3.5767, 3.2288, 3.1082, 2.8258, 2.6769, 3.4145, 2.2653], device='cuda:0'), covar=tensor([0.0330, 0.0110, 0.0162, 0.0165, 0.0335, 0.0267, 0.0127, 0.0424], device='cuda:0'), in_proj_covar=tensor([0.0190, 0.0165, 0.0156, 0.0183, 0.0205, 0.0199, 0.0161, 0.0203], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-15 17:17:26,913 INFO [finetune.py:992] (0/2) Epoch 1, batch 8750, loss[loss=0.1798, simple_loss=0.2686, pruned_loss=0.04545, over 12290.00 frames. ], tot_loss[loss=0.1776, simple_loss=0.2642, pruned_loss=0.04546, over 2346426.82 frames. ], batch size: 34, lr: 4.99e-03, grad_scale: 16.0 2023-05-15 17:17:45,030 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.0991, 4.4799, 4.1190, 4.8988, 4.4188, 2.8096, 4.2298, 3.0569], device='cuda:0'), covar=tensor([0.0775, 0.0729, 0.1231, 0.0313, 0.0935, 0.1514, 0.0822, 0.2881], device='cuda:0'), in_proj_covar=tensor([0.0288, 0.0366, 0.0344, 0.0259, 0.0353, 0.0259, 0.0325, 0.0358], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-15 17:17:53,149 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=108788.0, num_to_drop=1, layers_to_drop={1} 2023-05-15 17:17:55,441 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=108791.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 17:17:55,535 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.4456, 4.7878, 4.4991, 5.2183, 4.7741, 2.7615, 4.4781, 3.2385], device='cuda:0'), covar=tensor([0.0675, 0.0693, 0.0895, 0.0310, 0.0801, 0.1598, 0.0790, 0.2875], device='cuda:0'), in_proj_covar=tensor([0.0289, 0.0368, 0.0345, 0.0259, 0.0354, 0.0260, 0.0326, 0.0359], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-15 17:18:03,507 INFO [finetune.py:992] (0/2) Epoch 1, batch 8800, loss[loss=0.1785, simple_loss=0.2654, pruned_loss=0.04583, over 12286.00 frames. ], tot_loss[loss=0.1774, simple_loss=0.2642, pruned_loss=0.0453, over 2351360.58 frames. ], batch size: 33, lr: 4.99e-03, grad_scale: 16.0 2023-05-15 17:18:04,935 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.005e+02 2.996e+02 3.587e+02 4.204e+02 6.716e+02, threshold=7.173e+02, percent-clipped=2.0 2023-05-15 17:18:39,810 INFO [finetune.py:992] (0/2) Epoch 1, batch 8850, loss[loss=0.208, simple_loss=0.2941, pruned_loss=0.06089, over 12045.00 frames. ], tot_loss[loss=0.1777, simple_loss=0.2645, pruned_loss=0.04546, over 2351466.61 frames. ], batch size: 37, lr: 4.99e-03, grad_scale: 16.0 2023-05-15 17:18:40,735 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=108852.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 17:19:02,298 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.9581, 3.3508, 5.1675, 2.9866, 2.7073, 4.0569, 3.4671, 4.0385], device='cuda:0'), covar=tensor([0.0446, 0.1125, 0.0349, 0.1023, 0.2032, 0.1113, 0.1191, 0.1036], device='cuda:0'), in_proj_covar=tensor([0.0234, 0.0228, 0.0239, 0.0177, 0.0235, 0.0282, 0.0225, 0.0251], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-15 17:19:03,654 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.1532, 2.5236, 3.6007, 4.1774, 3.8075, 4.1560, 3.5936, 2.9259], device='cuda:0'), covar=tensor([0.0033, 0.0340, 0.0106, 0.0034, 0.0088, 0.0059, 0.0106, 0.0294], device='cuda:0'), in_proj_covar=tensor([0.0080, 0.0121, 0.0104, 0.0076, 0.0099, 0.0111, 0.0086, 0.0136], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-15 17:19:15,768 INFO [finetune.py:992] (0/2) Epoch 1, batch 8900, loss[loss=0.199, simple_loss=0.2899, pruned_loss=0.05403, over 11132.00 frames. ], tot_loss[loss=0.1776, simple_loss=0.2645, pruned_loss=0.04533, over 2353044.33 frames. ], batch size: 55, lr: 4.99e-03, grad_scale: 16.0 2023-05-15 17:19:17,231 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.029e+02 3.094e+02 3.535e+02 4.369e+02 7.573e+02, threshold=7.071e+02, percent-clipped=1.0 2023-05-15 17:19:30,506 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.43 vs. limit=2.0 2023-05-15 17:19:38,481 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=108933.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 17:19:51,746 INFO [finetune.py:992] (0/2) Epoch 1, batch 8950, loss[loss=0.1783, simple_loss=0.2642, pruned_loss=0.04623, over 12034.00 frames. ], tot_loss[loss=0.1776, simple_loss=0.2642, pruned_loss=0.04547, over 2358484.79 frames. ], batch size: 31, lr: 4.99e-03, grad_scale: 16.0 2023-05-15 17:19:57,616 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.3535, 5.1387, 5.2392, 5.2913, 4.9313, 5.0008, 4.8456, 5.2388], device='cuda:0'), covar=tensor([0.0497, 0.0516, 0.0623, 0.0453, 0.1647, 0.1070, 0.0476, 0.0818], device='cuda:0'), in_proj_covar=tensor([0.0487, 0.0621, 0.0530, 0.0584, 0.0775, 0.0704, 0.0515, 0.0457], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0005, 0.0004, 0.0005, 0.0005, 0.0005, 0.0004, 0.0004], device='cuda:0') 2023-05-15 17:19:57,956 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.40 vs. limit=2.0 2023-05-15 17:20:12,390 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.0361, 3.9778, 4.0830, 4.5059, 3.2227, 3.9075, 2.6518, 4.1240], device='cuda:0'), covar=tensor([0.1746, 0.0784, 0.0885, 0.0541, 0.1017, 0.0652, 0.1777, 0.1196], device='cuda:0'), in_proj_covar=tensor([0.0220, 0.0261, 0.0299, 0.0359, 0.0241, 0.0238, 0.0255, 0.0396], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-15 17:20:12,924 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=108981.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 17:20:28,038 INFO [finetune.py:992] (0/2) Epoch 1, batch 9000, loss[loss=0.1857, simple_loss=0.2713, pruned_loss=0.05001, over 11314.00 frames. ], tot_loss[loss=0.1776, simple_loss=0.2645, pruned_loss=0.04533, over 2363230.71 frames. ], batch size: 55, lr: 4.99e-03, grad_scale: 16.0 2023-05-15 17:20:28,039 INFO [finetune.py:1017] (0/2) Computing validation loss 2023-05-15 17:20:41,271 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.9268, 4.8413, 4.9786, 4.9570, 4.5379, 4.6021, 4.6184, 4.8637], device='cuda:0'), covar=tensor([0.0679, 0.0487, 0.0588, 0.0462, 0.1902, 0.1312, 0.0485, 0.1016], device='cuda:0'), in_proj_covar=tensor([0.0492, 0.0626, 0.0533, 0.0589, 0.0780, 0.0710, 0.0518, 0.0460], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0005, 0.0004, 0.0005, 0.0005, 0.0005, 0.0004, 0.0004], device='cuda:0') 2023-05-15 17:20:41,939 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.1557, 3.2407, 4.5736, 2.4326, 2.6063, 3.6917, 2.9824, 3.7901], device='cuda:0'), covar=tensor([0.0540, 0.0983, 0.0245, 0.1184, 0.1829, 0.1147, 0.1410, 0.0803], device='cuda:0'), in_proj_covar=tensor([0.0234, 0.0229, 0.0239, 0.0177, 0.0235, 0.0283, 0.0226, 0.0251], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-15 17:20:46,577 INFO [finetune.py:1026] (0/2) Epoch 1, validation: loss=0.357, simple_loss=0.4207, pruned_loss=0.1467, over 1020973.00 frames. 2023-05-15 17:20:46,578 INFO [finetune.py:1027] (0/2) Maximum memory allocated so far is 12608MB 2023-05-15 17:20:48,013 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.087e+02 2.872e+02 3.471e+02 4.611e+02 1.446e+03, threshold=6.943e+02, percent-clipped=2.0 2023-05-15 17:20:49,135 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-05-15 17:21:06,545 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=109029.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 17:21:12,407 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=109037.0, num_to_drop=1, layers_to_drop={0} 2023-05-15 17:21:22,768 INFO [finetune.py:992] (0/2) Epoch 1, batch 9050, loss[loss=0.1831, simple_loss=0.271, pruned_loss=0.04755, over 12107.00 frames. ], tot_loss[loss=0.1779, simple_loss=0.2649, pruned_loss=0.04549, over 2359442.92 frames. ], batch size: 33, lr: 4.99e-03, grad_scale: 16.0 2023-05-15 17:21:49,866 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=109088.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 17:21:56,924 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=109098.0, num_to_drop=1, layers_to_drop={2} 2023-05-15 17:21:58,826 INFO [finetune.py:992] (0/2) Epoch 1, batch 9100, loss[loss=0.1839, simple_loss=0.277, pruned_loss=0.04544, over 12001.00 frames. ], tot_loss[loss=0.177, simple_loss=0.2641, pruned_loss=0.04497, over 2365066.74 frames. ], batch size: 40, lr: 4.99e-03, grad_scale: 16.0 2023-05-15 17:22:00,286 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.912e+02 3.137e+02 3.755e+02 4.480e+02 1.020e+03, threshold=7.511e+02, percent-clipped=5.0 2023-05-15 17:22:23,926 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=109136.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 17:22:31,605 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=109147.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 17:22:33,183 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.7220, 3.7087, 3.4335, 3.3971, 3.0370, 2.9541, 3.6492, 2.4083], device='cuda:0'), covar=tensor([0.0289, 0.0117, 0.0126, 0.0150, 0.0320, 0.0273, 0.0110, 0.0398], device='cuda:0'), in_proj_covar=tensor([0.0188, 0.0165, 0.0155, 0.0183, 0.0206, 0.0198, 0.0162, 0.0202], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-15 17:22:34,376 INFO [finetune.py:992] (0/2) Epoch 1, batch 9150, loss[loss=0.1623, simple_loss=0.2592, pruned_loss=0.03267, over 12355.00 frames. ], tot_loss[loss=0.177, simple_loss=0.2642, pruned_loss=0.04492, over 2358342.69 frames. ], batch size: 36, lr: 4.99e-03, grad_scale: 16.0 2023-05-15 17:22:41,819 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.3147, 4.5585, 2.7456, 2.4817, 3.9662, 2.6582, 3.9645, 2.9706], device='cuda:0'), covar=tensor([0.0689, 0.0611, 0.1196, 0.1630, 0.0242, 0.1237, 0.0416, 0.0827], device='cuda:0'), in_proj_covar=tensor([0.0195, 0.0256, 0.0179, 0.0202, 0.0143, 0.0183, 0.0195, 0.0177], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-15 17:22:52,714 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.57 vs. limit=2.0 2023-05-15 17:23:05,496 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=109194.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 17:23:10,694 INFO [finetune.py:992] (0/2) Epoch 1, batch 9200, loss[loss=0.1618, simple_loss=0.2538, pruned_loss=0.03489, over 12029.00 frames. ], tot_loss[loss=0.1764, simple_loss=0.2636, pruned_loss=0.04464, over 2360375.70 frames. ], batch size: 31, lr: 4.99e-03, grad_scale: 16.0 2023-05-15 17:23:12,100 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.921e+02 2.734e+02 3.400e+02 4.293e+02 1.083e+03, threshold=6.801e+02, percent-clipped=3.0 2023-05-15 17:23:34,125 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=109233.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 17:23:46,645 INFO [finetune.py:992] (0/2) Epoch 1, batch 9250, loss[loss=0.1545, simple_loss=0.2498, pruned_loss=0.02964, over 12290.00 frames. ], tot_loss[loss=0.1767, simple_loss=0.2642, pruned_loss=0.04461, over 2362395.44 frames. ], batch size: 34, lr: 4.99e-03, grad_scale: 16.0 2023-05-15 17:23:49,832 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=109255.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 17:24:04,673 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.6323, 3.3358, 5.0841, 2.8036, 2.6488, 3.8606, 3.2506, 3.9478], device='cuda:0'), covar=tensor([0.0538, 0.1105, 0.0251, 0.1110, 0.1915, 0.1269, 0.1373, 0.1066], device='cuda:0'), in_proj_covar=tensor([0.0234, 0.0228, 0.0240, 0.0177, 0.0235, 0.0283, 0.0227, 0.0251], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-15 17:24:08,071 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=109281.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 17:24:17,285 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=109294.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 17:24:22,164 INFO [finetune.py:992] (0/2) Epoch 1, batch 9300, loss[loss=0.1601, simple_loss=0.2409, pruned_loss=0.03968, over 12339.00 frames. ], tot_loss[loss=0.1755, simple_loss=0.2632, pruned_loss=0.04392, over 2368710.63 frames. ], batch size: 30, lr: 4.99e-03, grad_scale: 16.0 2023-05-15 17:24:23,559 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.037e+02 2.953e+02 3.400e+02 3.968e+02 5.850e+02, threshold=6.799e+02, percent-clipped=0.0 2023-05-15 17:24:34,379 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=109318.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 17:24:58,224 INFO [finetune.py:992] (0/2) Epoch 1, batch 9350, loss[loss=0.1581, simple_loss=0.2324, pruned_loss=0.04189, over 12012.00 frames. ], tot_loss[loss=0.1743, simple_loss=0.2618, pruned_loss=0.04337, over 2378420.07 frames. ], batch size: 28, lr: 4.99e-03, grad_scale: 8.0 2023-05-15 17:25:01,381 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.5929, 2.7246, 3.3457, 4.3869, 2.5511, 4.5519, 4.5007, 4.7249], device='cuda:0'), covar=tensor([0.0085, 0.1060, 0.0399, 0.0101, 0.1101, 0.0187, 0.0123, 0.0063], device='cuda:0'), in_proj_covar=tensor([0.0121, 0.0202, 0.0189, 0.0115, 0.0187, 0.0181, 0.0170, 0.0119], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-15 17:25:01,405 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=109355.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 17:25:18,363 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=109379.0, num_to_drop=1, layers_to_drop={2} 2023-05-15 17:25:28,647 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=109393.0, num_to_drop=1, layers_to_drop={1} 2023-05-15 17:25:34,456 INFO [finetune.py:992] (0/2) Epoch 1, batch 9400, loss[loss=0.1789, simple_loss=0.2685, pruned_loss=0.04467, over 11758.00 frames. ], tot_loss[loss=0.1744, simple_loss=0.2617, pruned_loss=0.04357, over 2381190.41 frames. ], batch size: 44, lr: 4.99e-03, grad_scale: 8.0 2023-05-15 17:25:36,642 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.940e+02 3.155e+02 3.540e+02 4.244e+02 9.825e+02, threshold=7.080e+02, percent-clipped=5.0 2023-05-15 17:25:59,200 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.53 vs. limit=2.0 2023-05-15 17:26:07,468 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=109447.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 17:26:10,163 INFO [finetune.py:992] (0/2) Epoch 1, batch 9450, loss[loss=0.1876, simple_loss=0.2831, pruned_loss=0.04604, over 11765.00 frames. ], tot_loss[loss=0.1742, simple_loss=0.2613, pruned_loss=0.0435, over 2389503.99 frames. ], batch size: 48, lr: 4.99e-03, grad_scale: 8.0 2023-05-15 17:26:42,321 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=109495.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 17:26:42,725 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.41 vs. limit=2.0 2023-05-15 17:26:46,330 INFO [finetune.py:992] (0/2) Epoch 1, batch 9500, loss[loss=0.1881, simple_loss=0.2798, pruned_loss=0.04821, over 10384.00 frames. ], tot_loss[loss=0.1727, simple_loss=0.2599, pruned_loss=0.04271, over 2391519.04 frames. ], batch size: 68, lr: 4.99e-03, grad_scale: 8.0 2023-05-15 17:26:48,424 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.809e+02 2.725e+02 3.238e+02 3.980e+02 6.999e+02, threshold=6.476e+02, percent-clipped=0.0 2023-05-15 17:27:22,053 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=109550.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 17:27:22,690 INFO [finetune.py:992] (0/2) Epoch 1, batch 9550, loss[loss=0.1584, simple_loss=0.2386, pruned_loss=0.03914, over 12284.00 frames. ], tot_loss[loss=0.1724, simple_loss=0.2598, pruned_loss=0.04251, over 2393052.34 frames. ], batch size: 28, lr: 4.99e-03, grad_scale: 8.0 2023-05-15 17:27:42,705 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=109579.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 17:27:58,354 INFO [finetune.py:992] (0/2) Epoch 1, batch 9600, loss[loss=0.1615, simple_loss=0.2556, pruned_loss=0.03369, over 12190.00 frames. ], tot_loss[loss=0.1734, simple_loss=0.261, pruned_loss=0.04293, over 2390421.94 frames. ], batch size: 35, lr: 4.99e-03, grad_scale: 8.0 2023-05-15 17:28:00,544 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.865e+02 2.787e+02 3.129e+02 4.200e+02 2.494e+03, threshold=6.258e+02, percent-clipped=6.0 2023-05-15 17:28:07,873 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.7888, 3.8351, 3.4202, 3.3476, 3.0315, 2.9791, 3.7615, 2.4974], device='cuda:0'), covar=tensor([0.0293, 0.0094, 0.0124, 0.0138, 0.0302, 0.0262, 0.0088, 0.0378], device='cuda:0'), in_proj_covar=tensor([0.0186, 0.0163, 0.0153, 0.0181, 0.0204, 0.0195, 0.0159, 0.0199], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-15 17:28:15,914 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2023-05-15 17:28:27,040 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=109640.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 17:28:34,087 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=109650.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 17:28:34,752 INFO [finetune.py:992] (0/2) Epoch 1, batch 9650, loss[loss=0.1506, simple_loss=0.2333, pruned_loss=0.03396, over 12188.00 frames. ], tot_loss[loss=0.1742, simple_loss=0.2618, pruned_loss=0.04333, over 2381041.50 frames. ], batch size: 31, lr: 4.99e-03, grad_scale: 8.0 2023-05-15 17:28:40,494 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=109659.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 17:28:51,826 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=109674.0, num_to_drop=1, layers_to_drop={0} 2023-05-15 17:28:59,542 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.2184, 5.9991, 5.5294, 5.5492, 6.1244, 5.4720, 5.6347, 5.5654], device='cuda:0'), covar=tensor([0.1401, 0.0951, 0.0803, 0.1936, 0.0851, 0.2000, 0.1828, 0.1005], device='cuda:0'), in_proj_covar=tensor([0.0330, 0.0466, 0.0366, 0.0414, 0.0444, 0.0418, 0.0376, 0.0365], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-15 17:29:05,360 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=109693.0, num_to_drop=1, layers_to_drop={1} 2023-05-15 17:29:10,836 INFO [finetune.py:992] (0/2) Epoch 1, batch 9700, loss[loss=0.201, simple_loss=0.2852, pruned_loss=0.05846, over 11614.00 frames. ], tot_loss[loss=0.1744, simple_loss=0.2616, pruned_loss=0.04361, over 2382342.22 frames. ], batch size: 48, lr: 4.99e-03, grad_scale: 8.0 2023-05-15 17:29:12,535 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2023-05-15 17:29:12,851 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.182e+02 3.051e+02 3.582e+02 4.337e+02 1.129e+03, threshold=7.164e+02, percent-clipped=10.0 2023-05-15 17:29:24,230 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=109720.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 17:29:38,952 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=109741.0, num_to_drop=1, layers_to_drop={0} 2023-05-15 17:29:46,029 INFO [finetune.py:992] (0/2) Epoch 1, batch 9750, loss[loss=0.1566, simple_loss=0.2518, pruned_loss=0.03072, over 12196.00 frames. ], tot_loss[loss=0.1756, simple_loss=0.2626, pruned_loss=0.04432, over 2372467.94 frames. ], batch size: 35, lr: 4.99e-03, grad_scale: 8.0 2023-05-15 17:30:02,858 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.2421, 5.2026, 4.9504, 5.1233, 4.7892, 5.2470, 5.1829, 5.4347], device='cuda:0'), covar=tensor([0.0160, 0.0111, 0.0192, 0.0222, 0.0646, 0.0171, 0.0117, 0.0125], device='cuda:0'), in_proj_covar=tensor([0.0183, 0.0187, 0.0186, 0.0230, 0.0234, 0.0201, 0.0173, 0.0222], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0005, 0.0005, 0.0005, 0.0004, 0.0005], device='cuda:0') 2023-05-15 17:30:22,015 INFO [finetune.py:992] (0/2) Epoch 1, batch 9800, loss[loss=0.1582, simple_loss=0.2403, pruned_loss=0.03806, over 12137.00 frames. ], tot_loss[loss=0.1767, simple_loss=0.2634, pruned_loss=0.04495, over 2375662.69 frames. ], batch size: 30, lr: 4.99e-03, grad_scale: 8.0 2023-05-15 17:30:24,132 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.803e+02 2.923e+02 3.465e+02 4.220e+02 9.009e+02, threshold=6.931e+02, percent-clipped=2.0 2023-05-15 17:30:29,434 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=2.30 vs. limit=5.0 2023-05-15 17:30:37,043 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.9154, 2.1454, 3.4113, 2.9870, 3.3412, 3.0175, 2.1767, 3.4347], device='cuda:0'), covar=tensor([0.0119, 0.0355, 0.0125, 0.0186, 0.0130, 0.0158, 0.0333, 0.0104], device='cuda:0'), in_proj_covar=tensor([0.0156, 0.0191, 0.0169, 0.0174, 0.0191, 0.0150, 0.0181, 0.0162], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-15 17:30:57,255 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=109850.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 17:30:57,890 INFO [finetune.py:992] (0/2) Epoch 1, batch 9850, loss[loss=0.1949, simple_loss=0.2848, pruned_loss=0.05257, over 10332.00 frames. ], tot_loss[loss=0.1754, simple_loss=0.2626, pruned_loss=0.04407, over 2385167.68 frames. ], batch size: 68, lr: 4.99e-03, grad_scale: 8.0 2023-05-15 17:31:27,969 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.42 vs. limit=2.0 2023-05-15 17:31:31,187 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=109898.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 17:31:33,195 INFO [finetune.py:992] (0/2) Epoch 1, batch 9900, loss[loss=0.1616, simple_loss=0.2477, pruned_loss=0.03778, over 12130.00 frames. ], tot_loss[loss=0.1753, simple_loss=0.2626, pruned_loss=0.044, over 2374677.54 frames. ], batch size: 30, lr: 4.99e-03, grad_scale: 8.0 2023-05-15 17:31:34,017 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.6548, 2.7877, 3.8220, 4.8494, 4.1538, 4.7609, 4.1316, 3.3114], device='cuda:0'), covar=tensor([0.0027, 0.0327, 0.0096, 0.0019, 0.0088, 0.0046, 0.0072, 0.0286], device='cuda:0'), in_proj_covar=tensor([0.0079, 0.0120, 0.0103, 0.0074, 0.0099, 0.0110, 0.0085, 0.0135], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-15 17:31:35,261 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.138e+02 2.877e+02 3.253e+02 4.045e+02 6.927e+02, threshold=6.506e+02, percent-clipped=0.0 2023-05-15 17:31:57,919 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=109935.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 17:32:08,863 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=109950.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 17:32:09,474 INFO [finetune.py:992] (0/2) Epoch 1, batch 9950, loss[loss=0.1746, simple_loss=0.2637, pruned_loss=0.04277, over 12046.00 frames. ], tot_loss[loss=0.1753, simple_loss=0.2624, pruned_loss=0.0441, over 2374886.76 frames. ], batch size: 37, lr: 4.99e-03, grad_scale: 8.0 2023-05-15 17:32:26,681 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=109974.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 17:32:43,525 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=109998.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 17:32:45,212 INFO [checkpoint.py:75] (0/2) Saving checkpoint to pruned_transducer_stateless7/exp_giga_finetune/checkpoint-10000.pt 2023-05-15 17:32:49,157 INFO [finetune.py:992] (0/2) Epoch 1, batch 10000, loss[loss=0.2618, simple_loss=0.3296, pruned_loss=0.09705, over 8470.00 frames. ], tot_loss[loss=0.1763, simple_loss=0.2632, pruned_loss=0.04466, over 2367418.09 frames. ], batch size: 97, lr: 4.99e-03, grad_scale: 8.0 2023-05-15 17:32:51,301 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.780e+02 2.816e+02 3.590e+02 4.301e+02 9.697e+02, threshold=7.180e+02, percent-clipped=4.0 2023-05-15 17:32:59,274 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=110015.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 17:33:00,816 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=110017.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 17:33:04,260 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=110022.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 17:33:06,764 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.42 vs. limit=2.0 2023-05-15 17:33:18,604 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.2557, 4.9887, 5.2207, 5.1651, 5.0403, 5.1697, 5.0087, 3.3436], device='cuda:0'), covar=tensor([0.0091, 0.0059, 0.0057, 0.0050, 0.0037, 0.0077, 0.0074, 0.0452], device='cuda:0'), in_proj_covar=tensor([0.0065, 0.0075, 0.0078, 0.0071, 0.0059, 0.0087, 0.0077, 0.0091], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0004], device='cuda:0') 2023-05-15 17:33:20,068 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=110044.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 17:33:24,893 INFO [finetune.py:992] (0/2) Epoch 1, batch 10050, loss[loss=0.1524, simple_loss=0.2462, pruned_loss=0.02927, over 12296.00 frames. ], tot_loss[loss=0.1765, simple_loss=0.2634, pruned_loss=0.04484, over 2360190.61 frames. ], batch size: 33, lr: 4.99e-03, grad_scale: 8.0 2023-05-15 17:33:26,018 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=3.40 vs. limit=5.0 2023-05-15 17:33:44,470 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=110078.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 17:34:01,142 INFO [finetune.py:992] (0/2) Epoch 1, batch 10100, loss[loss=0.1735, simple_loss=0.2721, pruned_loss=0.03748, over 12276.00 frames. ], tot_loss[loss=0.1768, simple_loss=0.2636, pruned_loss=0.04495, over 2364461.67 frames. ], batch size: 37, lr: 4.99e-03, grad_scale: 8.0 2023-05-15 17:34:03,242 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.129e+02 2.928e+02 3.602e+02 4.397e+02 7.962e+02, threshold=7.204e+02, percent-clipped=1.0 2023-05-15 17:34:04,140 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=110105.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 17:34:07,654 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=110110.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 17:34:11,327 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=110115.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 17:34:14,031 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.1890, 4.8847, 4.9850, 5.0760, 4.7955, 4.9984, 4.9455, 3.1242], device='cuda:0'), covar=tensor([0.0069, 0.0063, 0.0062, 0.0049, 0.0052, 0.0080, 0.0066, 0.0521], device='cuda:0'), in_proj_covar=tensor([0.0064, 0.0074, 0.0078, 0.0071, 0.0059, 0.0086, 0.0076, 0.0090], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0004], device='cuda:0') 2023-05-15 17:34:35,968 INFO [finetune.py:992] (0/2) Epoch 1, batch 10150, loss[loss=0.1411, simple_loss=0.2227, pruned_loss=0.02974, over 12179.00 frames. ], tot_loss[loss=0.1762, simple_loss=0.2631, pruned_loss=0.04464, over 2366695.75 frames. ], batch size: 29, lr: 4.99e-03, grad_scale: 8.0 2023-05-15 17:34:37,798 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.90 vs. limit=2.0 2023-05-15 17:34:50,606 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=110171.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 17:34:54,103 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=110176.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 17:35:12,213 INFO [finetune.py:992] (0/2) Epoch 1, batch 10200, loss[loss=0.151, simple_loss=0.2346, pruned_loss=0.0337, over 12184.00 frames. ], tot_loss[loss=0.1756, simple_loss=0.2627, pruned_loss=0.04422, over 2373790.67 frames. ], batch size: 31, lr: 4.99e-03, grad_scale: 8.0 2023-05-15 17:35:14,275 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.910e+02 2.856e+02 3.463e+02 4.178e+02 6.748e+02, threshold=6.925e+02, percent-clipped=0.0 2023-05-15 17:35:36,335 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=110235.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 17:35:47,037 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.4507, 5.2423, 5.3448, 5.4054, 5.0145, 5.0877, 4.9424, 5.3508], device='cuda:0'), covar=tensor([0.0564, 0.0556, 0.0607, 0.0506, 0.1870, 0.1073, 0.0524, 0.0922], device='cuda:0'), in_proj_covar=tensor([0.0484, 0.0619, 0.0530, 0.0590, 0.0768, 0.0705, 0.0513, 0.0461], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0005, 0.0004, 0.0005, 0.0005, 0.0005, 0.0004, 0.0004], device='cuda:0') 2023-05-15 17:35:48,306 INFO [finetune.py:992] (0/2) Epoch 1, batch 10250, loss[loss=0.1722, simple_loss=0.2638, pruned_loss=0.04028, over 12277.00 frames. ], tot_loss[loss=0.1761, simple_loss=0.2635, pruned_loss=0.04436, over 2369165.24 frames. ], batch size: 33, lr: 4.99e-03, grad_scale: 8.0 2023-05-15 17:35:57,604 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([6.0713, 6.0396, 5.8358, 5.3715, 5.2180, 6.0274, 5.6110, 5.3671], device='cuda:0'), covar=tensor([0.0599, 0.0809, 0.0565, 0.1257, 0.0574, 0.0610, 0.1148, 0.0959], device='cuda:0'), in_proj_covar=tensor([0.0574, 0.0507, 0.0457, 0.0581, 0.0370, 0.0654, 0.0715, 0.0520], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0004, 0.0004, 0.0003], device='cuda:0') 2023-05-15 17:36:11,086 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=110283.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 17:36:11,213 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=110283.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 17:36:18,087 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=3.26 vs. limit=5.0 2023-05-15 17:36:23,711 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=3.15 vs. limit=5.0 2023-05-15 17:36:23,999 INFO [finetune.py:992] (0/2) Epoch 1, batch 10300, loss[loss=0.1957, simple_loss=0.2818, pruned_loss=0.05483, over 12075.00 frames. ], tot_loss[loss=0.1754, simple_loss=0.2629, pruned_loss=0.04394, over 2369859.60 frames. ], batch size: 42, lr: 4.99e-03, grad_scale: 8.0 2023-05-15 17:36:25,565 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=110303.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 17:36:26,050 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.644e+02 2.961e+02 3.396e+02 3.978e+02 7.164e+02, threshold=6.793e+02, percent-clipped=1.0 2023-05-15 17:36:34,000 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=110315.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 17:36:36,082 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=110318.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 17:36:55,359 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=110344.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 17:37:00,260 INFO [finetune.py:992] (0/2) Epoch 1, batch 10350, loss[loss=0.1738, simple_loss=0.27, pruned_loss=0.03873, over 12279.00 frames. ], tot_loss[loss=0.1761, simple_loss=0.2636, pruned_loss=0.04433, over 2367489.04 frames. ], batch size: 37, lr: 4.99e-03, grad_scale: 8.0 2023-05-15 17:37:08,667 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=110363.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 17:37:09,521 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=110364.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 17:37:15,798 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=110373.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 17:37:20,130 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=110379.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 17:37:23,638 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.6765, 2.5177, 4.5540, 4.8567, 3.0978, 2.6673, 2.9364, 1.7654], device='cuda:0'), covar=tensor([0.1417, 0.3295, 0.0421, 0.0316, 0.0978, 0.1990, 0.2488, 0.4754], device='cuda:0'), in_proj_covar=tensor([0.0276, 0.0367, 0.0266, 0.0287, 0.0248, 0.0274, 0.0349, 0.0345], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-15 17:37:35,888 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=110400.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 17:37:36,488 INFO [finetune.py:992] (0/2) Epoch 1, batch 10400, loss[loss=0.1519, simple_loss=0.2285, pruned_loss=0.03767, over 12313.00 frames. ], tot_loss[loss=0.1763, simple_loss=0.2638, pruned_loss=0.04439, over 2366011.90 frames. ], batch size: 28, lr: 4.99e-03, grad_scale: 8.0 2023-05-15 17:37:38,661 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.930e+02 3.067e+02 3.653e+02 4.594e+02 8.110e+02, threshold=7.306e+02, percent-clipped=3.0 2023-05-15 17:37:47,173 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2023-05-15 17:37:56,831 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.4387, 3.6134, 3.8233, 4.3137, 3.0643, 3.6130, 2.3907, 3.9335], device='cuda:0'), covar=tensor([0.1504, 0.0953, 0.1175, 0.0752, 0.1108, 0.0791, 0.1956, 0.1089], device='cuda:0'), in_proj_covar=tensor([0.0219, 0.0259, 0.0294, 0.0356, 0.0238, 0.0238, 0.0253, 0.0395], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-15 17:38:11,408 INFO [finetune.py:992] (0/2) Epoch 1, batch 10450, loss[loss=0.1683, simple_loss=0.2438, pruned_loss=0.04638, over 12173.00 frames. ], tot_loss[loss=0.1762, simple_loss=0.2634, pruned_loss=0.04454, over 2366460.00 frames. ], batch size: 31, lr: 4.99e-03, grad_scale: 8.0 2023-05-15 17:38:21,840 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=110466.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 17:38:25,342 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=110471.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 17:38:43,620 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.90 vs. limit=2.0 2023-05-15 17:38:47,248 INFO [finetune.py:992] (0/2) Epoch 1, batch 10500, loss[loss=0.184, simple_loss=0.2706, pruned_loss=0.04867, over 12153.00 frames. ], tot_loss[loss=0.175, simple_loss=0.2622, pruned_loss=0.04393, over 2373308.92 frames. ], batch size: 34, lr: 4.99e-03, grad_scale: 8.0 2023-05-15 17:38:49,441 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.763e+02 3.055e+02 3.705e+02 4.613e+02 1.350e+03, threshold=7.409e+02, percent-clipped=4.0 2023-05-15 17:39:06,106 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.35 vs. limit=2.0 2023-05-15 17:39:22,907 INFO [finetune.py:992] (0/2) Epoch 1, batch 10550, loss[loss=0.2956, simple_loss=0.3659, pruned_loss=0.1127, over 8241.00 frames. ], tot_loss[loss=0.1762, simple_loss=0.2633, pruned_loss=0.04456, over 2370072.83 frames. ], batch size: 98, lr: 4.99e-03, grad_scale: 8.0 2023-05-15 17:39:58,882 INFO [finetune.py:992] (0/2) Epoch 1, batch 10600, loss[loss=0.1656, simple_loss=0.2484, pruned_loss=0.04137, over 12183.00 frames. ], tot_loss[loss=0.1765, simple_loss=0.2635, pruned_loss=0.04474, over 2360517.73 frames. ], batch size: 31, lr: 4.99e-03, grad_scale: 8.0 2023-05-15 17:40:00,987 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.036e+02 2.797e+02 3.485e+02 4.135e+02 9.594e+02, threshold=6.971e+02, percent-clipped=1.0 2023-05-15 17:40:06,174 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=110611.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 17:40:06,800 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.5600, 5.1512, 4.8094, 4.6997, 5.2875, 4.6387, 4.8859, 4.6999], device='cuda:0'), covar=tensor([0.1346, 0.0919, 0.1023, 0.1901, 0.0927, 0.1840, 0.1415, 0.1129], device='cuda:0'), in_proj_covar=tensor([0.0329, 0.0462, 0.0363, 0.0417, 0.0444, 0.0413, 0.0374, 0.0359], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-15 17:40:26,982 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=110639.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 17:40:35,341 INFO [finetune.py:992] (0/2) Epoch 1, batch 10650, loss[loss=0.1448, simple_loss=0.2244, pruned_loss=0.03257, over 12001.00 frames. ], tot_loss[loss=0.1752, simple_loss=0.2623, pruned_loss=0.04403, over 2364279.84 frames. ], batch size: 28, lr: 4.99e-03, grad_scale: 8.0 2023-05-15 17:40:41,031 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=110659.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 17:40:50,346 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=110672.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 17:40:50,914 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=110673.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 17:40:51,619 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=110674.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 17:41:10,647 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=110700.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 17:41:11,215 INFO [finetune.py:992] (0/2) Epoch 1, batch 10700, loss[loss=0.1823, simple_loss=0.2654, pruned_loss=0.04965, over 12238.00 frames. ], tot_loss[loss=0.1767, simple_loss=0.2637, pruned_loss=0.04488, over 2358052.83 frames. ], batch size: 32, lr: 4.99e-03, grad_scale: 8.0 2023-05-15 17:41:13,391 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.622e+02 2.805e+02 3.521e+02 4.283e+02 1.056e+03, threshold=7.043e+02, percent-clipped=6.0 2023-05-15 17:41:25,434 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=110721.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 17:41:36,283 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=110736.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 17:41:44,713 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=110748.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 17:41:46,800 INFO [finetune.py:992] (0/2) Epoch 1, batch 10750, loss[loss=0.168, simple_loss=0.2593, pruned_loss=0.03831, over 12361.00 frames. ], tot_loss[loss=0.1765, simple_loss=0.2638, pruned_loss=0.04464, over 2358500.53 frames. ], batch size: 35, lr: 4.99e-03, grad_scale: 8.0 2023-05-15 17:41:57,964 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=110766.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 17:42:01,462 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=110771.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 17:42:20,593 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=110797.0, num_to_drop=1, layers_to_drop={1} 2023-05-15 17:42:23,504 INFO [finetune.py:992] (0/2) Epoch 1, batch 10800, loss[loss=0.1785, simple_loss=0.2676, pruned_loss=0.04468, over 12120.00 frames. ], tot_loss[loss=0.1761, simple_loss=0.2633, pruned_loss=0.04443, over 2360659.51 frames. ], batch size: 33, lr: 4.99e-03, grad_scale: 8.0 2023-05-15 17:42:25,616 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.652e+02 3.029e+02 3.519e+02 4.086e+02 7.256e+02, threshold=7.039e+02, percent-clipped=1.0 2023-05-15 17:42:31,516 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.4462, 3.5262, 3.2913, 3.2842, 2.8386, 2.6414, 3.5304, 2.2775], device='cuda:0'), covar=tensor([0.0346, 0.0146, 0.0155, 0.0142, 0.0355, 0.0315, 0.0111, 0.0426], device='cuda:0'), in_proj_covar=tensor([0.0184, 0.0160, 0.0150, 0.0178, 0.0202, 0.0193, 0.0156, 0.0196], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-15 17:42:32,743 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=110814.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 17:42:36,335 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=110819.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 17:42:53,288 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.82 vs. limit=2.0 2023-05-15 17:42:59,719 INFO [finetune.py:992] (0/2) Epoch 1, batch 10850, loss[loss=0.2071, simple_loss=0.2902, pruned_loss=0.062, over 11832.00 frames. ], tot_loss[loss=0.1768, simple_loss=0.264, pruned_loss=0.04482, over 2360085.44 frames. ], batch size: 44, lr: 4.99e-03, grad_scale: 8.0 2023-05-15 17:43:12,553 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=110868.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 17:43:36,033 INFO [finetune.py:992] (0/2) Epoch 1, batch 10900, loss[loss=0.1775, simple_loss=0.2761, pruned_loss=0.03947, over 12353.00 frames. ], tot_loss[loss=0.1778, simple_loss=0.2652, pruned_loss=0.04516, over 2360509.00 frames. ], batch size: 35, lr: 4.99e-03, grad_scale: 8.0 2023-05-15 17:43:38,038 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.219e+02 2.906e+02 3.302e+02 4.025e+02 7.029e+02, threshold=6.603e+02, percent-clipped=0.0 2023-05-15 17:43:43,300 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=2.53 vs. limit=5.0 2023-05-15 17:43:44,558 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.5888, 2.7254, 3.8485, 4.7073, 4.2186, 4.6521, 4.0082, 3.2470], device='cuda:0'), covar=tensor([0.0024, 0.0319, 0.0092, 0.0023, 0.0067, 0.0053, 0.0066, 0.0293], device='cuda:0'), in_proj_covar=tensor([0.0079, 0.0120, 0.0102, 0.0074, 0.0099, 0.0111, 0.0086, 0.0136], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-15 17:43:56,660 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=110929.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 17:44:03,662 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=110939.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 17:44:07,586 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.47 vs. limit=2.0 2023-05-15 17:44:12,065 INFO [finetune.py:992] (0/2) Epoch 1, batch 10950, loss[loss=0.2064, simple_loss=0.2933, pruned_loss=0.05972, over 12053.00 frames. ], tot_loss[loss=0.1774, simple_loss=0.2649, pruned_loss=0.045, over 2367462.63 frames. ], batch size: 42, lr: 4.99e-03, grad_scale: 8.0 2023-05-15 17:44:17,213 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=110958.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 17:44:17,815 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=110959.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 17:44:23,450 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=110967.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 17:44:29,129 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=110974.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 17:44:38,267 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=110987.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 17:44:48,268 INFO [finetune.py:992] (0/2) Epoch 1, batch 11000, loss[loss=0.1828, simple_loss=0.2599, pruned_loss=0.05285, over 12326.00 frames. ], tot_loss[loss=0.1793, simple_loss=0.2665, pruned_loss=0.04607, over 2360988.85 frames. ], batch size: 31, lr: 4.98e-03, grad_scale: 8.0 2023-05-15 17:44:50,411 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.024e+02 3.036e+02 3.496e+02 4.197e+02 8.188e+02, threshold=6.991e+02, percent-clipped=4.0 2023-05-15 17:44:52,569 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=111007.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 17:44:56,285 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.8359, 3.2459, 2.4714, 2.2254, 2.8758, 2.3041, 3.0731, 2.6167], device='cuda:0'), covar=tensor([0.0545, 0.0752, 0.0758, 0.1156, 0.0268, 0.0947, 0.0466, 0.0668], device='cuda:0'), in_proj_covar=tensor([0.0192, 0.0253, 0.0178, 0.0198, 0.0141, 0.0180, 0.0194, 0.0175], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-15 17:45:01,103 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=111019.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 17:45:03,068 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=111022.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 17:45:03,272 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.5191, 4.8573, 4.2151, 5.1443, 4.6735, 3.0814, 4.3930, 3.3189], device='cuda:0'), covar=tensor([0.0746, 0.0760, 0.1342, 0.0322, 0.1117, 0.1525, 0.0884, 0.2693], device='cuda:0'), in_proj_covar=tensor([0.0295, 0.0372, 0.0351, 0.0262, 0.0360, 0.0263, 0.0332, 0.0363], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-15 17:45:12,124 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.2185, 2.1272, 2.6278, 3.1627, 2.1846, 3.2939, 3.1683, 3.3204], device='cuda:0'), covar=tensor([0.0144, 0.0960, 0.0421, 0.0155, 0.0891, 0.0290, 0.0265, 0.0107], device='cuda:0'), in_proj_covar=tensor([0.0122, 0.0207, 0.0191, 0.0115, 0.0190, 0.0182, 0.0173, 0.0122], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-15 17:45:23,259 INFO [finetune.py:992] (0/2) Epoch 1, batch 11050, loss[loss=0.1821, simple_loss=0.2719, pruned_loss=0.04615, over 12199.00 frames. ], tot_loss[loss=0.1826, simple_loss=0.2692, pruned_loss=0.04801, over 2328279.71 frames. ], batch size: 35, lr: 4.98e-03, grad_scale: 8.0 2023-05-15 17:45:27,929 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=3.25 vs. limit=5.0 2023-05-15 17:45:54,094 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=111092.0, num_to_drop=1, layers_to_drop={2} 2023-05-15 17:46:00,251 INFO [finetune.py:992] (0/2) Epoch 1, batch 11100, loss[loss=0.181, simple_loss=0.2742, pruned_loss=0.04396, over 12306.00 frames. ], tot_loss[loss=0.1879, simple_loss=0.2735, pruned_loss=0.05111, over 2269746.01 frames. ], batch size: 34, lr: 4.98e-03, grad_scale: 8.0 2023-05-15 17:46:02,289 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.324e+02 3.326e+02 4.024e+02 5.027e+02 9.650e+02, threshold=8.047e+02, percent-clipped=3.0 2023-05-15 17:46:14,628 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.3395, 3.5360, 3.2950, 3.6560, 3.4997, 2.5802, 3.2834, 2.8729], device='cuda:0'), covar=tensor([0.0718, 0.0855, 0.1147, 0.0530, 0.0938, 0.1379, 0.0990, 0.2149], device='cuda:0'), in_proj_covar=tensor([0.0292, 0.0366, 0.0347, 0.0257, 0.0355, 0.0259, 0.0327, 0.0357], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-15 17:46:36,135 INFO [finetune.py:992] (0/2) Epoch 1, batch 11150, loss[loss=0.3207, simple_loss=0.372, pruned_loss=0.1347, over 6556.00 frames. ], tot_loss[loss=0.1948, simple_loss=0.2794, pruned_loss=0.05508, over 2228248.32 frames. ], batch size: 99, lr: 4.98e-03, grad_scale: 8.0 2023-05-15 17:47:02,904 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.73 vs. limit=2.0 2023-05-15 17:47:08,023 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([1.9491, 2.1547, 2.2550, 2.2503, 1.9011, 1.8977, 2.2437, 1.6529], device='cuda:0'), covar=tensor([0.0234, 0.0153, 0.0107, 0.0117, 0.0230, 0.0162, 0.0091, 0.0261], device='cuda:0'), in_proj_covar=tensor([0.0181, 0.0156, 0.0146, 0.0175, 0.0198, 0.0188, 0.0153, 0.0194], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], device='cuda:0') 2023-05-15 17:47:11,539 INFO [finetune.py:992] (0/2) Epoch 1, batch 11200, loss[loss=0.346, simple_loss=0.3833, pruned_loss=0.1544, over 6873.00 frames. ], tot_loss[loss=0.2071, simple_loss=0.2892, pruned_loss=0.06254, over 2153669.70 frames. ], batch size: 98, lr: 4.98e-03, grad_scale: 8.0 2023-05-15 17:47:12,672 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=3.11 vs. limit=5.0 2023-05-15 17:47:13,736 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.651e+02 3.669e+02 4.380e+02 5.080e+02 1.411e+03, threshold=8.760e+02, percent-clipped=3.0 2023-05-15 17:47:17,530 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=2.10 vs. limit=2.0 2023-05-15 17:47:28,128 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=111224.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 17:47:28,922 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=111225.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 17:47:47,220 INFO [finetune.py:992] (0/2) Epoch 1, batch 11250, loss[loss=0.2301, simple_loss=0.3223, pruned_loss=0.06888, over 11061.00 frames. ], tot_loss[loss=0.2163, simple_loss=0.297, pruned_loss=0.06777, over 2107767.62 frames. ], batch size: 55, lr: 4.98e-03, grad_scale: 8.0 2023-05-15 17:47:58,427 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=111267.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 17:47:58,854 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.34 vs. limit=2.0 2023-05-15 17:48:11,551 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=111286.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 17:48:22,095 INFO [finetune.py:992] (0/2) Epoch 1, batch 11300, loss[loss=0.3168, simple_loss=0.3679, pruned_loss=0.1328, over 6990.00 frames. ], tot_loss[loss=0.2246, simple_loss=0.3037, pruned_loss=0.07272, over 2053269.23 frames. ], batch size: 99, lr: 4.98e-03, grad_scale: 8.0 2023-05-15 17:48:24,238 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.301e+02 3.763e+02 4.460e+02 5.493e+02 1.220e+03, threshold=8.920e+02, percent-clipped=6.0 2023-05-15 17:48:31,905 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=111314.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 17:48:32,583 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=111315.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 17:48:53,372 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.32 vs. limit=2.0 2023-05-15 17:48:57,080 INFO [finetune.py:992] (0/2) Epoch 1, batch 11350, loss[loss=0.3282, simple_loss=0.3845, pruned_loss=0.1359, over 6575.00 frames. ], tot_loss[loss=0.2318, simple_loss=0.3099, pruned_loss=0.07686, over 1998623.68 frames. ], batch size: 102, lr: 4.98e-03, grad_scale: 16.0 2023-05-15 17:49:26,205 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=111392.0, num_to_drop=1, layers_to_drop={0} 2023-05-15 17:49:32,422 INFO [finetune.py:992] (0/2) Epoch 1, batch 11400, loss[loss=0.2938, simple_loss=0.3539, pruned_loss=0.1169, over 7077.00 frames. ], tot_loss[loss=0.2387, simple_loss=0.3152, pruned_loss=0.08104, over 1943882.53 frames. ], batch size: 100, lr: 4.98e-03, grad_scale: 16.0 2023-05-15 17:49:34,493 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.662e+02 3.744e+02 4.367e+02 5.298e+02 9.715e+02, threshold=8.735e+02, percent-clipped=1.0 2023-05-15 17:50:00,478 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=111440.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 17:50:05,484 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.46 vs. limit=2.0 2023-05-15 17:50:06,937 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.85 vs. limit=2.0 2023-05-15 17:50:07,907 INFO [finetune.py:992] (0/2) Epoch 1, batch 11450, loss[loss=0.1629, simple_loss=0.2447, pruned_loss=0.04056, over 12188.00 frames. ], tot_loss[loss=0.243, simple_loss=0.3182, pruned_loss=0.08392, over 1900254.86 frames. ], batch size: 29, lr: 4.98e-03, grad_scale: 16.0 2023-05-15 17:50:19,791 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=111468.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 17:50:42,433 INFO [finetune.py:992] (0/2) Epoch 1, batch 11500, loss[loss=0.2345, simple_loss=0.3175, pruned_loss=0.07579, over 11042.00 frames. ], tot_loss[loss=0.2481, simple_loss=0.3219, pruned_loss=0.08714, over 1847450.05 frames. ], batch size: 55, lr: 4.98e-03, grad_scale: 16.0 2023-05-15 17:50:44,445 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.242e+02 3.745e+02 4.269e+02 5.103e+02 9.433e+02, threshold=8.538e+02, percent-clipped=0.0 2023-05-15 17:50:56,108 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.1597, 2.3862, 3.5734, 2.9901, 3.3848, 3.1448, 2.2296, 3.5777], device='cuda:0'), covar=tensor([0.0100, 0.0313, 0.0078, 0.0198, 0.0108, 0.0148, 0.0368, 0.0084], device='cuda:0'), in_proj_covar=tensor([0.0146, 0.0180, 0.0155, 0.0163, 0.0177, 0.0141, 0.0171, 0.0149], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-15 17:50:56,312 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 2023-05-15 17:50:58,785 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=111524.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 17:51:02,194 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=111529.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 17:51:16,795 INFO [finetune.py:992] (0/2) Epoch 1, batch 11550, loss[loss=0.2724, simple_loss=0.3369, pruned_loss=0.1039, over 7020.00 frames. ], tot_loss[loss=0.2544, simple_loss=0.326, pruned_loss=0.09143, over 1793996.37 frames. ], batch size: 99, lr: 4.98e-03, grad_scale: 8.0 2023-05-15 17:51:26,199 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.3289, 3.4661, 3.2335, 3.5642, 3.4043, 2.5890, 3.2862, 2.8695], device='cuda:0'), covar=tensor([0.0808, 0.0897, 0.1187, 0.0544, 0.1191, 0.1452, 0.1100, 0.2510], device='cuda:0'), in_proj_covar=tensor([0.0279, 0.0347, 0.0327, 0.0239, 0.0338, 0.0248, 0.0310, 0.0342], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-15 17:51:32,047 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=111572.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 17:51:38,237 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=111581.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 17:51:52,045 INFO [finetune.py:992] (0/2) Epoch 1, batch 11600, loss[loss=0.3088, simple_loss=0.3643, pruned_loss=0.1267, over 6834.00 frames. ], tot_loss[loss=0.2568, simple_loss=0.3275, pruned_loss=0.09309, over 1767930.02 frames. ], batch size: 99, lr: 4.98e-03, grad_scale: 8.0 2023-05-15 17:51:54,673 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.870e+02 3.662e+02 4.318e+02 5.254e+02 1.085e+03, threshold=8.637e+02, percent-clipped=4.0 2023-05-15 17:52:01,746 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=111614.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 17:52:05,761 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.7713, 3.6796, 3.6783, 3.8043, 3.3259, 3.7749, 3.8180, 3.8860], device='cuda:0'), covar=tensor([0.0220, 0.0182, 0.0198, 0.0238, 0.0798, 0.0337, 0.0165, 0.0233], device='cuda:0'), in_proj_covar=tensor([0.0158, 0.0160, 0.0159, 0.0196, 0.0203, 0.0172, 0.0149, 0.0192], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003, 0.0005], device='cuda:0') 2023-05-15 17:52:15,057 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.8031, 3.1576, 2.5037, 2.2576, 2.8717, 2.3445, 3.0753, 2.5893], device='cuda:0'), covar=tensor([0.0484, 0.0565, 0.0854, 0.1306, 0.0233, 0.1002, 0.0431, 0.0789], device='cuda:0'), in_proj_covar=tensor([0.0182, 0.0234, 0.0170, 0.0191, 0.0132, 0.0174, 0.0183, 0.0169], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-15 17:52:28,660 INFO [finetune.py:992] (0/2) Epoch 1, batch 11650, loss[loss=0.2967, simple_loss=0.3565, pruned_loss=0.1185, over 6798.00 frames. ], tot_loss[loss=0.2571, simple_loss=0.3271, pruned_loss=0.09354, over 1742791.98 frames. ], batch size: 100, lr: 4.98e-03, grad_scale: 8.0 2023-05-15 17:52:36,720 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=111662.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 17:52:48,809 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.7584, 3.1352, 2.4276, 2.1607, 2.8339, 2.2836, 3.0437, 2.5534], device='cuda:0'), covar=tensor([0.0512, 0.0634, 0.0871, 0.1385, 0.0235, 0.1132, 0.0428, 0.0809], device='cuda:0'), in_proj_covar=tensor([0.0184, 0.0235, 0.0171, 0.0193, 0.0133, 0.0176, 0.0183, 0.0171], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-15 17:53:03,927 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.44 vs. limit=2.0 2023-05-15 17:53:04,045 INFO [finetune.py:992] (0/2) Epoch 1, batch 11700, loss[loss=0.2924, simple_loss=0.3438, pruned_loss=0.1205, over 6964.00 frames. ], tot_loss[loss=0.2553, simple_loss=0.3256, pruned_loss=0.09248, over 1751669.38 frames. ], batch size: 100, lr: 4.98e-03, grad_scale: 8.0 2023-05-15 17:53:06,689 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.650e+02 3.667e+02 3.963e+02 4.752e+02 8.265e+02, threshold=7.925e+02, percent-clipped=0.0 2023-05-15 17:53:28,965 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.6492, 3.5759, 3.5441, 3.6869, 3.3873, 3.6513, 3.6628, 3.7544], device='cuda:0'), covar=tensor([0.0230, 0.0170, 0.0205, 0.0251, 0.0561, 0.0296, 0.0195, 0.0228], device='cuda:0'), in_proj_covar=tensor([0.0157, 0.0160, 0.0158, 0.0196, 0.0202, 0.0171, 0.0148, 0.0191], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003, 0.0005], device='cuda:0') 2023-05-15 17:53:38,229 INFO [finetune.py:992] (0/2) Epoch 1, batch 11750, loss[loss=0.2375, simple_loss=0.322, pruned_loss=0.07648, over 10267.00 frames. ], tot_loss[loss=0.2566, simple_loss=0.3261, pruned_loss=0.09355, over 1727286.29 frames. ], batch size: 68, lr: 4.98e-03, grad_scale: 8.0 2023-05-15 17:54:04,577 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.9776, 2.8935, 4.4247, 2.4548, 2.5563, 3.5297, 3.0039, 3.6330], device='cuda:0'), covar=tensor([0.0567, 0.1471, 0.0229, 0.1346, 0.2158, 0.1161, 0.1518, 0.0838], device='cuda:0'), in_proj_covar=tensor([0.0221, 0.0218, 0.0217, 0.0172, 0.0226, 0.0263, 0.0215, 0.0233], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:0') 2023-05-15 17:54:06,465 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=111791.0, num_to_drop=1, layers_to_drop={1} 2023-05-15 17:54:13,728 INFO [finetune.py:992] (0/2) Epoch 1, batch 11800, loss[loss=0.3868, simple_loss=0.4008, pruned_loss=0.1863, over 6061.00 frames. ], tot_loss[loss=0.2604, simple_loss=0.3291, pruned_loss=0.09588, over 1712532.11 frames. ], batch size: 99, lr: 4.98e-03, grad_scale: 8.0 2023-05-15 17:54:16,478 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 3.032e+02 3.813e+02 4.515e+02 5.504e+02 1.380e+03, threshold=9.030e+02, percent-clipped=6.0 2023-05-15 17:54:29,029 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.2487, 4.8925, 5.2150, 4.6884, 5.0231, 4.7913, 5.2395, 4.8993], device='cuda:0'), covar=tensor([0.0258, 0.0320, 0.0259, 0.0242, 0.0281, 0.0235, 0.0197, 0.0243], device='cuda:0'), in_proj_covar=tensor([0.0202, 0.0213, 0.0227, 0.0208, 0.0208, 0.0207, 0.0190, 0.0175], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-15 17:54:30,321 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=111824.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 17:54:46,542 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.4004, 2.8366, 3.6989, 2.3315, 2.5710, 3.0882, 2.8929, 3.1537], device='cuda:0'), covar=tensor([0.0532, 0.1116, 0.0300, 0.1271, 0.1759, 0.1117, 0.1119, 0.0952], device='cuda:0'), in_proj_covar=tensor([0.0218, 0.0216, 0.0214, 0.0170, 0.0224, 0.0260, 0.0213, 0.0230], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:0') 2023-05-15 17:54:48,314 INFO [finetune.py:992] (0/2) Epoch 1, batch 11850, loss[loss=0.2148, simple_loss=0.3102, pruned_loss=0.05977, over 10296.00 frames. ], tot_loss[loss=0.2616, simple_loss=0.3306, pruned_loss=0.09637, over 1696654.69 frames. ], batch size: 68, lr: 4.98e-03, grad_scale: 8.0 2023-05-15 17:54:49,152 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=111852.0, num_to_drop=1, layers_to_drop={2} 2023-05-15 17:55:10,086 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=111881.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 17:55:12,962 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.7380, 3.0935, 2.4371, 2.1756, 2.7111, 2.2950, 2.9769, 2.5745], device='cuda:0'), covar=tensor([0.0606, 0.0715, 0.0892, 0.1425, 0.0350, 0.1153, 0.0522, 0.0792], device='cuda:0'), in_proj_covar=tensor([0.0183, 0.0233, 0.0171, 0.0193, 0.0131, 0.0176, 0.0182, 0.0171], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-15 17:55:16,985 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.3693, 3.1528, 3.1755, 3.4535, 2.8920, 3.1374, 2.5674, 2.8372], device='cuda:0'), covar=tensor([0.1597, 0.0720, 0.0776, 0.0483, 0.0906, 0.0713, 0.1521, 0.0750], device='cuda:0'), in_proj_covar=tensor([0.0215, 0.0249, 0.0283, 0.0337, 0.0230, 0.0227, 0.0246, 0.0371], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-15 17:55:23,524 INFO [finetune.py:992] (0/2) Epoch 1, batch 11900, loss[loss=0.3127, simple_loss=0.355, pruned_loss=0.1351, over 7023.00 frames. ], tot_loss[loss=0.2608, simple_loss=0.3303, pruned_loss=0.09563, over 1676639.68 frames. ], batch size: 100, lr: 4.98e-03, grad_scale: 8.0 2023-05-15 17:55:26,249 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.186e+02 3.677e+02 4.314e+02 4.976e+02 1.689e+03, threshold=8.628e+02, percent-clipped=4.0 2023-05-15 17:55:28,554 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.5444, 4.4819, 4.3950, 4.0831, 4.1676, 4.5555, 4.2475, 4.2030], device='cuda:0'), covar=tensor([0.0786, 0.0890, 0.0654, 0.1124, 0.2182, 0.0691, 0.1343, 0.0762], device='cuda:0'), in_proj_covar=tensor([0.0541, 0.0472, 0.0433, 0.0543, 0.0356, 0.0604, 0.0653, 0.0484], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-05-15 17:55:43,163 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=111929.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 17:55:58,572 INFO [finetune.py:992] (0/2) Epoch 1, batch 11950, loss[loss=0.2025, simple_loss=0.2858, pruned_loss=0.05963, over 7283.00 frames. ], tot_loss[loss=0.2555, simple_loss=0.3265, pruned_loss=0.09225, over 1667075.42 frames. ], batch size: 99, lr: 4.98e-03, grad_scale: 8.0 2023-05-15 17:56:33,078 INFO [checkpoint.py:75] (0/2) Saving checkpoint to pruned_transducer_stateless7/exp_giga_finetune/checkpoint-12000.pt 2023-05-15 17:56:37,210 INFO [finetune.py:992] (0/2) Epoch 1, batch 12000, loss[loss=0.2031, simple_loss=0.2926, pruned_loss=0.0568, over 10496.00 frames. ], tot_loss[loss=0.2492, simple_loss=0.3216, pruned_loss=0.08835, over 1664514.89 frames. ], batch size: 68, lr: 4.98e-03, grad_scale: 8.0 2023-05-15 17:56:37,211 INFO [finetune.py:1017] (0/2) Computing validation loss 2023-05-15 17:56:43,348 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.0381, 3.7003, 3.8570, 4.2334, 3.0836, 3.8170, 2.4050, 3.6859], device='cuda:0'), covar=tensor([0.2126, 0.0982, 0.1091, 0.0549, 0.1211, 0.0788, 0.2308, 0.1477], device='cuda:0'), in_proj_covar=tensor([0.0216, 0.0250, 0.0284, 0.0337, 0.0230, 0.0228, 0.0247, 0.0371], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-15 17:56:53,329 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.1237, 2.6806, 3.4404, 2.1840, 2.4699, 3.0320, 2.6936, 3.0865], device='cuda:0'), covar=tensor([0.0515, 0.0965, 0.0258, 0.1116, 0.1519, 0.0898, 0.0978, 0.0706], device='cuda:0'), in_proj_covar=tensor([0.0213, 0.0212, 0.0208, 0.0168, 0.0219, 0.0253, 0.0208, 0.0225], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:0') 2023-05-15 17:56:55,919 INFO [finetune.py:1026] (0/2) Epoch 1, validation: loss=0.2946, simple_loss=0.3701, pruned_loss=0.1095, over 1020973.00 frames. 2023-05-15 17:56:55,920 INFO [finetune.py:1027] (0/2) Maximum memory allocated so far is 12745MB 2023-05-15 17:56:58,670 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.354e+02 3.236e+02 3.961e+02 5.158e+02 1.207e+03, threshold=7.921e+02, percent-clipped=1.0 2023-05-15 17:57:30,934 INFO [finetune.py:992] (0/2) Epoch 1, batch 12050, loss[loss=0.2161, simple_loss=0.3086, pruned_loss=0.06182, over 11787.00 frames. ], tot_loss[loss=0.2433, simple_loss=0.3173, pruned_loss=0.08462, over 1677676.39 frames. ], batch size: 44, lr: 4.98e-03, grad_scale: 8.0 2023-05-15 17:57:40,916 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=112066.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 17:57:51,014 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([1.9913, 2.1751, 2.2475, 2.2733, 1.9892, 1.7902, 2.2676, 1.7533], device='cuda:0'), covar=tensor([0.0280, 0.0142, 0.0137, 0.0150, 0.0274, 0.0234, 0.0117, 0.0344], device='cuda:0'), in_proj_covar=tensor([0.0171, 0.0145, 0.0136, 0.0164, 0.0184, 0.0179, 0.0141, 0.0184], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], device='cuda:0') 2023-05-15 17:58:03,541 INFO [finetune.py:992] (0/2) Epoch 1, batch 12100, loss[loss=0.2494, simple_loss=0.311, pruned_loss=0.09387, over 7116.00 frames. ], tot_loss[loss=0.2415, simple_loss=0.3159, pruned_loss=0.08356, over 1670348.52 frames. ], batch size: 100, lr: 4.98e-03, grad_scale: 8.0 2023-05-15 17:58:03,987 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.36 vs. limit=2.0 2023-05-15 17:58:06,084 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.361e+02 3.222e+02 3.757e+02 4.693e+02 1.119e+03, threshold=7.514e+02, percent-clipped=4.0 2023-05-15 17:58:18,338 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=112124.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 17:58:20,307 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=112127.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 17:58:30,695 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.7996, 3.1305, 5.2590, 2.5472, 2.7493, 4.2930, 3.2461, 4.2269], device='cuda:0'), covar=tensor([0.0372, 0.1306, 0.0144, 0.1375, 0.2067, 0.0943, 0.1465, 0.0859], device='cuda:0'), in_proj_covar=tensor([0.0213, 0.0212, 0.0208, 0.0168, 0.0220, 0.0253, 0.0209, 0.0225], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:0') 2023-05-15 17:58:33,052 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=112147.0, num_to_drop=1, layers_to_drop={3} 2023-05-15 17:58:35,566 INFO [finetune.py:992] (0/2) Epoch 1, batch 12150, loss[loss=0.3284, simple_loss=0.3694, pruned_loss=0.1437, over 6485.00 frames. ], tot_loss[loss=0.2416, simple_loss=0.3166, pruned_loss=0.08326, over 1692192.00 frames. ], batch size: 98, lr: 4.98e-03, grad_scale: 8.0 2023-05-15 17:58:38,394 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.7781, 3.1457, 2.4287, 2.1939, 2.8414, 2.2456, 3.0422, 2.5615], device='cuda:0'), covar=tensor([0.0540, 0.0697, 0.0892, 0.1476, 0.0244, 0.1184, 0.0425, 0.0772], device='cuda:0'), in_proj_covar=tensor([0.0180, 0.0227, 0.0168, 0.0190, 0.0127, 0.0174, 0.0179, 0.0168], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-15 17:58:49,591 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=112172.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 17:59:08,146 INFO [finetune.py:992] (0/2) Epoch 1, batch 12200, loss[loss=0.2693, simple_loss=0.3354, pruned_loss=0.1015, over 7137.00 frames. ], tot_loss[loss=0.2429, simple_loss=0.3174, pruned_loss=0.08417, over 1669826.59 frames. ], batch size: 99, lr: 4.98e-03, grad_scale: 8.0 2023-05-15 17:59:10,534 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.169e+02 3.288e+02 3.899e+02 4.371e+02 8.916e+02, threshold=7.798e+02, percent-clipped=1.0 2023-05-15 17:59:13,747 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.7932, 3.7729, 3.7270, 3.8337, 3.6500, 3.6761, 3.6545, 3.7517], device='cuda:0'), covar=tensor([0.0801, 0.0664, 0.1285, 0.0687, 0.1670, 0.1240, 0.0544, 0.1054], device='cuda:0'), in_proj_covar=tensor([0.0420, 0.0536, 0.0463, 0.0507, 0.0651, 0.0604, 0.0444, 0.0407], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0004, 0.0004, 0.0005, 0.0004, 0.0003, 0.0003], device='cuda:0') 2023-05-15 17:59:18,181 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.9772, 2.1863, 2.5655, 2.9477, 2.2504, 3.1211, 2.9908, 3.1042], device='cuda:0'), covar=tensor([0.0157, 0.1022, 0.0403, 0.0195, 0.1049, 0.0297, 0.0279, 0.0136], device='cuda:0'), in_proj_covar=tensor([0.0113, 0.0196, 0.0177, 0.0106, 0.0182, 0.0164, 0.0156, 0.0111], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-15 17:59:29,938 INFO [checkpoint.py:75] (0/2) Saving checkpoint to pruned_transducer_stateless7/exp_giga_finetune/epoch-1.pt 2023-05-15 17:59:54,045 INFO [finetune.py:992] (0/2) Epoch 2, batch 0, loss[loss=0.2181, simple_loss=0.3096, pruned_loss=0.06334, over 12092.00 frames. ], tot_loss[loss=0.2181, simple_loss=0.3096, pruned_loss=0.06334, over 12092.00 frames. ], batch size: 45, lr: 4.98e-03, grad_scale: 8.0 2023-05-15 17:59:54,045 INFO [finetune.py:1017] (0/2) Computing validation loss 2023-05-15 18:00:11,581 INFO [finetune.py:1026] (0/2) Epoch 2, validation: loss=0.2987, simple_loss=0.3712, pruned_loss=0.1132, over 1020973.00 frames. 2023-05-15 18:00:11,582 INFO [finetune.py:1027] (0/2) Maximum memory allocated so far is 12745MB 2023-05-15 18:00:19,206 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.5862, 2.4930, 3.3842, 4.4464, 2.4186, 4.6911, 4.6268, 4.6178], device='cuda:0'), covar=tensor([0.0106, 0.1249, 0.0356, 0.0143, 0.1375, 0.0095, 0.0105, 0.0106], device='cuda:0'), in_proj_covar=tensor([0.0112, 0.0196, 0.0176, 0.0105, 0.0181, 0.0163, 0.0155, 0.0111], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-15 18:00:46,504 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=4.76 vs. limit=5.0 2023-05-15 18:00:47,302 INFO [finetune.py:992] (0/2) Epoch 2, batch 50, loss[loss=0.2078, simple_loss=0.2889, pruned_loss=0.06338, over 12296.00 frames. ], tot_loss[loss=0.1909, simple_loss=0.2775, pruned_loss=0.05214, over 533045.81 frames. ], batch size: 33, lr: 4.98e-03, grad_scale: 8.0 2023-05-15 18:01:01,429 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.225e+02 3.270e+02 3.935e+02 4.941e+02 1.663e+03, threshold=7.870e+02, percent-clipped=3.0 2023-05-15 18:01:04,521 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.7169, 2.7547, 4.5950, 4.8332, 2.9612, 2.6317, 2.8612, 2.0533], device='cuda:0'), covar=tensor([0.1431, 0.3104, 0.0413, 0.0332, 0.1114, 0.2059, 0.2641, 0.4048], device='cuda:0'), in_proj_covar=tensor([0.0268, 0.0355, 0.0251, 0.0273, 0.0239, 0.0268, 0.0342, 0.0344], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-15 18:01:19,127 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.5569, 2.4362, 3.1123, 4.3729, 2.3267, 4.5352, 4.4795, 4.6018], device='cuda:0'), covar=tensor([0.0086, 0.1419, 0.0535, 0.0133, 0.1432, 0.0163, 0.0135, 0.0066], device='cuda:0'), in_proj_covar=tensor([0.0114, 0.0199, 0.0180, 0.0107, 0.0185, 0.0167, 0.0159, 0.0113], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-15 18:01:22,474 INFO [finetune.py:992] (0/2) Epoch 2, batch 100, loss[loss=0.1832, simple_loss=0.2682, pruned_loss=0.04909, over 12088.00 frames. ], tot_loss[loss=0.1871, simple_loss=0.2748, pruned_loss=0.0497, over 944281.14 frames. ], batch size: 32, lr: 4.98e-03, grad_scale: 8.0 2023-05-15 18:01:58,370 INFO [finetune.py:992] (0/2) Epoch 2, batch 150, loss[loss=0.1681, simple_loss=0.2505, pruned_loss=0.04281, over 12250.00 frames. ], tot_loss[loss=0.1858, simple_loss=0.2734, pruned_loss=0.0491, over 1264987.52 frames. ], batch size: 32, lr: 4.98e-03, grad_scale: 8.0 2023-05-15 18:02:01,499 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=112389.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 18:02:08,701 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-05-15 18:02:12,810 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.396e+02 2.983e+02 3.524e+02 4.187e+02 7.539e+02, threshold=7.047e+02, percent-clipped=0.0 2023-05-15 18:02:25,636 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=112422.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 18:02:34,901 INFO [finetune.py:992] (0/2) Epoch 2, batch 200, loss[loss=0.1768, simple_loss=0.2697, pruned_loss=0.042, over 12164.00 frames. ], tot_loss[loss=0.1834, simple_loss=0.2707, pruned_loss=0.04802, over 1510474.22 frames. ], batch size: 34, lr: 4.98e-03, grad_scale: 8.0 2023-05-15 18:02:43,302 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=112447.0, num_to_drop=1, layers_to_drop={0} 2023-05-15 18:02:45,599 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=112450.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 18:02:47,638 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.0075, 5.8820, 5.4764, 5.3945, 5.9462, 5.2680, 5.4828, 5.5054], device='cuda:0'), covar=tensor([0.1513, 0.0828, 0.0818, 0.2081, 0.0983, 0.2158, 0.1683, 0.0847], device='cuda:0'), in_proj_covar=tensor([0.0314, 0.0430, 0.0345, 0.0392, 0.0416, 0.0393, 0.0351, 0.0340], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-15 18:03:10,362 INFO [finetune.py:992] (0/2) Epoch 2, batch 250, loss[loss=0.1458, simple_loss=0.2321, pruned_loss=0.02979, over 12360.00 frames. ], tot_loss[loss=0.1831, simple_loss=0.2708, pruned_loss=0.04772, over 1706552.19 frames. ], batch size: 31, lr: 4.98e-03, grad_scale: 8.0 2023-05-15 18:03:18,524 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=112495.0, num_to_drop=1, layers_to_drop={1} 2023-05-15 18:03:20,932 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-05-15 18:03:25,434 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.114e+02 2.915e+02 3.506e+02 4.058e+02 6.967e+02, threshold=7.011e+02, percent-clipped=0.0 2023-05-15 18:03:46,660 INFO [finetune.py:992] (0/2) Epoch 2, batch 300, loss[loss=0.1761, simple_loss=0.2654, pruned_loss=0.04342, over 12132.00 frames. ], tot_loss[loss=0.1817, simple_loss=0.2695, pruned_loss=0.0469, over 1857017.87 frames. ], batch size: 38, lr: 4.98e-03, grad_scale: 8.0 2023-05-15 18:04:06,067 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=112562.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 18:04:17,251 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=112577.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 18:04:22,779 INFO [finetune.py:992] (0/2) Epoch 2, batch 350, loss[loss=0.1869, simple_loss=0.2788, pruned_loss=0.0475, over 11851.00 frames. ], tot_loss[loss=0.1796, simple_loss=0.2671, pruned_loss=0.04603, over 1969189.06 frames. ], batch size: 44, lr: 4.98e-03, grad_scale: 8.0 2023-05-15 18:04:37,325 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.052e+02 2.824e+02 3.340e+02 4.009e+02 1.325e+03, threshold=6.681e+02, percent-clipped=2.0 2023-05-15 18:04:45,970 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=112617.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 18:04:49,703 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=2.32 vs. limit=5.0 2023-05-15 18:04:50,384 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=112623.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 18:04:56,029 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.9164, 4.9410, 4.6762, 4.8195, 4.3742, 4.9022, 4.8806, 5.0937], device='cuda:0'), covar=tensor([0.0205, 0.0139, 0.0220, 0.0276, 0.0805, 0.0260, 0.0147, 0.0198], device='cuda:0'), in_proj_covar=tensor([0.0165, 0.0169, 0.0170, 0.0209, 0.0215, 0.0183, 0.0158, 0.0204], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0005, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:0') 2023-05-15 18:04:58,650 INFO [finetune.py:992] (0/2) Epoch 2, batch 400, loss[loss=0.1606, simple_loss=0.2337, pruned_loss=0.04376, over 12004.00 frames. ], tot_loss[loss=0.1785, simple_loss=0.2661, pruned_loss=0.04543, over 2068382.13 frames. ], batch size: 28, lr: 4.98e-03, grad_scale: 8.0 2023-05-15 18:05:01,044 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=112638.0, num_to_drop=1, layers_to_drop={2} 2023-05-15 18:05:06,560 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.7204, 5.4452, 5.0355, 4.9240, 5.4849, 4.7890, 5.0709, 4.9512], device='cuda:0'), covar=tensor([0.1497, 0.0931, 0.1107, 0.2151, 0.1107, 0.2382, 0.1815, 0.1207], device='cuda:0'), in_proj_covar=tensor([0.0317, 0.0433, 0.0347, 0.0397, 0.0421, 0.0398, 0.0355, 0.0346], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-15 18:05:29,932 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=112678.0, num_to_drop=1, layers_to_drop={1} 2023-05-15 18:05:34,540 INFO [finetune.py:992] (0/2) Epoch 2, batch 450, loss[loss=0.1858, simple_loss=0.2653, pruned_loss=0.0532, over 12354.00 frames. ], tot_loss[loss=0.1794, simple_loss=0.2667, pruned_loss=0.04606, over 2136983.09 frames. ], batch size: 30, lr: 4.98e-03, grad_scale: 8.0 2023-05-15 18:05:48,795 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.011e+02 2.826e+02 3.576e+02 4.360e+02 8.764e+02, threshold=7.152e+02, percent-clipped=3.0 2023-05-15 18:05:49,772 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.3391, 3.1577, 3.1115, 2.9751, 2.6969, 2.4306, 3.2810, 2.0908], device='cuda:0'), covar=tensor([0.0341, 0.0143, 0.0152, 0.0161, 0.0336, 0.0362, 0.0110, 0.0454], device='cuda:0'), in_proj_covar=tensor([0.0177, 0.0152, 0.0141, 0.0170, 0.0192, 0.0185, 0.0147, 0.0191], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], device='cuda:0') 2023-05-15 18:05:53,958 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.8516, 2.8159, 5.3204, 2.3508, 2.4519, 4.2357, 3.2475, 4.0704], device='cuda:0'), covar=tensor([0.0368, 0.1469, 0.0204, 0.1417, 0.2364, 0.1004, 0.1381, 0.1067], device='cuda:0'), in_proj_covar=tensor([0.0225, 0.0223, 0.0222, 0.0178, 0.0233, 0.0269, 0.0220, 0.0240], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:0') 2023-05-15 18:05:54,798 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.70 vs. limit=2.0 2023-05-15 18:06:00,942 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=112722.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 18:06:10,648 INFO [finetune.py:992] (0/2) Epoch 2, batch 500, loss[loss=0.1491, simple_loss=0.2368, pruned_loss=0.03067, over 12038.00 frames. ], tot_loss[loss=0.1789, simple_loss=0.266, pruned_loss=0.04593, over 2190847.43 frames. ], batch size: 31, lr: 4.98e-03, grad_scale: 8.0 2023-05-15 18:06:17,967 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=112745.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 18:06:35,543 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=112770.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 18:06:46,037 INFO [finetune.py:992] (0/2) Epoch 2, batch 550, loss[loss=0.1842, simple_loss=0.2774, pruned_loss=0.04546, over 12156.00 frames. ], tot_loss[loss=0.1778, simple_loss=0.2655, pruned_loss=0.04508, over 2243120.06 frames. ], batch size: 39, lr: 4.98e-03, grad_scale: 8.0 2023-05-15 18:07:01,099 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.806e+02 2.848e+02 3.230e+02 3.802e+02 1.094e+03, threshold=6.461e+02, percent-clipped=1.0 2023-05-15 18:07:21,259 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.2625, 4.0873, 3.9583, 4.4086, 2.8992, 4.1422, 2.5084, 4.1274], device='cuda:0'), covar=tensor([0.1660, 0.0663, 0.1012, 0.0668, 0.1162, 0.0510, 0.1932, 0.1368], device='cuda:0'), in_proj_covar=tensor([0.0224, 0.0262, 0.0295, 0.0350, 0.0240, 0.0237, 0.0255, 0.0386], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-15 18:07:22,410 INFO [finetune.py:992] (0/2) Epoch 2, batch 600, loss[loss=0.1951, simple_loss=0.2776, pruned_loss=0.05635, over 12121.00 frames. ], tot_loss[loss=0.1782, simple_loss=0.2663, pruned_loss=0.04507, over 2279650.46 frames. ], batch size: 38, lr: 4.98e-03, grad_scale: 8.0 2023-05-15 18:07:25,953 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.9646, 4.8992, 4.7847, 4.9118, 4.0836, 5.0527, 4.9636, 5.1058], device='cuda:0'), covar=tensor([0.0299, 0.0158, 0.0217, 0.0286, 0.1230, 0.0232, 0.0166, 0.0217], device='cuda:0'), in_proj_covar=tensor([0.0167, 0.0173, 0.0172, 0.0213, 0.0219, 0.0186, 0.0160, 0.0206], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0005, 0.0005, 0.0004, 0.0004, 0.0005], device='cuda:0') 2023-05-15 18:07:58,206 INFO [finetune.py:992] (0/2) Epoch 2, batch 650, loss[loss=0.1711, simple_loss=0.2705, pruned_loss=0.0359, over 12344.00 frames. ], tot_loss[loss=0.1784, simple_loss=0.2659, pruned_loss=0.04545, over 2292359.42 frames. ], batch size: 35, lr: 4.98e-03, grad_scale: 8.0 2023-05-15 18:08:12,430 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.261e+02 2.862e+02 3.449e+02 4.206e+02 6.989e+02, threshold=6.897e+02, percent-clipped=2.0 2023-05-15 18:08:16,494 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.91 vs. limit=2.0 2023-05-15 18:08:21,925 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=112918.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 18:08:24,112 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=112921.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 18:08:32,638 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=112933.0, num_to_drop=1, layers_to_drop={3} 2023-05-15 18:08:33,934 INFO [finetune.py:992] (0/2) Epoch 2, batch 700, loss[loss=0.1766, simple_loss=0.2665, pruned_loss=0.04334, over 12350.00 frames. ], tot_loss[loss=0.1786, simple_loss=0.2662, pruned_loss=0.04544, over 2305262.63 frames. ], batch size: 35, lr: 4.98e-03, grad_scale: 8.0 2023-05-15 18:09:01,386 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=112973.0, num_to_drop=1, layers_to_drop={3} 2023-05-15 18:09:06,571 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-05-15 18:09:07,933 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=112982.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 18:09:09,722 INFO [finetune.py:992] (0/2) Epoch 2, batch 750, loss[loss=0.1812, simple_loss=0.273, pruned_loss=0.04468, over 12094.00 frames. ], tot_loss[loss=0.1783, simple_loss=0.266, pruned_loss=0.04529, over 2320474.26 frames. ], batch size: 33, lr: 4.98e-03, grad_scale: 8.0 2023-05-15 18:09:09,925 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.0313, 2.2918, 3.0088, 3.9414, 2.2299, 4.0356, 3.9919, 4.1567], device='cuda:0'), covar=tensor([0.0134, 0.1200, 0.0463, 0.0107, 0.1276, 0.0227, 0.0180, 0.0070], device='cuda:0'), in_proj_covar=tensor([0.0118, 0.0202, 0.0184, 0.0109, 0.0188, 0.0173, 0.0165, 0.0116], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-15 18:09:14,917 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.5889, 5.0608, 5.4576, 4.7835, 5.1414, 4.8868, 5.5846, 5.1964], device='cuda:0'), covar=tensor([0.0210, 0.0298, 0.0255, 0.0269, 0.0298, 0.0274, 0.0163, 0.0179], device='cuda:0'), in_proj_covar=tensor([0.0213, 0.0223, 0.0239, 0.0219, 0.0219, 0.0218, 0.0200, 0.0186], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-15 18:09:24,413 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.175e+02 2.857e+02 3.349e+02 4.061e+02 6.781e+02, threshold=6.697e+02, percent-clipped=0.0 2023-05-15 18:09:40,162 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=3.18 vs. limit=5.0 2023-05-15 18:09:46,146 INFO [finetune.py:992] (0/2) Epoch 2, batch 800, loss[loss=0.1669, simple_loss=0.249, pruned_loss=0.04243, over 12281.00 frames. ], tot_loss[loss=0.1775, simple_loss=0.2652, pruned_loss=0.04487, over 2332792.34 frames. ], batch size: 28, lr: 4.98e-03, grad_scale: 8.0 2023-05-15 18:09:53,364 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=113045.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 18:10:21,389 INFO [finetune.py:992] (0/2) Epoch 2, batch 850, loss[loss=0.1547, simple_loss=0.2494, pruned_loss=0.03002, over 12415.00 frames. ], tot_loss[loss=0.1782, simple_loss=0.2659, pruned_loss=0.04531, over 2341767.53 frames. ], batch size: 32, lr: 4.98e-03, grad_scale: 8.0 2023-05-15 18:10:27,744 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=113093.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 18:10:36,173 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.012e+02 3.121e+02 3.766e+02 4.532e+02 1.104e+03, threshold=7.531e+02, percent-clipped=12.0 2023-05-15 18:10:57,556 INFO [finetune.py:992] (0/2) Epoch 2, batch 900, loss[loss=0.1635, simple_loss=0.2548, pruned_loss=0.03613, over 12105.00 frames. ], tot_loss[loss=0.1771, simple_loss=0.2647, pruned_loss=0.04473, over 2358256.23 frames. ], batch size: 33, lr: 4.98e-03, grad_scale: 8.0 2023-05-15 18:11:00,091 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2023-05-15 18:11:33,565 INFO [finetune.py:992] (0/2) Epoch 2, batch 950, loss[loss=0.1964, simple_loss=0.2989, pruned_loss=0.04688, over 12252.00 frames. ], tot_loss[loss=0.1772, simple_loss=0.2653, pruned_loss=0.04457, over 2358058.73 frames. ], batch size: 37, lr: 4.98e-03, grad_scale: 8.0 2023-05-15 18:11:43,894 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.1421, 2.5430, 3.6663, 3.1551, 3.5148, 3.2206, 2.4516, 3.6336], device='cuda:0'), covar=tensor([0.0097, 0.0276, 0.0121, 0.0200, 0.0128, 0.0149, 0.0283, 0.0095], device='cuda:0'), in_proj_covar=tensor([0.0151, 0.0190, 0.0163, 0.0170, 0.0184, 0.0146, 0.0179, 0.0156], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-15 18:11:48,148 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.177e+02 2.906e+02 3.318e+02 3.826e+02 7.167e+02, threshold=6.636e+02, percent-clipped=0.0 2023-05-15 18:11:57,290 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=113218.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 18:12:06,010 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.7035, 2.6536, 3.3396, 4.5422, 2.7185, 4.6150, 4.6053, 4.8219], device='cuda:0'), covar=tensor([0.0090, 0.1150, 0.0426, 0.0096, 0.1169, 0.0158, 0.0107, 0.0052], device='cuda:0'), in_proj_covar=tensor([0.0117, 0.0201, 0.0184, 0.0109, 0.0188, 0.0172, 0.0165, 0.0115], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-15 18:12:08,082 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=113233.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 18:12:08,891 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=113234.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 18:12:09,392 INFO [finetune.py:992] (0/2) Epoch 2, batch 1000, loss[loss=0.2134, simple_loss=0.2966, pruned_loss=0.06512, over 12076.00 frames. ], tot_loss[loss=0.1768, simple_loss=0.2646, pruned_loss=0.04451, over 2361606.28 frames. ], batch size: 42, lr: 4.98e-03, grad_scale: 8.0 2023-05-15 18:12:32,400 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=113266.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 18:12:37,488 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=113273.0, num_to_drop=1, layers_to_drop={2} 2023-05-15 18:12:40,284 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=113277.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 18:12:43,104 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=113281.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 18:12:45,832 INFO [finetune.py:992] (0/2) Epoch 2, batch 1050, loss[loss=0.1806, simple_loss=0.2745, pruned_loss=0.04338, over 11761.00 frames. ], tot_loss[loss=0.1758, simple_loss=0.2635, pruned_loss=0.04409, over 2363380.67 frames. ], batch size: 44, lr: 4.98e-03, grad_scale: 8.0 2023-05-15 18:12:53,176 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=113295.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 18:13:00,151 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.997e+02 2.902e+02 3.523e+02 4.453e+02 1.735e+03, threshold=7.046e+02, percent-clipped=8.0 2023-05-15 18:13:11,490 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=113321.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 18:13:22,315 INFO [finetune.py:992] (0/2) Epoch 2, batch 1100, loss[loss=0.1963, simple_loss=0.2885, pruned_loss=0.05212, over 11326.00 frames. ], tot_loss[loss=0.1763, simple_loss=0.2638, pruned_loss=0.04437, over 2369203.16 frames. ], batch size: 55, lr: 4.98e-03, grad_scale: 8.0 2023-05-15 18:13:57,242 INFO [finetune.py:992] (0/2) Epoch 2, batch 1150, loss[loss=0.2034, simple_loss=0.2884, pruned_loss=0.05925, over 12113.00 frames. ], tot_loss[loss=0.1767, simple_loss=0.2642, pruned_loss=0.04461, over 2368679.03 frames. ], batch size: 38, lr: 4.98e-03, grad_scale: 8.0 2023-05-15 18:14:11,172 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.4409, 4.6110, 4.1826, 4.9860, 4.7088, 2.9010, 4.2891, 3.0180], device='cuda:0'), covar=tensor([0.0663, 0.0765, 0.1260, 0.0387, 0.0836, 0.1541, 0.1069, 0.3246], device='cuda:0'), in_proj_covar=tensor([0.0291, 0.0364, 0.0343, 0.0250, 0.0352, 0.0261, 0.0326, 0.0357], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-15 18:14:12,257 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.916e+02 2.863e+02 3.318e+02 3.829e+02 6.998e+02, threshold=6.637e+02, percent-clipped=0.0 2023-05-15 18:14:20,815 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.4231, 2.1389, 3.7221, 4.5305, 4.0038, 4.3731, 4.0303, 2.9585], device='cuda:0'), covar=tensor([0.0026, 0.0427, 0.0097, 0.0024, 0.0078, 0.0058, 0.0071, 0.0337], device='cuda:0'), in_proj_covar=tensor([0.0077, 0.0120, 0.0099, 0.0072, 0.0095, 0.0107, 0.0083, 0.0133], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-15 18:14:33,300 INFO [finetune.py:992] (0/2) Epoch 2, batch 1200, loss[loss=0.1885, simple_loss=0.276, pruned_loss=0.05046, over 12093.00 frames. ], tot_loss[loss=0.1767, simple_loss=0.2641, pruned_loss=0.04464, over 2376594.95 frames. ], batch size: 39, lr: 4.98e-03, grad_scale: 8.0 2023-05-15 18:15:09,606 INFO [finetune.py:992] (0/2) Epoch 2, batch 1250, loss[loss=0.1572, simple_loss=0.2428, pruned_loss=0.03577, over 12352.00 frames. ], tot_loss[loss=0.1764, simple_loss=0.2638, pruned_loss=0.04447, over 2378775.66 frames. ], batch size: 31, lr: 4.98e-03, grad_scale: 8.0 2023-05-15 18:15:23,694 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.287e+02 2.869e+02 3.347e+02 3.921e+02 1.251e+03, threshold=6.694e+02, percent-clipped=4.0 2023-05-15 18:15:45,173 INFO [finetune.py:992] (0/2) Epoch 2, batch 1300, loss[loss=0.1768, simple_loss=0.2719, pruned_loss=0.04085, over 11290.00 frames. ], tot_loss[loss=0.1769, simple_loss=0.2645, pruned_loss=0.04467, over 2382167.21 frames. ], batch size: 55, lr: 4.98e-03, grad_scale: 8.0 2023-05-15 18:15:46,924 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.1626, 3.9912, 4.0434, 4.4716, 2.7984, 4.1037, 2.5463, 4.0627], device='cuda:0'), covar=tensor([0.1748, 0.0761, 0.0879, 0.0658, 0.1262, 0.0546, 0.1945, 0.1412], device='cuda:0'), in_proj_covar=tensor([0.0226, 0.0266, 0.0298, 0.0358, 0.0243, 0.0243, 0.0259, 0.0394], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-15 18:16:15,933 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=113577.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 18:16:17,390 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.0622, 5.0324, 4.8717, 4.9583, 4.5650, 5.0060, 5.0032, 5.3148], device='cuda:0'), covar=tensor([0.0198, 0.0141, 0.0194, 0.0275, 0.0759, 0.0315, 0.0168, 0.0157], device='cuda:0'), in_proj_covar=tensor([0.0175, 0.0178, 0.0178, 0.0220, 0.0225, 0.0193, 0.0165, 0.0212], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0005, 0.0005, 0.0005, 0.0004, 0.0005], device='cuda:0') 2023-05-15 18:16:21,374 INFO [finetune.py:992] (0/2) Epoch 2, batch 1350, loss[loss=0.1517, simple_loss=0.2404, pruned_loss=0.0315, over 12350.00 frames. ], tot_loss[loss=0.1766, simple_loss=0.2642, pruned_loss=0.04451, over 2377625.40 frames. ], batch size: 31, lr: 4.98e-03, grad_scale: 8.0 2023-05-15 18:16:25,052 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=113590.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 18:16:36,576 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.039e+02 2.988e+02 3.437e+02 3.961e+02 7.082e+02, threshold=6.873e+02, percent-clipped=1.0 2023-05-15 18:16:44,440 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=113617.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 18:16:50,801 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=113625.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 18:16:58,004 INFO [finetune.py:992] (0/2) Epoch 2, batch 1400, loss[loss=0.2043, simple_loss=0.2985, pruned_loss=0.05507, over 11153.00 frames. ], tot_loss[loss=0.1763, simple_loss=0.2641, pruned_loss=0.04428, over 2380162.53 frames. ], batch size: 55, lr: 4.98e-03, grad_scale: 8.0 2023-05-15 18:17:05,923 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=113646.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 18:17:06,748 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.8600, 3.4232, 5.1369, 2.6303, 2.7983, 3.8558, 3.2183, 3.8772], device='cuda:0'), covar=tensor([0.0406, 0.1062, 0.0296, 0.1237, 0.1943, 0.1433, 0.1359, 0.1127], device='cuda:0'), in_proj_covar=tensor([0.0231, 0.0229, 0.0231, 0.0182, 0.0238, 0.0278, 0.0225, 0.0248], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:0') 2023-05-15 18:17:28,240 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=113678.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 18:17:33,065 INFO [finetune.py:992] (0/2) Epoch 2, batch 1450, loss[loss=0.1566, simple_loss=0.2467, pruned_loss=0.0333, over 12077.00 frames. ], tot_loss[loss=0.1755, simple_loss=0.2628, pruned_loss=0.04405, over 2388422.18 frames. ], batch size: 32, lr: 4.98e-03, grad_scale: 8.0 2023-05-15 18:17:40,619 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 2023-05-15 18:17:41,292 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2023-05-15 18:17:48,634 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.119e+02 2.823e+02 3.362e+02 3.969e+02 1.640e+03, threshold=6.724e+02, percent-clipped=5.0 2023-05-15 18:17:49,484 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=113707.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 18:18:08,653 INFO [finetune.py:992] (0/2) Epoch 2, batch 1500, loss[loss=0.1677, simple_loss=0.2542, pruned_loss=0.0406, over 12185.00 frames. ], tot_loss[loss=0.1762, simple_loss=0.2639, pruned_loss=0.04421, over 2381445.78 frames. ], batch size: 31, lr: 4.98e-03, grad_scale: 8.0 2023-05-15 18:18:44,172 INFO [finetune.py:992] (0/2) Epoch 2, batch 1550, loss[loss=0.171, simple_loss=0.2644, pruned_loss=0.03884, over 12311.00 frames. ], tot_loss[loss=0.1761, simple_loss=0.2641, pruned_loss=0.04412, over 2386698.39 frames. ], batch size: 34, lr: 4.98e-03, grad_scale: 8.0 2023-05-15 18:18:59,279 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.034e+02 2.900e+02 3.326e+02 4.041e+02 9.467e+02, threshold=6.652e+02, percent-clipped=1.0 2023-05-15 18:19:14,518 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.8395, 4.5348, 4.7020, 4.7929, 4.5652, 4.7901, 4.6441, 2.6109], device='cuda:0'), covar=tensor([0.0121, 0.0054, 0.0070, 0.0057, 0.0050, 0.0069, 0.0067, 0.0672], device='cuda:0'), in_proj_covar=tensor([0.0063, 0.0072, 0.0075, 0.0069, 0.0056, 0.0084, 0.0073, 0.0089], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-15 18:19:20,722 INFO [finetune.py:992] (0/2) Epoch 2, batch 1600, loss[loss=0.1808, simple_loss=0.2712, pruned_loss=0.04516, over 12139.00 frames. ], tot_loss[loss=0.1764, simple_loss=0.2646, pruned_loss=0.04416, over 2390361.35 frames. ], batch size: 38, lr: 4.98e-03, grad_scale: 8.0 2023-05-15 18:19:56,672 INFO [finetune.py:992] (0/2) Epoch 2, batch 1650, loss[loss=0.1539, simple_loss=0.2345, pruned_loss=0.0367, over 12182.00 frames. ], tot_loss[loss=0.1761, simple_loss=0.2642, pruned_loss=0.04402, over 2386595.50 frames. ], batch size: 29, lr: 4.98e-03, grad_scale: 8.0 2023-05-15 18:20:00,333 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=113890.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 18:20:11,662 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.174e+02 2.961e+02 3.386e+02 4.032e+02 9.524e+02, threshold=6.773e+02, percent-clipped=3.0 2023-05-15 18:20:33,101 INFO [finetune.py:992] (0/2) Epoch 2, batch 1700, loss[loss=0.1792, simple_loss=0.2791, pruned_loss=0.03959, over 12274.00 frames. ], tot_loss[loss=0.1765, simple_loss=0.2642, pruned_loss=0.0444, over 2385636.16 frames. ], batch size: 37, lr: 4.98e-03, grad_scale: 8.0 2023-05-15 18:20:33,290 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.1801, 2.1754, 3.6087, 4.2556, 3.7024, 4.0820, 3.8284, 2.9354], device='cuda:0'), covar=tensor([0.0040, 0.0428, 0.0117, 0.0034, 0.0113, 0.0079, 0.0083, 0.0321], device='cuda:0'), in_proj_covar=tensor([0.0077, 0.0118, 0.0098, 0.0072, 0.0095, 0.0107, 0.0083, 0.0131], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-15 18:20:35,347 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=113938.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 18:20:37,490 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.5559, 2.8072, 3.6853, 4.6635, 3.9296, 4.6695, 4.1231, 3.5313], device='cuda:0'), covar=tensor([0.0034, 0.0325, 0.0130, 0.0034, 0.0133, 0.0063, 0.0089, 0.0258], device='cuda:0'), in_proj_covar=tensor([0.0077, 0.0118, 0.0098, 0.0072, 0.0095, 0.0107, 0.0083, 0.0131], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-15 18:20:59,872 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=113973.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 18:21:08,329 INFO [finetune.py:992] (0/2) Epoch 2, batch 1750, loss[loss=0.1806, simple_loss=0.2833, pruned_loss=0.03894, over 12148.00 frames. ], tot_loss[loss=0.1767, simple_loss=0.2646, pruned_loss=0.04434, over 2381536.14 frames. ], batch size: 34, lr: 4.98e-03, grad_scale: 8.0 2023-05-15 18:21:19,890 INFO [checkpoint.py:75] (0/2) Saving checkpoint to pruned_transducer_stateless7/exp_giga_finetune/checkpoint-14000.pt 2023-05-15 18:21:24,116 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=114002.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 18:21:26,783 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.193e+02 2.960e+02 3.494e+02 4.137e+02 7.553e+02, threshold=6.989e+02, percent-clipped=3.0 2023-05-15 18:21:47,098 INFO [finetune.py:992] (0/2) Epoch 2, batch 1800, loss[loss=0.1977, simple_loss=0.2838, pruned_loss=0.05581, over 12350.00 frames. ], tot_loss[loss=0.1757, simple_loss=0.2638, pruned_loss=0.04386, over 2390172.80 frames. ], batch size: 36, lr: 4.98e-03, grad_scale: 8.0 2023-05-15 18:22:05,815 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 2023-05-15 18:22:22,824 INFO [finetune.py:992] (0/2) Epoch 2, batch 1850, loss[loss=0.1629, simple_loss=0.2435, pruned_loss=0.04111, over 11779.00 frames. ], tot_loss[loss=0.1753, simple_loss=0.263, pruned_loss=0.04383, over 2383428.98 frames. ], batch size: 26, lr: 4.98e-03, grad_scale: 8.0 2023-05-15 18:22:30,599 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.33 vs. limit=2.0 2023-05-15 18:22:31,715 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.6417, 2.7635, 4.7063, 4.8782, 3.0156, 2.6480, 2.8306, 2.1576], device='cuda:0'), covar=tensor([0.1314, 0.2798, 0.0363, 0.0315, 0.1004, 0.1863, 0.2385, 0.3411], device='cuda:0'), in_proj_covar=tensor([0.0269, 0.0364, 0.0258, 0.0280, 0.0243, 0.0269, 0.0345, 0.0345], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-15 18:22:37,774 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.825e+02 2.921e+02 3.335e+02 4.221e+02 7.050e+02, threshold=6.670e+02, percent-clipped=1.0 2023-05-15 18:22:52,671 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=114127.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 18:22:59,062 INFO [finetune.py:992] (0/2) Epoch 2, batch 1900, loss[loss=0.2609, simple_loss=0.3331, pruned_loss=0.09433, over 8280.00 frames. ], tot_loss[loss=0.1759, simple_loss=0.2634, pruned_loss=0.04424, over 2376133.21 frames. ], batch size: 97, lr: 4.98e-03, grad_scale: 8.0 2023-05-15 18:23:08,435 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.7510, 3.4399, 5.1285, 2.7003, 2.7616, 3.9037, 3.2566, 3.8501], device='cuda:0'), covar=tensor([0.0455, 0.1032, 0.0285, 0.1145, 0.2009, 0.1232, 0.1396, 0.1006], device='cuda:0'), in_proj_covar=tensor([0.0235, 0.0232, 0.0235, 0.0183, 0.0241, 0.0283, 0.0229, 0.0253], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-15 18:23:29,057 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.3301, 2.3763, 3.1959, 4.1464, 2.3465, 4.3245, 4.1939, 4.4378], device='cuda:0'), covar=tensor([0.0117, 0.1108, 0.0402, 0.0132, 0.1183, 0.0174, 0.0155, 0.0076], device='cuda:0'), in_proj_covar=tensor([0.0119, 0.0202, 0.0183, 0.0110, 0.0188, 0.0173, 0.0168, 0.0116], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-15 18:23:29,911 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.0131, 3.5398, 5.3929, 2.9729, 2.9383, 4.0266, 3.3621, 4.0706], device='cuda:0'), covar=tensor([0.0351, 0.0985, 0.0200, 0.1044, 0.1871, 0.1253, 0.1306, 0.0911], device='cuda:0'), in_proj_covar=tensor([0.0235, 0.0231, 0.0234, 0.0183, 0.0240, 0.0282, 0.0228, 0.0252], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-15 18:23:34,313 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=2.00 vs. limit=2.0 2023-05-15 18:23:34,600 INFO [finetune.py:992] (0/2) Epoch 2, batch 1950, loss[loss=0.1506, simple_loss=0.2326, pruned_loss=0.03437, over 12179.00 frames. ], tot_loss[loss=0.1755, simple_loss=0.263, pruned_loss=0.04399, over 2381726.50 frames. ], batch size: 29, lr: 4.98e-03, grad_scale: 8.0 2023-05-15 18:23:36,804 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=114188.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 18:23:49,661 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.675e+02 2.996e+02 3.482e+02 4.338e+02 7.908e+02, threshold=6.963e+02, percent-clipped=3.0 2023-05-15 18:24:01,101 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.9278, 3.5867, 5.1804, 2.7699, 2.9153, 3.8709, 3.5156, 3.7895], device='cuda:0'), covar=tensor([0.0348, 0.0967, 0.0270, 0.1185, 0.1930, 0.1368, 0.1282, 0.1206], device='cuda:0'), in_proj_covar=tensor([0.0234, 0.0231, 0.0233, 0.0183, 0.0240, 0.0281, 0.0227, 0.0251], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-15 18:24:03,290 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.6883, 3.2858, 4.9972, 2.5555, 2.7244, 3.7026, 3.2771, 3.6818], device='cuda:0'), covar=tensor([0.0391, 0.1076, 0.0337, 0.1206, 0.2052, 0.1409, 0.1331, 0.1108], device='cuda:0'), in_proj_covar=tensor([0.0234, 0.0231, 0.0233, 0.0183, 0.0240, 0.0281, 0.0227, 0.0251], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-15 18:24:10,848 INFO [finetune.py:992] (0/2) Epoch 2, batch 2000, loss[loss=0.1751, simple_loss=0.2611, pruned_loss=0.0446, over 12189.00 frames. ], tot_loss[loss=0.1751, simple_loss=0.2628, pruned_loss=0.04367, over 2382641.63 frames. ], batch size: 31, lr: 4.97e-03, grad_scale: 8.0 2023-05-15 18:24:38,097 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=114273.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 18:24:46,217 INFO [finetune.py:992] (0/2) Epoch 2, batch 2050, loss[loss=0.1932, simple_loss=0.2749, pruned_loss=0.05574, over 12141.00 frames. ], tot_loss[loss=0.1746, simple_loss=0.2625, pruned_loss=0.04341, over 2389387.02 frames. ], batch size: 30, lr: 4.97e-03, grad_scale: 8.0 2023-05-15 18:24:59,100 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=114302.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 18:25:01,714 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.869e+02 2.702e+02 3.223e+02 3.685e+02 5.606e+02, threshold=6.446e+02, percent-clipped=0.0 2023-05-15 18:25:12,488 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=114321.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 18:25:22,478 INFO [finetune.py:992] (0/2) Epoch 2, batch 2100, loss[loss=0.1691, simple_loss=0.2478, pruned_loss=0.04519, over 11815.00 frames. ], tot_loss[loss=0.1745, simple_loss=0.262, pruned_loss=0.04346, over 2390394.99 frames. ], batch size: 26, lr: 4.97e-03, grad_scale: 8.0 2023-05-15 18:25:32,958 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=114350.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 18:25:58,390 INFO [finetune.py:992] (0/2) Epoch 2, batch 2150, loss[loss=0.1387, simple_loss=0.2228, pruned_loss=0.02734, over 12279.00 frames. ], tot_loss[loss=0.1758, simple_loss=0.2632, pruned_loss=0.04422, over 2370451.14 frames. ], batch size: 28, lr: 4.97e-03, grad_scale: 8.0 2023-05-15 18:26:09,193 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.3546, 6.0517, 5.5278, 5.5946, 6.1276, 5.4596, 5.6952, 5.6927], device='cuda:0'), covar=tensor([0.1329, 0.0912, 0.1047, 0.1962, 0.0878, 0.2175, 0.1612, 0.1103], device='cuda:0'), in_proj_covar=tensor([0.0328, 0.0451, 0.0360, 0.0407, 0.0437, 0.0408, 0.0369, 0.0363], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-15 18:26:13,342 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.133e+02 3.063e+02 3.595e+02 4.621e+02 1.301e+03, threshold=7.190e+02, percent-clipped=5.0 2023-05-15 18:26:34,482 INFO [finetune.py:992] (0/2) Epoch 2, batch 2200, loss[loss=0.1516, simple_loss=0.2318, pruned_loss=0.03569, over 12281.00 frames. ], tot_loss[loss=0.1768, simple_loss=0.264, pruned_loss=0.0448, over 2362087.45 frames. ], batch size: 28, lr: 4.97e-03, grad_scale: 8.0 2023-05-15 18:26:41,301 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=4.49 vs. limit=5.0 2023-05-15 18:27:08,554 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=114483.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 18:27:09,928 INFO [finetune.py:992] (0/2) Epoch 2, batch 2250, loss[loss=0.1522, simple_loss=0.2327, pruned_loss=0.03585, over 12001.00 frames. ], tot_loss[loss=0.1755, simple_loss=0.263, pruned_loss=0.04402, over 2375158.10 frames. ], batch size: 28, lr: 4.97e-03, grad_scale: 8.0 2023-05-15 18:27:25,722 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.121e+02 2.927e+02 3.341e+02 4.008e+02 6.848e+02, threshold=6.682e+02, percent-clipped=0.0 2023-05-15 18:27:44,304 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=2.31 vs. limit=5.0 2023-05-15 18:27:46,131 INFO [finetune.py:992] (0/2) Epoch 2, batch 2300, loss[loss=0.1549, simple_loss=0.2466, pruned_loss=0.03159, over 12284.00 frames. ], tot_loss[loss=0.1754, simple_loss=0.2627, pruned_loss=0.04405, over 2374453.27 frames. ], batch size: 33, lr: 4.97e-03, grad_scale: 8.0 2023-05-15 18:28:21,600 INFO [finetune.py:992] (0/2) Epoch 2, batch 2350, loss[loss=0.2047, simple_loss=0.2981, pruned_loss=0.05567, over 12043.00 frames. ], tot_loss[loss=0.1755, simple_loss=0.2631, pruned_loss=0.04394, over 2373355.47 frames. ], batch size: 40, lr: 4.97e-03, grad_scale: 8.0 2023-05-15 18:28:36,737 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.032e+02 2.856e+02 3.351e+02 4.065e+02 8.891e+02, threshold=6.701e+02, percent-clipped=2.0 2023-05-15 18:28:38,079 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=4.47 vs. limit=5.0 2023-05-15 18:28:57,359 INFO [finetune.py:992] (0/2) Epoch 2, batch 2400, loss[loss=0.1786, simple_loss=0.2665, pruned_loss=0.04538, over 12174.00 frames. ], tot_loss[loss=0.1749, simple_loss=0.2625, pruned_loss=0.04361, over 2379722.34 frames. ], batch size: 36, lr: 4.97e-03, grad_scale: 8.0 2023-05-15 18:29:27,013 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.1969, 4.8357, 5.0576, 4.4716, 4.9140, 4.5094, 5.0583, 4.9242], device='cuda:0'), covar=tensor([0.0346, 0.0412, 0.0517, 0.0340, 0.0323, 0.0342, 0.0408, 0.0396], device='cuda:0'), in_proj_covar=tensor([0.0227, 0.0236, 0.0255, 0.0230, 0.0229, 0.0233, 0.0211, 0.0197], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-05-15 18:29:33,897 INFO [finetune.py:992] (0/2) Epoch 2, batch 2450, loss[loss=0.1749, simple_loss=0.2685, pruned_loss=0.04061, over 10476.00 frames. ], tot_loss[loss=0.1745, simple_loss=0.2624, pruned_loss=0.04333, over 2378009.46 frames. ], batch size: 68, lr: 4.97e-03, grad_scale: 8.0 2023-05-15 18:29:48,848 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.883e+02 2.690e+02 3.055e+02 4.005e+02 7.689e+02, threshold=6.109e+02, percent-clipped=1.0 2023-05-15 18:29:56,824 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=114716.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 18:30:10,130 INFO [finetune.py:992] (0/2) Epoch 2, batch 2500, loss[loss=0.2047, simple_loss=0.2886, pruned_loss=0.06044, over 12143.00 frames. ], tot_loss[loss=0.1749, simple_loss=0.2628, pruned_loss=0.04353, over 2375466.43 frames. ], batch size: 39, lr: 4.97e-03, grad_scale: 8.0 2023-05-15 18:30:40,245 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=114777.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 18:30:44,399 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=114783.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 18:30:45,770 INFO [finetune.py:992] (0/2) Epoch 2, batch 2550, loss[loss=0.2, simple_loss=0.2891, pruned_loss=0.05544, over 12024.00 frames. ], tot_loss[loss=0.175, simple_loss=0.2628, pruned_loss=0.04362, over 2375280.87 frames. ], batch size: 40, lr: 4.97e-03, grad_scale: 8.0 2023-05-15 18:31:01,298 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.981e+02 2.977e+02 3.410e+02 4.074e+02 7.350e+02, threshold=6.819e+02, percent-clipped=2.0 2023-05-15 18:31:13,238 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.58 vs. limit=2.0 2023-05-15 18:31:19,229 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=114831.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 18:31:21,915 INFO [finetune.py:992] (0/2) Epoch 2, batch 2600, loss[loss=0.161, simple_loss=0.2461, pruned_loss=0.0379, over 12195.00 frames. ], tot_loss[loss=0.1749, simple_loss=0.2628, pruned_loss=0.04348, over 2377178.78 frames. ], batch size: 29, lr: 4.97e-03, grad_scale: 8.0 2023-05-15 18:31:48,326 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=114871.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 18:31:57,991 INFO [finetune.py:992] (0/2) Epoch 2, batch 2650, loss[loss=0.1946, simple_loss=0.276, pruned_loss=0.05662, over 11098.00 frames. ], tot_loss[loss=0.1749, simple_loss=0.2626, pruned_loss=0.04359, over 2372124.17 frames. ], batch size: 55, lr: 4.97e-03, grad_scale: 8.0 2023-05-15 18:32:12,516 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.207e+02 2.908e+02 3.486e+02 4.320e+02 1.243e+03, threshold=6.972e+02, percent-clipped=4.0 2023-05-15 18:32:30,996 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=114932.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 18:32:32,965 INFO [finetune.py:992] (0/2) Epoch 2, batch 2700, loss[loss=0.1746, simple_loss=0.2694, pruned_loss=0.03995, over 12154.00 frames. ], tot_loss[loss=0.1743, simple_loss=0.2625, pruned_loss=0.04308, over 2373067.45 frames. ], batch size: 36, lr: 4.97e-03, grad_scale: 8.0 2023-05-15 18:33:09,001 INFO [finetune.py:992] (0/2) Epoch 2, batch 2750, loss[loss=0.1931, simple_loss=0.278, pruned_loss=0.05412, over 12100.00 frames. ], tot_loss[loss=0.1744, simple_loss=0.2625, pruned_loss=0.04317, over 2374872.14 frames. ], batch size: 33, lr: 4.97e-03, grad_scale: 8.0 2023-05-15 18:33:17,252 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.4525, 4.9015, 3.0904, 2.8462, 4.1555, 2.7140, 4.0652, 3.4640], device='cuda:0'), covar=tensor([0.0529, 0.0384, 0.0954, 0.1267, 0.0216, 0.1109, 0.0425, 0.0635], device='cuda:0'), in_proj_covar=tensor([0.0191, 0.0248, 0.0175, 0.0198, 0.0136, 0.0180, 0.0192, 0.0174], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-15 18:33:24,488 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.576e+02 2.889e+02 3.342e+02 4.054e+02 5.766e+02, threshold=6.684e+02, percent-clipped=0.0 2023-05-15 18:33:45,722 INFO [finetune.py:992] (0/2) Epoch 2, batch 2800, loss[loss=0.1694, simple_loss=0.2618, pruned_loss=0.0385, over 12144.00 frames. ], tot_loss[loss=0.1741, simple_loss=0.2621, pruned_loss=0.04306, over 2376482.92 frames. ], batch size: 36, lr: 4.97e-03, grad_scale: 8.0 2023-05-15 18:34:04,428 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 2023-05-15 18:34:12,027 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=115072.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 18:34:21,196 INFO [finetune.py:992] (0/2) Epoch 2, batch 2850, loss[loss=0.1955, simple_loss=0.2832, pruned_loss=0.05388, over 12048.00 frames. ], tot_loss[loss=0.1739, simple_loss=0.2618, pruned_loss=0.04299, over 2374132.70 frames. ], batch size: 42, lr: 4.97e-03, grad_scale: 8.0 2023-05-15 18:34:22,066 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=115086.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 18:34:36,494 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.197e+02 2.884e+02 3.510e+02 4.253e+02 7.197e+02, threshold=7.019e+02, percent-clipped=0.0 2023-05-15 18:34:39,785 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-05-15 18:34:53,544 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.4927, 2.8352, 3.5299, 4.6051, 3.9567, 4.5915, 3.9450, 3.1427], device='cuda:0'), covar=tensor([0.0029, 0.0325, 0.0135, 0.0027, 0.0095, 0.0050, 0.0097, 0.0339], device='cuda:0'), in_proj_covar=tensor([0.0078, 0.0121, 0.0101, 0.0073, 0.0098, 0.0109, 0.0085, 0.0134], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-15 18:34:56,915 INFO [finetune.py:992] (0/2) Epoch 2, batch 2900, loss[loss=0.1747, simple_loss=0.2672, pruned_loss=0.04111, over 12371.00 frames. ], tot_loss[loss=0.1743, simple_loss=0.2618, pruned_loss=0.04338, over 2367064.22 frames. ], batch size: 38, lr: 4.97e-03, grad_scale: 8.0 2023-05-15 18:34:57,064 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.2421, 5.2181, 5.0388, 4.5620, 4.7251, 5.1839, 4.8072, 4.6397], device='cuda:0'), covar=tensor([0.0690, 0.0836, 0.0608, 0.1292, 0.1039, 0.0698, 0.1400, 0.1014], device='cuda:0'), in_proj_covar=tensor([0.0574, 0.0506, 0.0460, 0.0577, 0.0373, 0.0651, 0.0707, 0.0514], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0004, 0.0004, 0.0003], device='cuda:0') 2023-05-15 18:35:05,536 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.2012, 4.8012, 5.1146, 4.4922, 4.8050, 4.5680, 5.1740, 4.7708], device='cuda:0'), covar=tensor([0.0222, 0.0303, 0.0262, 0.0263, 0.0293, 0.0267, 0.0185, 0.0294], device='cuda:0'), in_proj_covar=tensor([0.0224, 0.0234, 0.0254, 0.0229, 0.0229, 0.0230, 0.0210, 0.0195], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-05-15 18:35:05,594 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=115147.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 18:35:09,297 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 2023-05-15 18:35:33,044 INFO [finetune.py:992] (0/2) Epoch 2, batch 2950, loss[loss=0.1584, simple_loss=0.2475, pruned_loss=0.03469, over 12303.00 frames. ], tot_loss[loss=0.1739, simple_loss=0.2618, pruned_loss=0.04299, over 2363734.12 frames. ], batch size: 33, lr: 4.97e-03, grad_scale: 8.0 2023-05-15 18:35:35,632 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.31 vs. limit=2.0 2023-05-15 18:35:48,185 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.064e+02 2.824e+02 3.306e+02 3.914e+02 8.627e+02, threshold=6.612e+02, percent-clipped=2.0 2023-05-15 18:36:03,023 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=115227.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 18:36:08,591 INFO [finetune.py:992] (0/2) Epoch 2, batch 3000, loss[loss=0.1437, simple_loss=0.2267, pruned_loss=0.03038, over 12116.00 frames. ], tot_loss[loss=0.174, simple_loss=0.2616, pruned_loss=0.04319, over 2367840.36 frames. ], batch size: 30, lr: 4.97e-03, grad_scale: 8.0 2023-05-15 18:36:08,591 INFO [finetune.py:1017] (0/2) Computing validation loss 2023-05-15 18:36:26,382 INFO [finetune.py:1026] (0/2) Epoch 2, validation: loss=0.3401, simple_loss=0.4115, pruned_loss=0.1344, over 1020973.00 frames. 2023-05-15 18:36:26,383 INFO [finetune.py:1027] (0/2) Maximum memory allocated so far is 12745MB 2023-05-15 18:36:33,714 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.1788, 5.0778, 5.1529, 5.1995, 4.8227, 4.7710, 4.6690, 5.1597], device='cuda:0'), covar=tensor([0.0792, 0.0571, 0.0816, 0.0618, 0.1968, 0.1563, 0.0590, 0.0921], device='cuda:0'), in_proj_covar=tensor([0.0495, 0.0627, 0.0535, 0.0600, 0.0782, 0.0709, 0.0514, 0.0466], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0005, 0.0004, 0.0005, 0.0005, 0.0005, 0.0004, 0.0004], device='cuda:0') 2023-05-15 18:36:46,154 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.45 vs. limit=2.0 2023-05-15 18:37:02,839 INFO [finetune.py:992] (0/2) Epoch 2, batch 3050, loss[loss=0.1667, simple_loss=0.2488, pruned_loss=0.04237, over 12251.00 frames. ], tot_loss[loss=0.1723, simple_loss=0.2598, pruned_loss=0.04237, over 2380084.61 frames. ], batch size: 32, lr: 4.97e-03, grad_scale: 8.0 2023-05-15 18:37:17,765 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.082e+02 2.803e+02 3.228e+02 3.994e+02 5.351e+02, threshold=6.456e+02, percent-clipped=0.0 2023-05-15 18:37:38,444 INFO [finetune.py:992] (0/2) Epoch 2, batch 3100, loss[loss=0.1843, simple_loss=0.28, pruned_loss=0.04432, over 12161.00 frames. ], tot_loss[loss=0.1723, simple_loss=0.26, pruned_loss=0.04227, over 2381000.14 frames. ], batch size: 36, lr: 4.97e-03, grad_scale: 8.0 2023-05-15 18:38:04,158 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=115370.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 18:38:04,910 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=115371.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 18:38:05,491 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=115372.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 18:38:14,444 INFO [finetune.py:992] (0/2) Epoch 2, batch 3150, loss[loss=0.138, simple_loss=0.2179, pruned_loss=0.02903, over 11831.00 frames. ], tot_loss[loss=0.1734, simple_loss=0.2613, pruned_loss=0.04275, over 2373883.84 frames. ], batch size: 26, lr: 4.97e-03, grad_scale: 8.0 2023-05-15 18:38:16,844 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.4593, 4.2234, 4.2040, 4.6181, 3.3348, 4.1529, 2.9615, 4.2743], device='cuda:0'), covar=tensor([0.1406, 0.0592, 0.0862, 0.0523, 0.0912, 0.0519, 0.1459, 0.1264], device='cuda:0'), in_proj_covar=tensor([0.0223, 0.0262, 0.0294, 0.0354, 0.0238, 0.0239, 0.0253, 0.0384], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-15 18:38:29,738 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.907e+02 2.799e+02 3.424e+02 4.301e+02 1.225e+03, threshold=6.849e+02, percent-clipped=7.0 2023-05-15 18:38:36,689 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.57 vs. limit=2.0 2023-05-15 18:38:40,574 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=115420.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 18:38:46,969 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.5017, 4.9447, 2.9506, 2.4674, 4.3603, 2.5655, 4.1143, 3.5812], device='cuda:0'), covar=tensor([0.0579, 0.0421, 0.1056, 0.1626, 0.0215, 0.1338, 0.0389, 0.0640], device='cuda:0'), in_proj_covar=tensor([0.0191, 0.0248, 0.0173, 0.0198, 0.0136, 0.0179, 0.0192, 0.0174], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-15 18:38:48,374 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=115431.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 18:38:49,016 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=115432.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 18:38:50,952 INFO [finetune.py:992] (0/2) Epoch 2, batch 3200, loss[loss=0.1564, simple_loss=0.2436, pruned_loss=0.0346, over 12133.00 frames. ], tot_loss[loss=0.1725, simple_loss=0.2601, pruned_loss=0.04248, over 2377996.80 frames. ], batch size: 30, lr: 4.97e-03, grad_scale: 8.0 2023-05-15 18:38:55,950 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=115442.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 18:39:26,326 INFO [finetune.py:992] (0/2) Epoch 2, batch 3250, loss[loss=0.1676, simple_loss=0.2648, pruned_loss=0.0352, over 12149.00 frames. ], tot_loss[loss=0.1735, simple_loss=0.2612, pruned_loss=0.04288, over 2368580.07 frames. ], batch size: 36, lr: 4.97e-03, grad_scale: 8.0 2023-05-15 18:39:41,620 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.258e+02 3.049e+02 3.540e+02 3.983e+02 9.276e+02, threshold=7.080e+02, percent-clipped=3.0 2023-05-15 18:39:46,625 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.0585, 5.7151, 5.4173, 5.2774, 5.8640, 5.1686, 5.4263, 5.3033], device='cuda:0'), covar=tensor([0.1294, 0.0887, 0.0823, 0.2017, 0.0858, 0.2152, 0.1575, 0.1097], device='cuda:0'), in_proj_covar=tensor([0.0325, 0.0444, 0.0355, 0.0405, 0.0431, 0.0404, 0.0366, 0.0358], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-15 18:39:56,766 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=115527.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 18:40:02,261 INFO [finetune.py:992] (0/2) Epoch 2, batch 3300, loss[loss=0.1704, simple_loss=0.2612, pruned_loss=0.03978, over 12345.00 frames. ], tot_loss[loss=0.1738, simple_loss=0.2616, pruned_loss=0.04302, over 2378176.60 frames. ], batch size: 35, lr: 4.97e-03, grad_scale: 8.0 2023-05-15 18:40:31,690 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=115575.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 18:40:32,880 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.56 vs. limit=2.0 2023-05-15 18:40:38,834 INFO [finetune.py:992] (0/2) Epoch 2, batch 3350, loss[loss=0.1959, simple_loss=0.289, pruned_loss=0.05137, over 12118.00 frames. ], tot_loss[loss=0.1741, simple_loss=0.2623, pruned_loss=0.043, over 2379441.93 frames. ], batch size: 38, lr: 4.97e-03, grad_scale: 16.0 2023-05-15 18:40:48,105 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.78 vs. limit=2.0 2023-05-15 18:40:54,234 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.063e+02 2.912e+02 3.332e+02 3.940e+02 7.157e+02, threshold=6.663e+02, percent-clipped=1.0 2023-05-15 18:41:14,827 INFO [finetune.py:992] (0/2) Epoch 2, batch 3400, loss[loss=0.1588, simple_loss=0.2529, pruned_loss=0.03234, over 12101.00 frames. ], tot_loss[loss=0.174, simple_loss=0.2624, pruned_loss=0.04282, over 2381110.12 frames. ], batch size: 32, lr: 4.97e-03, grad_scale: 16.0 2023-05-15 18:41:51,019 INFO [finetune.py:992] (0/2) Epoch 2, batch 3450, loss[loss=0.1886, simple_loss=0.2848, pruned_loss=0.04625, over 12166.00 frames. ], tot_loss[loss=0.1735, simple_loss=0.2617, pruned_loss=0.04264, over 2384982.29 frames. ], batch size: 36, lr: 4.97e-03, grad_scale: 16.0 2023-05-15 18:42:06,755 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.982e+02 2.941e+02 3.439e+02 4.034e+02 9.078e+02, threshold=6.877e+02, percent-clipped=3.0 2023-05-15 18:42:21,106 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=115726.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 18:42:21,732 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=115727.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 18:42:25,584 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=2.90 vs. limit=5.0 2023-05-15 18:42:27,390 INFO [finetune.py:992] (0/2) Epoch 2, batch 3500, loss[loss=0.1879, simple_loss=0.2764, pruned_loss=0.04967, over 12133.00 frames. ], tot_loss[loss=0.1734, simple_loss=0.2616, pruned_loss=0.04259, over 2385215.37 frames. ], batch size: 39, lr: 4.97e-03, grad_scale: 16.0 2023-05-15 18:42:32,391 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=115742.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 18:43:02,731 INFO [finetune.py:992] (0/2) Epoch 2, batch 3550, loss[loss=0.1856, simple_loss=0.2806, pruned_loss=0.04524, over 12168.00 frames. ], tot_loss[loss=0.1733, simple_loss=0.2613, pruned_loss=0.04264, over 2378931.73 frames. ], batch size: 39, lr: 4.97e-03, grad_scale: 16.0 2023-05-15 18:43:06,342 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=115790.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 18:43:18,424 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.268e+02 3.032e+02 3.512e+02 4.471e+02 9.547e+02, threshold=7.024e+02, percent-clipped=4.0 2023-05-15 18:43:22,226 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.4930, 5.3326, 5.4017, 5.4306, 5.1145, 5.1079, 4.8411, 5.3774], device='cuda:0'), covar=tensor([0.0578, 0.0493, 0.0578, 0.0524, 0.1503, 0.1103, 0.0546, 0.0854], device='cuda:0'), in_proj_covar=tensor([0.0489, 0.0625, 0.0535, 0.0591, 0.0774, 0.0694, 0.0512, 0.0462], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0005, 0.0004, 0.0005, 0.0005, 0.0005, 0.0004, 0.0004], device='cuda:0') 2023-05-15 18:43:34,466 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.43 vs. limit=2.0 2023-05-15 18:43:38,972 INFO [finetune.py:992] (0/2) Epoch 2, batch 3600, loss[loss=0.1522, simple_loss=0.2383, pruned_loss=0.03303, over 12275.00 frames. ], tot_loss[loss=0.1741, simple_loss=0.2622, pruned_loss=0.04301, over 2378260.32 frames. ], batch size: 28, lr: 4.97e-03, grad_scale: 8.0 2023-05-15 18:43:47,014 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.4398, 2.2970, 3.1102, 4.2597, 2.5440, 4.5257, 4.3840, 4.5569], device='cuda:0'), covar=tensor([0.0114, 0.1233, 0.0475, 0.0160, 0.1041, 0.0124, 0.0101, 0.0078], device='cuda:0'), in_proj_covar=tensor([0.0118, 0.0199, 0.0181, 0.0110, 0.0186, 0.0171, 0.0167, 0.0117], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-15 18:44:14,851 INFO [finetune.py:992] (0/2) Epoch 2, batch 3650, loss[loss=0.1515, simple_loss=0.2292, pruned_loss=0.03687, over 12198.00 frames. ], tot_loss[loss=0.1732, simple_loss=0.2612, pruned_loss=0.04259, over 2380821.08 frames. ], batch size: 29, lr: 4.97e-03, grad_scale: 8.0 2023-05-15 18:44:20,083 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=115892.0, num_to_drop=1, layers_to_drop={1} 2023-05-15 18:44:30,537 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.807e+02 2.927e+02 3.317e+02 4.260e+02 7.259e+02, threshold=6.633e+02, percent-clipped=3.0 2023-05-15 18:44:50,647 INFO [finetune.py:992] (0/2) Epoch 2, batch 3700, loss[loss=0.1925, simple_loss=0.2883, pruned_loss=0.04838, over 12370.00 frames. ], tot_loss[loss=0.1739, simple_loss=0.2616, pruned_loss=0.04307, over 2379370.00 frames. ], batch size: 38, lr: 4.97e-03, grad_scale: 8.0 2023-05-15 18:45:02,367 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.6028, 3.5480, 3.3235, 3.1250, 2.8160, 2.7107, 3.4705, 2.1677], device='cuda:0'), covar=tensor([0.0286, 0.0098, 0.0121, 0.0137, 0.0294, 0.0300, 0.0094, 0.0373], device='cuda:0'), in_proj_covar=tensor([0.0180, 0.0155, 0.0144, 0.0172, 0.0194, 0.0187, 0.0150, 0.0193], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], device='cuda:0') 2023-05-15 18:45:04,495 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=115953.0, num_to_drop=1, layers_to_drop={1} 2023-05-15 18:45:27,084 INFO [finetune.py:992] (0/2) Epoch 2, batch 3750, loss[loss=0.1577, simple_loss=0.2479, pruned_loss=0.03377, over 12272.00 frames. ], tot_loss[loss=0.1735, simple_loss=0.2615, pruned_loss=0.04272, over 2382824.61 frames. ], batch size: 28, lr: 4.97e-03, grad_scale: 8.0 2023-05-15 18:45:36,509 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 2023-05-15 18:45:38,556 INFO [checkpoint.py:75] (0/2) Saving checkpoint to pruned_transducer_stateless7/exp_giga_finetune/checkpoint-16000.pt 2023-05-15 18:45:42,539 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.36 vs. limit=2.0 2023-05-15 18:45:46,262 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.103e+02 3.010e+02 3.508e+02 3.936e+02 8.193e+02, threshold=7.016e+02, percent-clipped=1.0 2023-05-15 18:45:51,389 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=116014.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 18:45:59,986 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=116026.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 18:46:00,634 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=116027.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 18:46:06,053 INFO [finetune.py:992] (0/2) Epoch 2, batch 3800, loss[loss=0.1865, simple_loss=0.2694, pruned_loss=0.05184, over 12356.00 frames. ], tot_loss[loss=0.1731, simple_loss=0.2608, pruned_loss=0.04272, over 2388890.40 frames. ], batch size: 35, lr: 4.97e-03, grad_scale: 8.0 2023-05-15 18:46:30,938 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.5043, 2.5572, 3.6282, 4.5148, 4.0333, 4.5686, 3.8949, 3.4700], device='cuda:0'), covar=tensor([0.0028, 0.0356, 0.0116, 0.0026, 0.0085, 0.0056, 0.0081, 0.0254], device='cuda:0'), in_proj_covar=tensor([0.0078, 0.0120, 0.0100, 0.0073, 0.0098, 0.0109, 0.0084, 0.0133], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-15 18:46:33,679 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=116074.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 18:46:34,432 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=116075.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 18:46:34,570 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=116075.0, num_to_drop=1, layers_to_drop={0} 2023-05-15 18:46:41,063 INFO [finetune.py:992] (0/2) Epoch 2, batch 3850, loss[loss=0.1729, simple_loss=0.2668, pruned_loss=0.03953, over 12085.00 frames. ], tot_loss[loss=0.174, simple_loss=0.2617, pruned_loss=0.04311, over 2384967.07 frames. ], batch size: 32, lr: 4.97e-03, grad_scale: 8.0 2023-05-15 18:46:52,685 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.46 vs. limit=2.0 2023-05-15 18:46:55,799 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=116105.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 18:46:56,949 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.021e+02 3.031e+02 3.770e+02 4.514e+02 9.662e+02, threshold=7.540e+02, percent-clipped=3.0 2023-05-15 18:47:03,565 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=116116.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 18:47:17,410 INFO [finetune.py:992] (0/2) Epoch 2, batch 3900, loss[loss=0.142, simple_loss=0.2192, pruned_loss=0.03235, over 12342.00 frames. ], tot_loss[loss=0.1739, simple_loss=0.2615, pruned_loss=0.04319, over 2391936.16 frames. ], batch size: 30, lr: 4.97e-03, grad_scale: 8.0 2023-05-15 18:47:38,963 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.39 vs. limit=2.0 2023-05-15 18:47:39,341 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=116166.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 18:47:47,184 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=116177.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 18:47:52,787 INFO [finetune.py:992] (0/2) Epoch 2, batch 3950, loss[loss=0.1634, simple_loss=0.2489, pruned_loss=0.0389, over 10588.00 frames. ], tot_loss[loss=0.1738, simple_loss=0.2612, pruned_loss=0.04315, over 2391305.28 frames. ], batch size: 68, lr: 4.97e-03, grad_scale: 8.0 2023-05-15 18:47:58,626 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.4251, 6.1113, 5.6543, 5.6628, 6.2261, 5.4546, 5.8773, 5.7772], device='cuda:0'), covar=tensor([0.1176, 0.0820, 0.0928, 0.1953, 0.0810, 0.2085, 0.1263, 0.0939], device='cuda:0'), in_proj_covar=tensor([0.0318, 0.0436, 0.0348, 0.0395, 0.0423, 0.0396, 0.0358, 0.0352], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-15 18:48:08,575 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.014e+02 2.779e+02 3.406e+02 4.033e+02 7.927e+02, threshold=6.812e+02, percent-clipped=2.0 2023-05-15 18:48:25,231 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.44 vs. limit=2.0 2023-05-15 18:48:28,887 INFO [finetune.py:992] (0/2) Epoch 2, batch 4000, loss[loss=0.1441, simple_loss=0.2241, pruned_loss=0.03203, over 11977.00 frames. ], tot_loss[loss=0.1743, simple_loss=0.2618, pruned_loss=0.04334, over 2384304.13 frames. ], batch size: 28, lr: 4.97e-03, grad_scale: 8.0 2023-05-15 18:48:38,187 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=116248.0, num_to_drop=1, layers_to_drop={1} 2023-05-15 18:48:46,817 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.1124, 2.2403, 2.9874, 4.0140, 2.1165, 4.1751, 4.0442, 4.2465], device='cuda:0'), covar=tensor([0.0129, 0.1189, 0.0471, 0.0136, 0.1279, 0.0191, 0.0153, 0.0100], device='cuda:0'), in_proj_covar=tensor([0.0120, 0.0204, 0.0186, 0.0112, 0.0190, 0.0176, 0.0171, 0.0120], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-15 18:48:49,619 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.9346, 4.8929, 4.7741, 4.8036, 4.3980, 4.8928, 4.8703, 5.0997], device='cuda:0'), covar=tensor([0.0218, 0.0148, 0.0170, 0.0365, 0.0795, 0.0250, 0.0152, 0.0183], device='cuda:0'), in_proj_covar=tensor([0.0181, 0.0185, 0.0183, 0.0226, 0.0232, 0.0197, 0.0170, 0.0218], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0005, 0.0005, 0.0005, 0.0004, 0.0005], device='cuda:0') 2023-05-15 18:49:01,589 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.9771, 3.6202, 5.0648, 2.7271, 3.0121, 3.7998, 3.4960, 3.8459], device='cuda:0'), covar=tensor([0.0284, 0.0898, 0.0268, 0.1067, 0.1714, 0.1302, 0.1044, 0.1028], device='cuda:0'), in_proj_covar=tensor([0.0231, 0.0226, 0.0230, 0.0179, 0.0233, 0.0276, 0.0221, 0.0247], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:0') 2023-05-15 18:49:05,000 INFO [finetune.py:992] (0/2) Epoch 2, batch 4050, loss[loss=0.1643, simple_loss=0.2424, pruned_loss=0.04314, over 11817.00 frames. ], tot_loss[loss=0.1734, simple_loss=0.2608, pruned_loss=0.04299, over 2381890.51 frames. ], batch size: 26, lr: 4.97e-03, grad_scale: 8.0 2023-05-15 18:49:20,558 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.874e+02 2.857e+02 3.441e+02 3.984e+02 7.490e+02, threshold=6.881e+02, percent-clipped=3.0 2023-05-15 18:49:32,043 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=116323.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 18:49:38,943 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.5487, 4.2310, 4.5018, 3.9426, 4.2478, 4.0114, 4.5102, 4.2521], device='cuda:0'), covar=tensor([0.0299, 0.0321, 0.0339, 0.0283, 0.0308, 0.0326, 0.0241, 0.0489], device='cuda:0'), in_proj_covar=tensor([0.0227, 0.0237, 0.0256, 0.0230, 0.0233, 0.0233, 0.0212, 0.0195], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-05-15 18:49:40,133 INFO [finetune.py:992] (0/2) Epoch 2, batch 4100, loss[loss=0.1922, simple_loss=0.2829, pruned_loss=0.05072, over 11992.00 frames. ], tot_loss[loss=0.1741, simple_loss=0.2614, pruned_loss=0.04338, over 2375774.31 frames. ], batch size: 40, lr: 4.97e-03, grad_scale: 8.0 2023-05-15 18:49:41,369 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.91 vs. limit=2.0 2023-05-15 18:49:45,619 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=3.00 vs. limit=5.0 2023-05-15 18:49:49,368 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.36 vs. limit=2.0 2023-05-15 18:50:05,299 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=116370.0, num_to_drop=1, layers_to_drop={3} 2023-05-15 18:50:09,161 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=5.03 vs. limit=5.0 2023-05-15 18:50:16,029 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=116384.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 18:50:16,525 INFO [finetune.py:992] (0/2) Epoch 2, batch 4150, loss[loss=0.1392, simple_loss=0.2341, pruned_loss=0.02218, over 12178.00 frames. ], tot_loss[loss=0.1737, simple_loss=0.2613, pruned_loss=0.04309, over 2385478.00 frames. ], batch size: 31, lr: 4.97e-03, grad_scale: 8.0 2023-05-15 18:50:32,533 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.000e+02 2.865e+02 3.386e+02 4.375e+02 9.798e+02, threshold=6.772e+02, percent-clipped=6.0 2023-05-15 18:50:53,144 INFO [finetune.py:992] (0/2) Epoch 2, batch 4200, loss[loss=0.1794, simple_loss=0.2671, pruned_loss=0.04586, over 12258.00 frames. ], tot_loss[loss=0.1736, simple_loss=0.2614, pruned_loss=0.0429, over 2383938.86 frames. ], batch size: 37, lr: 4.97e-03, grad_scale: 8.0 2023-05-15 18:50:54,870 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.81 vs. limit=2.0 2023-05-15 18:51:00,831 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=116446.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 18:51:11,254 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=116461.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 18:51:18,574 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.6473, 2.7167, 3.3940, 4.5490, 2.5770, 4.6150, 4.5740, 4.8475], device='cuda:0'), covar=tensor([0.0110, 0.1062, 0.0394, 0.0131, 0.1125, 0.0173, 0.0131, 0.0082], device='cuda:0'), in_proj_covar=tensor([0.0119, 0.0203, 0.0185, 0.0111, 0.0189, 0.0175, 0.0169, 0.0120], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-15 18:51:19,187 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=116472.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 18:51:28,452 INFO [finetune.py:992] (0/2) Epoch 2, batch 4250, loss[loss=0.1643, simple_loss=0.2574, pruned_loss=0.03556, over 11774.00 frames. ], tot_loss[loss=0.1742, simple_loss=0.2619, pruned_loss=0.04324, over 2386955.64 frames. ], batch size: 44, lr: 4.97e-03, grad_scale: 8.0 2023-05-15 18:51:42,739 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=116505.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 18:51:43,982 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.761e+02 2.921e+02 3.307e+02 4.423e+02 2.186e+03, threshold=6.613e+02, percent-clipped=3.0 2023-05-15 18:51:44,241 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=116507.0, num_to_drop=1, layers_to_drop={1} 2023-05-15 18:51:54,962 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.2643, 4.7119, 2.8519, 2.4623, 4.1451, 2.3222, 4.0629, 3.0732], device='cuda:0'), covar=tensor([0.0630, 0.0367, 0.0958, 0.1378, 0.0239, 0.1306, 0.0340, 0.0810], device='cuda:0'), in_proj_covar=tensor([0.0192, 0.0249, 0.0174, 0.0198, 0.0136, 0.0182, 0.0193, 0.0176], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-15 18:52:04,674 INFO [finetune.py:992] (0/2) Epoch 2, batch 4300, loss[loss=0.1972, simple_loss=0.2781, pruned_loss=0.05817, over 8382.00 frames. ], tot_loss[loss=0.1751, simple_loss=0.2629, pruned_loss=0.0437, over 2383699.10 frames. ], batch size: 98, lr: 4.97e-03, grad_scale: 8.0 2023-05-15 18:52:13,973 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=116548.0, num_to_drop=1, layers_to_drop={0} 2023-05-15 18:52:26,834 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=116566.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 18:52:40,691 INFO [finetune.py:992] (0/2) Epoch 2, batch 4350, loss[loss=0.1802, simple_loss=0.2698, pruned_loss=0.04532, over 12319.00 frames. ], tot_loss[loss=0.1758, simple_loss=0.2635, pruned_loss=0.04407, over 2380994.88 frames. ], batch size: 34, lr: 4.97e-03, grad_scale: 8.0 2023-05-15 18:52:48,467 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=116596.0, num_to_drop=1, layers_to_drop={0} 2023-05-15 18:52:53,038 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.8252, 4.5199, 4.6903, 4.7882, 4.4767, 4.7282, 4.6472, 2.7325], device='cuda:0'), covar=tensor([0.0079, 0.0055, 0.0074, 0.0053, 0.0048, 0.0080, 0.0065, 0.0606], device='cuda:0'), in_proj_covar=tensor([0.0063, 0.0072, 0.0076, 0.0070, 0.0056, 0.0085, 0.0074, 0.0087], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-15 18:52:56,312 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.055e+02 3.059e+02 3.531e+02 4.344e+02 1.283e+03, threshold=7.062e+02, percent-clipped=4.0 2023-05-15 18:53:01,677 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 2023-05-15 18:53:04,246 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=116618.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 18:53:16,090 INFO [finetune.py:992] (0/2) Epoch 2, batch 4400, loss[loss=0.1741, simple_loss=0.2664, pruned_loss=0.04088, over 12303.00 frames. ], tot_loss[loss=0.1752, simple_loss=0.2625, pruned_loss=0.04393, over 2375479.44 frames. ], batch size: 34, lr: 4.97e-03, grad_scale: 8.0 2023-05-15 18:53:41,642 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=116670.0, num_to_drop=1, layers_to_drop={0} 2023-05-15 18:53:47,801 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=116679.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 18:53:47,949 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=116679.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 18:53:49,294 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.3109, 4.9886, 5.2112, 5.2006, 5.0049, 5.1897, 5.0227, 3.0077], device='cuda:0'), covar=tensor([0.0078, 0.0056, 0.0062, 0.0048, 0.0036, 0.0081, 0.0079, 0.0510], device='cuda:0'), in_proj_covar=tensor([0.0063, 0.0073, 0.0076, 0.0070, 0.0057, 0.0086, 0.0075, 0.0088], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-15 18:53:51,913 INFO [finetune.py:992] (0/2) Epoch 2, batch 4450, loss[loss=0.174, simple_loss=0.2599, pruned_loss=0.04405, over 11239.00 frames. ], tot_loss[loss=0.1743, simple_loss=0.2617, pruned_loss=0.04347, over 2378423.22 frames. ], batch size: 55, lr: 4.97e-03, grad_scale: 8.0 2023-05-15 18:53:53,286 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=4.69 vs. limit=5.0 2023-05-15 18:54:05,689 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=116704.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 18:54:07,667 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.882e+02 2.725e+02 3.200e+02 3.613e+02 8.473e+02, threshold=6.400e+02, percent-clipped=2.0 2023-05-15 18:54:08,661 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2023-05-15 18:54:15,290 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=116718.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 18:54:15,450 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.2120, 4.5789, 2.6896, 2.4997, 3.9509, 2.4665, 3.9103, 2.9876], device='cuda:0'), covar=tensor([0.0646, 0.0497, 0.1118, 0.1520, 0.0228, 0.1194, 0.0437, 0.0829], device='cuda:0'), in_proj_covar=tensor([0.0189, 0.0246, 0.0172, 0.0196, 0.0135, 0.0179, 0.0192, 0.0174], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-15 18:54:28,081 INFO [finetune.py:992] (0/2) Epoch 2, batch 4500, loss[loss=0.2192, simple_loss=0.2988, pruned_loss=0.06978, over 12120.00 frames. ], tot_loss[loss=0.1743, simple_loss=0.2617, pruned_loss=0.04345, over 2378779.33 frames. ], batch size: 39, lr: 4.97e-03, grad_scale: 8.0 2023-05-15 18:54:37,361 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.9805, 4.6352, 4.9483, 4.4125, 4.6572, 4.4738, 5.0024, 4.6650], device='cuda:0'), covar=tensor([0.0250, 0.0286, 0.0268, 0.0220, 0.0291, 0.0240, 0.0175, 0.0354], device='cuda:0'), in_proj_covar=tensor([0.0225, 0.0233, 0.0252, 0.0228, 0.0230, 0.0230, 0.0209, 0.0194], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-05-15 18:54:46,722 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=116761.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 18:54:47,464 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.9996, 4.9028, 4.9150, 5.0019, 4.6247, 4.6959, 4.5436, 4.9692], device='cuda:0'), covar=tensor([0.0707, 0.0542, 0.0779, 0.0547, 0.1822, 0.1114, 0.0514, 0.0803], device='cuda:0'), in_proj_covar=tensor([0.0485, 0.0629, 0.0533, 0.0590, 0.0765, 0.0693, 0.0510, 0.0455], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0005, 0.0004, 0.0005, 0.0005, 0.0005, 0.0004, 0.0004], device='cuda:0') 2023-05-15 18:54:49,666 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=116765.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 18:54:54,657 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=116772.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 18:55:03,568 INFO [finetune.py:992] (0/2) Epoch 2, batch 4550, loss[loss=0.1844, simple_loss=0.2587, pruned_loss=0.05507, over 12088.00 frames. ], tot_loss[loss=0.1745, simple_loss=0.262, pruned_loss=0.04355, over 2381495.06 frames. ], batch size: 32, lr: 4.97e-03, grad_scale: 8.0 2023-05-15 18:55:13,820 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 2023-05-15 18:55:16,608 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=116802.0, num_to_drop=1, layers_to_drop={3} 2023-05-15 18:55:16,737 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=116802.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 18:55:20,078 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.014e+02 2.993e+02 3.616e+02 4.229e+02 7.475e+02, threshold=7.232e+02, percent-clipped=4.0 2023-05-15 18:55:21,580 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=116809.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 18:55:29,381 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=116820.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 18:55:40,106 INFO [finetune.py:992] (0/2) Epoch 2, batch 4600, loss[loss=0.1774, simple_loss=0.265, pruned_loss=0.04486, over 12195.00 frames. ], tot_loss[loss=0.1748, simple_loss=0.2625, pruned_loss=0.04355, over 2378693.90 frames. ], batch size: 35, lr: 4.97e-03, grad_scale: 8.0 2023-05-15 18:55:48,979 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.3918, 4.6182, 4.0849, 5.0185, 4.6016, 2.8520, 4.3541, 3.0933], device='cuda:0'), covar=tensor([0.0648, 0.0716, 0.1319, 0.0306, 0.1002, 0.1553, 0.0903, 0.3029], device='cuda:0'), in_proj_covar=tensor([0.0290, 0.0364, 0.0342, 0.0253, 0.0354, 0.0257, 0.0326, 0.0356], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-15 18:55:52,331 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.9019, 4.9085, 4.7492, 4.8609, 4.2788, 4.9545, 4.9207, 5.0636], device='cuda:0'), covar=tensor([0.0197, 0.0129, 0.0191, 0.0214, 0.0781, 0.0221, 0.0147, 0.0187], device='cuda:0'), in_proj_covar=tensor([0.0176, 0.0181, 0.0180, 0.0221, 0.0226, 0.0194, 0.0167, 0.0214], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0005, 0.0005, 0.0005, 0.0004, 0.0005], device='cuda:0') 2023-05-15 18:55:59,263 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=116861.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 18:56:00,824 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=116863.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 18:56:16,380 INFO [finetune.py:992] (0/2) Epoch 2, batch 4650, loss[loss=0.1638, simple_loss=0.2349, pruned_loss=0.04636, over 12113.00 frames. ], tot_loss[loss=0.1751, simple_loss=0.2627, pruned_loss=0.04373, over 2381023.22 frames. ], batch size: 30, lr: 4.96e-03, grad_scale: 8.0 2023-05-15 18:56:31,908 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.613e+02 2.962e+02 3.393e+02 3.992e+02 1.065e+03, threshold=6.785e+02, percent-clipped=1.0 2023-05-15 18:56:51,493 INFO [finetune.py:992] (0/2) Epoch 2, batch 4700, loss[loss=0.1345, simple_loss=0.2137, pruned_loss=0.02759, over 12013.00 frames. ], tot_loss[loss=0.1751, simple_loss=0.263, pruned_loss=0.04363, over 2380221.43 frames. ], batch size: 28, lr: 4.96e-03, grad_scale: 8.0 2023-05-15 18:56:52,617 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.84 vs. limit=2.0 2023-05-15 18:57:19,850 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=116974.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 18:57:23,491 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=116979.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 18:57:27,773 INFO [finetune.py:992] (0/2) Epoch 2, batch 4750, loss[loss=0.1836, simple_loss=0.2754, pruned_loss=0.04592, over 12295.00 frames. ], tot_loss[loss=0.1753, simple_loss=0.2631, pruned_loss=0.04375, over 2376761.58 frames. ], batch size: 34, lr: 4.96e-03, grad_scale: 8.0 2023-05-15 18:57:30,771 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.3404, 5.1968, 5.2149, 5.3184, 4.9103, 4.9876, 4.7830, 5.3030], device='cuda:0'), covar=tensor([0.0707, 0.0573, 0.0766, 0.0570, 0.1761, 0.1297, 0.0556, 0.0854], device='cuda:0'), in_proj_covar=tensor([0.0490, 0.0631, 0.0537, 0.0592, 0.0768, 0.0702, 0.0516, 0.0457], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0005, 0.0004, 0.0005, 0.0005, 0.0005, 0.0004, 0.0004], device='cuda:0') 2023-05-15 18:57:44,289 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.002e+02 2.939e+02 3.466e+02 4.060e+02 9.560e+02, threshold=6.933e+02, percent-clipped=2.0 2023-05-15 18:57:50,121 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.3923, 4.9134, 5.3765, 4.6976, 5.0264, 4.7665, 5.4213, 5.1022], device='cuda:0'), covar=tensor([0.0241, 0.0320, 0.0268, 0.0222, 0.0265, 0.0255, 0.0155, 0.0256], device='cuda:0'), in_proj_covar=tensor([0.0226, 0.0234, 0.0253, 0.0230, 0.0230, 0.0231, 0.0210, 0.0194], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-05-15 18:57:58,579 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=117027.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 18:58:04,250 INFO [finetune.py:992] (0/2) Epoch 2, batch 4800, loss[loss=0.191, simple_loss=0.2708, pruned_loss=0.05557, over 12091.00 frames. ], tot_loss[loss=0.1762, simple_loss=0.264, pruned_loss=0.04424, over 2373212.11 frames. ], batch size: 32, lr: 4.96e-03, grad_scale: 8.0 2023-05-15 18:58:22,020 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=117060.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 18:58:27,565 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.3868, 2.3552, 3.5980, 4.3554, 3.8991, 4.4170, 3.8822, 2.9417], device='cuda:0'), covar=tensor([0.0033, 0.0394, 0.0120, 0.0031, 0.0102, 0.0062, 0.0092, 0.0336], device='cuda:0'), in_proj_covar=tensor([0.0077, 0.0118, 0.0100, 0.0072, 0.0097, 0.0108, 0.0082, 0.0131], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-15 18:58:39,686 INFO [finetune.py:992] (0/2) Epoch 2, batch 4850, loss[loss=0.1961, simple_loss=0.2912, pruned_loss=0.05047, over 12167.00 frames. ], tot_loss[loss=0.1769, simple_loss=0.2646, pruned_loss=0.04463, over 2366610.79 frames. ], batch size: 36, lr: 4.96e-03, grad_scale: 8.0 2023-05-15 18:58:52,666 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=117102.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 18:58:53,716 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.32 vs. limit=2.0 2023-05-15 18:58:56,077 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.262e+02 2.964e+02 3.334e+02 4.015e+02 7.759e+02, threshold=6.668e+02, percent-clipped=1.0 2023-05-15 18:59:16,084 INFO [finetune.py:992] (0/2) Epoch 2, batch 4900, loss[loss=0.1776, simple_loss=0.2705, pruned_loss=0.0424, over 12086.00 frames. ], tot_loss[loss=0.1762, simple_loss=0.2641, pruned_loss=0.04417, over 2377273.89 frames. ], batch size: 32, lr: 4.96e-03, grad_scale: 8.0 2023-05-15 18:59:21,224 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.7666, 2.2199, 2.8637, 3.6776, 2.1445, 3.8375, 3.6770, 3.9182], device='cuda:0'), covar=tensor([0.0137, 0.1090, 0.0455, 0.0121, 0.1087, 0.0220, 0.0192, 0.0082], device='cuda:0'), in_proj_covar=tensor([0.0120, 0.0203, 0.0185, 0.0113, 0.0189, 0.0176, 0.0168, 0.0119], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-15 18:59:26,761 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=117150.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 18:59:33,152 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=117158.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 18:59:35,305 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=117161.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 18:59:52,178 INFO [finetune.py:992] (0/2) Epoch 2, batch 4950, loss[loss=0.1615, simple_loss=0.2404, pruned_loss=0.04129, over 12172.00 frames. ], tot_loss[loss=0.177, simple_loss=0.2649, pruned_loss=0.04459, over 2367192.27 frames. ], batch size: 29, lr: 4.96e-03, grad_scale: 8.0 2023-05-15 19:00:08,167 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.664e+02 2.817e+02 3.366e+02 4.085e+02 9.670e+02, threshold=6.732e+02, percent-clipped=4.0 2023-05-15 19:00:09,695 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=117209.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 19:00:18,353 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=117221.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 19:00:28,755 INFO [finetune.py:992] (0/2) Epoch 2, batch 5000, loss[loss=0.1887, simple_loss=0.2813, pruned_loss=0.04803, over 12362.00 frames. ], tot_loss[loss=0.1762, simple_loss=0.2639, pruned_loss=0.04428, over 2366163.95 frames. ], batch size: 35, lr: 4.96e-03, grad_scale: 8.0 2023-05-15 19:00:40,505 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.34 vs. limit=2.0 2023-05-15 19:00:56,704 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=117274.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 19:01:01,060 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.6196, 2.8220, 4.4282, 4.5396, 2.9870, 2.6631, 2.8007, 2.1006], device='cuda:0'), covar=tensor([0.1398, 0.2756, 0.0395, 0.0380, 0.1026, 0.1859, 0.2356, 0.3474], device='cuda:0'), in_proj_covar=tensor([0.0276, 0.0370, 0.0263, 0.0286, 0.0250, 0.0276, 0.0352, 0.0350], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-15 19:01:02,433 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=117282.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 19:01:04,409 INFO [finetune.py:992] (0/2) Epoch 2, batch 5050, loss[loss=0.1605, simple_loss=0.2539, pruned_loss=0.03353, over 12109.00 frames. ], tot_loss[loss=0.1755, simple_loss=0.263, pruned_loss=0.04398, over 2364934.61 frames. ], batch size: 39, lr: 4.96e-03, grad_scale: 8.0 2023-05-15 19:01:20,674 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.881e+02 3.022e+02 3.419e+02 3.930e+02 8.423e+02, threshold=6.837e+02, percent-clipped=1.0 2023-05-15 19:01:31,445 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=117322.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 19:01:36,468 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=117329.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 19:01:40,331 INFO [finetune.py:992] (0/2) Epoch 2, batch 5100, loss[loss=0.1899, simple_loss=0.2746, pruned_loss=0.05258, over 11307.00 frames. ], tot_loss[loss=0.1759, simple_loss=0.2634, pruned_loss=0.04421, over 2362722.94 frames. ], batch size: 55, lr: 4.96e-03, grad_scale: 8.0 2023-05-15 19:01:58,286 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=117360.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 19:02:06,863 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.6393, 3.7226, 3.3308, 3.2376, 2.9507, 2.7269, 3.7469, 2.2973], device='cuda:0'), covar=tensor([0.0296, 0.0121, 0.0129, 0.0142, 0.0291, 0.0268, 0.0082, 0.0375], device='cuda:0'), in_proj_covar=tensor([0.0181, 0.0155, 0.0146, 0.0174, 0.0196, 0.0191, 0.0153, 0.0197], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], device='cuda:0') 2023-05-15 19:02:16,507 INFO [finetune.py:992] (0/2) Epoch 2, batch 5150, loss[loss=0.2435, simple_loss=0.3166, pruned_loss=0.0852, over 10635.00 frames. ], tot_loss[loss=0.1754, simple_loss=0.2629, pruned_loss=0.04398, over 2364145.52 frames. ], batch size: 68, lr: 4.96e-03, grad_scale: 8.0 2023-05-15 19:02:20,295 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=117390.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 19:02:32,751 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.024e+02 2.982e+02 3.418e+02 4.155e+02 8.976e+02, threshold=6.836e+02, percent-clipped=2.0 2023-05-15 19:02:33,543 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=117408.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 19:02:52,219 INFO [finetune.py:992] (0/2) Epoch 2, batch 5200, loss[loss=0.1773, simple_loss=0.2748, pruned_loss=0.03989, over 12035.00 frames. ], tot_loss[loss=0.1756, simple_loss=0.2631, pruned_loss=0.04405, over 2366122.08 frames. ], batch size: 40, lr: 4.96e-03, grad_scale: 8.0 2023-05-15 19:03:08,981 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=117458.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 19:03:27,701 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=2.28 vs. limit=5.0 2023-05-15 19:03:28,091 INFO [finetune.py:992] (0/2) Epoch 2, batch 5250, loss[loss=0.1603, simple_loss=0.2429, pruned_loss=0.03879, over 12354.00 frames. ], tot_loss[loss=0.1756, simple_loss=0.2631, pruned_loss=0.04403, over 2371562.03 frames. ], batch size: 31, lr: 4.96e-03, grad_scale: 8.0 2023-05-15 19:03:29,244 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 2023-05-15 19:03:42,937 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=117506.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 19:03:43,578 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.118e+02 2.779e+02 3.417e+02 4.487e+02 9.626e+02, threshold=6.834e+02, percent-clipped=5.0 2023-05-15 19:04:04,143 INFO [finetune.py:992] (0/2) Epoch 2, batch 5300, loss[loss=0.1539, simple_loss=0.2393, pruned_loss=0.03427, over 12302.00 frames. ], tot_loss[loss=0.1757, simple_loss=0.2631, pruned_loss=0.04409, over 2368843.81 frames. ], batch size: 33, lr: 4.96e-03, grad_scale: 8.0 2023-05-15 19:04:21,789 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.1990, 5.1695, 4.9872, 5.0915, 4.6845, 5.1430, 5.1679, 5.3496], device='cuda:0'), covar=tensor([0.0154, 0.0118, 0.0152, 0.0227, 0.0631, 0.0219, 0.0139, 0.0152], device='cuda:0'), in_proj_covar=tensor([0.0177, 0.0180, 0.0179, 0.0221, 0.0226, 0.0194, 0.0166, 0.0214], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0005, 0.0005, 0.0005, 0.0004, 0.0005], device='cuda:0') 2023-05-15 19:04:33,721 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=117577.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 19:04:39,283 INFO [finetune.py:992] (0/2) Epoch 2, batch 5350, loss[loss=0.1619, simple_loss=0.2616, pruned_loss=0.03105, over 12015.00 frames. ], tot_loss[loss=0.1753, simple_loss=0.2627, pruned_loss=0.04399, over 2372171.80 frames. ], batch size: 40, lr: 4.96e-03, grad_scale: 8.0 2023-05-15 19:04:55,725 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.861e+02 2.946e+02 3.399e+02 4.247e+02 1.173e+03, threshold=6.799e+02, percent-clipped=4.0 2023-05-15 19:05:15,435 INFO [finetune.py:992] (0/2) Epoch 2, batch 5400, loss[loss=0.1858, simple_loss=0.2706, pruned_loss=0.05056, over 12068.00 frames. ], tot_loss[loss=0.1745, simple_loss=0.2618, pruned_loss=0.04357, over 2381833.88 frames. ], batch size: 40, lr: 4.96e-03, grad_scale: 8.0 2023-05-15 19:05:51,783 INFO [finetune.py:992] (0/2) Epoch 2, batch 5450, loss[loss=0.1456, simple_loss=0.2308, pruned_loss=0.0302, over 12251.00 frames. ], tot_loss[loss=0.1736, simple_loss=0.261, pruned_loss=0.04308, over 2383187.97 frames. ], batch size: 32, lr: 4.96e-03, grad_scale: 8.0 2023-05-15 19:05:51,882 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=117685.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 19:06:03,430 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-05-15 19:06:03,541 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=3.37 vs. limit=5.0 2023-05-15 19:06:07,351 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.684e+02 2.792e+02 3.327e+02 3.916e+02 6.376e+02, threshold=6.653e+02, percent-clipped=0.0 2023-05-15 19:06:08,302 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.5333, 3.6072, 3.1545, 3.2414, 2.8386, 2.6353, 3.6190, 2.1562], device='cuda:0'), covar=tensor([0.0295, 0.0114, 0.0155, 0.0130, 0.0325, 0.0313, 0.0101, 0.0404], device='cuda:0'), in_proj_covar=tensor([0.0181, 0.0155, 0.0146, 0.0173, 0.0195, 0.0189, 0.0153, 0.0197], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], device='cuda:0') 2023-05-15 19:06:21,985 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2023-05-15 19:06:27,139 INFO [finetune.py:992] (0/2) Epoch 2, batch 5500, loss[loss=0.1477, simple_loss=0.2366, pruned_loss=0.02941, over 12342.00 frames. ], tot_loss[loss=0.1749, simple_loss=0.2625, pruned_loss=0.04361, over 2371165.20 frames. ], batch size: 31, lr: 4.96e-03, grad_scale: 8.0 2023-05-15 19:06:59,918 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.85 vs. limit=2.0 2023-05-15 19:07:03,036 INFO [finetune.py:992] (0/2) Epoch 2, batch 5550, loss[loss=0.1599, simple_loss=0.2442, pruned_loss=0.0378, over 12286.00 frames. ], tot_loss[loss=0.1746, simple_loss=0.2623, pruned_loss=0.0434, over 2380142.06 frames. ], batch size: 33, lr: 4.96e-03, grad_scale: 8.0 2023-05-15 19:07:19,628 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.047e+02 2.985e+02 3.531e+02 4.289e+02 1.406e+03, threshold=7.062e+02, percent-clipped=6.0 2023-05-15 19:07:38,884 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.4122, 5.1802, 5.3306, 5.3194, 4.9087, 4.9925, 4.7721, 5.3092], device='cuda:0'), covar=tensor([0.0565, 0.0617, 0.0540, 0.0575, 0.2141, 0.1216, 0.0524, 0.0893], device='cuda:0'), in_proj_covar=tensor([0.0491, 0.0637, 0.0534, 0.0594, 0.0779, 0.0706, 0.0517, 0.0462], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0005, 0.0004, 0.0005, 0.0005, 0.0005, 0.0004, 0.0004], device='cuda:0') 2023-05-15 19:07:39,499 INFO [finetune.py:992] (0/2) Epoch 2, batch 5600, loss[loss=0.178, simple_loss=0.2622, pruned_loss=0.04688, over 12099.00 frames. ], tot_loss[loss=0.1748, simple_loss=0.2625, pruned_loss=0.04354, over 2384081.32 frames. ], batch size: 33, lr: 4.96e-03, grad_scale: 16.0 2023-05-15 19:08:09,116 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=117877.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 19:08:15,298 INFO [finetune.py:992] (0/2) Epoch 2, batch 5650, loss[loss=0.1879, simple_loss=0.2717, pruned_loss=0.05209, over 12308.00 frames. ], tot_loss[loss=0.1754, simple_loss=0.2627, pruned_loss=0.04408, over 2379465.09 frames. ], batch size: 34, lr: 4.96e-03, grad_scale: 16.0 2023-05-15 19:08:30,776 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.114e+02 2.959e+02 3.524e+02 4.205e+02 8.301e+02, threshold=7.048e+02, percent-clipped=3.0 2023-05-15 19:08:43,538 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=117925.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 19:08:50,582 INFO [finetune.py:992] (0/2) Epoch 2, batch 5700, loss[loss=0.1426, simple_loss=0.2396, pruned_loss=0.02286, over 12155.00 frames. ], tot_loss[loss=0.1763, simple_loss=0.2635, pruned_loss=0.04452, over 2376634.90 frames. ], batch size: 36, lr: 4.96e-03, grad_scale: 16.0 2023-05-15 19:09:02,212 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=2.97 vs. limit=5.0 2023-05-15 19:09:26,522 INFO [finetune.py:992] (0/2) Epoch 2, batch 5750, loss[loss=0.159, simple_loss=0.2601, pruned_loss=0.02895, over 12149.00 frames. ], tot_loss[loss=0.1769, simple_loss=0.2641, pruned_loss=0.04488, over 2372273.18 frames. ], batch size: 34, lr: 4.96e-03, grad_scale: 16.0 2023-05-15 19:09:26,670 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=117985.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 19:09:30,282 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.3111, 4.9994, 5.2003, 5.2359, 5.0351, 5.2501, 5.0625, 2.9997], device='cuda:0'), covar=tensor([0.0071, 0.0049, 0.0048, 0.0041, 0.0038, 0.0054, 0.0056, 0.0511], device='cuda:0'), in_proj_covar=tensor([0.0063, 0.0072, 0.0075, 0.0069, 0.0056, 0.0085, 0.0075, 0.0087], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-15 19:09:37,563 INFO [checkpoint.py:75] (0/2) Saving checkpoint to pruned_transducer_stateless7/exp_giga_finetune/checkpoint-18000.pt 2023-05-15 19:09:45,346 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.023e+02 2.769e+02 3.355e+02 4.095e+02 8.466e+02, threshold=6.710e+02, percent-clipped=3.0 2023-05-15 19:09:57,971 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=4.35 vs. limit=5.0 2023-05-15 19:10:04,797 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=118033.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 19:10:06,161 INFO [finetune.py:992] (0/2) Epoch 2, batch 5800, loss[loss=0.1971, simple_loss=0.2867, pruned_loss=0.05376, over 10460.00 frames. ], tot_loss[loss=0.1776, simple_loss=0.265, pruned_loss=0.04516, over 2364930.40 frames. ], batch size: 68, lr: 4.96e-03, grad_scale: 16.0 2023-05-15 19:10:09,860 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=118040.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 19:10:41,889 INFO [finetune.py:992] (0/2) Epoch 2, batch 5850, loss[loss=0.1573, simple_loss=0.243, pruned_loss=0.03582, over 12254.00 frames. ], tot_loss[loss=0.1774, simple_loss=0.2645, pruned_loss=0.04514, over 2360398.18 frames. ], batch size: 32, lr: 4.96e-03, grad_scale: 16.0 2023-05-15 19:10:54,031 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=118101.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 19:10:58,041 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.016e+02 2.830e+02 3.539e+02 4.021e+02 7.766e+02, threshold=7.079e+02, percent-clipped=1.0 2023-05-15 19:11:15,240 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.1854, 4.3715, 4.2439, 4.5129, 2.9467, 4.3312, 2.9845, 4.2448], device='cuda:0'), covar=tensor([0.1430, 0.0489, 0.0604, 0.0493, 0.0984, 0.0424, 0.1439, 0.0846], device='cuda:0'), in_proj_covar=tensor([0.0225, 0.0264, 0.0296, 0.0357, 0.0239, 0.0240, 0.0257, 0.0389], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-15 19:11:17,736 INFO [finetune.py:992] (0/2) Epoch 2, batch 5900, loss[loss=0.1831, simple_loss=0.2729, pruned_loss=0.04663, over 10556.00 frames. ], tot_loss[loss=0.1761, simple_loss=0.2636, pruned_loss=0.04427, over 2369526.45 frames. ], batch size: 68, lr: 4.96e-03, grad_scale: 16.0 2023-05-15 19:11:41,841 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=118168.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 19:11:53,730 INFO [finetune.py:992] (0/2) Epoch 2, batch 5950, loss[loss=0.1691, simple_loss=0.2605, pruned_loss=0.03881, over 12336.00 frames. ], tot_loss[loss=0.1767, simple_loss=0.2639, pruned_loss=0.04479, over 2370692.42 frames. ], batch size: 31, lr: 4.96e-03, grad_scale: 16.0 2023-05-15 19:11:56,139 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.7778, 3.1530, 5.1004, 2.6916, 2.9873, 3.8394, 3.3674, 3.8472], device='cuda:0'), covar=tensor([0.0408, 0.1192, 0.0240, 0.1141, 0.1725, 0.1211, 0.1228, 0.1053], device='cuda:0'), in_proj_covar=tensor([0.0233, 0.0228, 0.0232, 0.0179, 0.0235, 0.0277, 0.0223, 0.0249], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-15 19:12:09,708 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.193e+02 2.846e+02 3.350e+02 3.911e+02 9.426e+02, threshold=6.700e+02, percent-clipped=2.0 2023-05-15 19:12:25,184 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=118229.0, num_to_drop=1, layers_to_drop={2} 2023-05-15 19:12:27,233 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.9637, 4.7748, 4.8614, 4.9191, 4.6616, 4.9813, 4.8734, 2.5760], device='cuda:0'), covar=tensor([0.0129, 0.0053, 0.0070, 0.0057, 0.0051, 0.0067, 0.0057, 0.0676], device='cuda:0'), in_proj_covar=tensor([0.0063, 0.0072, 0.0076, 0.0070, 0.0057, 0.0085, 0.0075, 0.0087], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-15 19:12:29,091 INFO [finetune.py:992] (0/2) Epoch 2, batch 6000, loss[loss=0.1722, simple_loss=0.2578, pruned_loss=0.0433, over 12195.00 frames. ], tot_loss[loss=0.1771, simple_loss=0.2644, pruned_loss=0.04492, over 2372163.31 frames. ], batch size: 35, lr: 4.96e-03, grad_scale: 16.0 2023-05-15 19:12:29,091 INFO [finetune.py:1017] (0/2) Computing validation loss 2023-05-15 19:12:47,237 INFO [finetune.py:1026] (0/2) Epoch 2, validation: loss=0.3365, simple_loss=0.4084, pruned_loss=0.1323, over 1020973.00 frames. 2023-05-15 19:12:47,237 INFO [finetune.py:1027] (0/2) Maximum memory allocated so far is 12745MB 2023-05-15 19:13:11,036 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=118269.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 19:13:15,286 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.6687, 5.0036, 3.2293, 2.8573, 4.2422, 2.9272, 4.2162, 3.5604], device='cuda:0'), covar=tensor([0.0586, 0.0373, 0.0941, 0.1281, 0.0216, 0.1073, 0.0386, 0.0716], device='cuda:0'), in_proj_covar=tensor([0.0187, 0.0244, 0.0172, 0.0194, 0.0136, 0.0178, 0.0191, 0.0172], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-15 19:13:22,961 INFO [finetune.py:992] (0/2) Epoch 2, batch 6050, loss[loss=0.1468, simple_loss=0.2272, pruned_loss=0.03325, over 12265.00 frames. ], tot_loss[loss=0.1766, simple_loss=0.2638, pruned_loss=0.04464, over 2378880.60 frames. ], batch size: 28, lr: 4.96e-03, grad_scale: 16.0 2023-05-15 19:13:26,042 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=118289.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 19:13:38,428 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.286e+02 2.970e+02 3.436e+02 4.188e+02 8.999e+02, threshold=6.871e+02, percent-clipped=4.0 2023-05-15 19:13:53,935 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=118329.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 19:13:54,762 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=118330.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 19:13:58,735 INFO [finetune.py:992] (0/2) Epoch 2, batch 6100, loss[loss=0.1396, simple_loss=0.2201, pruned_loss=0.02954, over 12280.00 frames. ], tot_loss[loss=0.176, simple_loss=0.2632, pruned_loss=0.04439, over 2379577.93 frames. ], batch size: 28, lr: 4.96e-03, grad_scale: 16.0 2023-05-15 19:14:09,605 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=118350.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 19:14:33,337 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=2.00 vs. limit=2.0 2023-05-15 19:14:34,231 INFO [finetune.py:992] (0/2) Epoch 2, batch 6150, loss[loss=0.1528, simple_loss=0.2408, pruned_loss=0.03241, over 12345.00 frames. ], tot_loss[loss=0.1763, simple_loss=0.2635, pruned_loss=0.04452, over 2374890.20 frames. ], batch size: 36, lr: 4.96e-03, grad_scale: 16.0 2023-05-15 19:14:38,134 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=118390.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 19:14:42,279 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=118396.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 19:14:50,159 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.033e+02 2.985e+02 3.360e+02 4.001e+02 1.245e+03, threshold=6.721e+02, percent-clipped=3.0 2023-05-15 19:15:10,798 INFO [finetune.py:992] (0/2) Epoch 2, batch 6200, loss[loss=0.1998, simple_loss=0.3, pruned_loss=0.0498, over 12093.00 frames. ], tot_loss[loss=0.1763, simple_loss=0.2634, pruned_loss=0.04461, over 2367185.10 frames. ], batch size: 42, lr: 4.96e-03, grad_scale: 16.0 2023-05-15 19:15:21,966 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2023-05-15 19:15:46,837 INFO [finetune.py:992] (0/2) Epoch 2, batch 6250, loss[loss=0.181, simple_loss=0.2734, pruned_loss=0.0443, over 12284.00 frames. ], tot_loss[loss=0.1772, simple_loss=0.2641, pruned_loss=0.04515, over 2363942.44 frames. ], batch size: 37, lr: 4.96e-03, grad_scale: 16.0 2023-05-15 19:15:50,022 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-05-15 19:15:50,746 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.15 vs. limit=2.0 2023-05-15 19:16:02,306 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.910e+02 2.962e+02 3.437e+02 4.041e+02 7.964e+02, threshold=6.874e+02, percent-clipped=1.0 2023-05-15 19:16:14,270 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=118524.0, num_to_drop=1, layers_to_drop={3} 2023-05-15 19:16:21,813 INFO [finetune.py:992] (0/2) Epoch 2, batch 6300, loss[loss=0.1493, simple_loss=0.2351, pruned_loss=0.03179, over 12176.00 frames. ], tot_loss[loss=0.1774, simple_loss=0.2641, pruned_loss=0.04533, over 2368916.97 frames. ], batch size: 29, lr: 4.96e-03, grad_scale: 16.0 2023-05-15 19:16:34,825 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.1421, 4.9285, 5.0948, 5.0833, 4.8894, 5.0412, 5.0159, 3.1443], device='cuda:0'), covar=tensor([0.0075, 0.0051, 0.0056, 0.0049, 0.0036, 0.0069, 0.0051, 0.0500], device='cuda:0'), in_proj_covar=tensor([0.0064, 0.0073, 0.0076, 0.0070, 0.0057, 0.0086, 0.0075, 0.0088], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-15 19:16:57,696 INFO [finetune.py:992] (0/2) Epoch 2, batch 6350, loss[loss=0.1812, simple_loss=0.271, pruned_loss=0.04567, over 12350.00 frames. ], tot_loss[loss=0.1775, simple_loss=0.2642, pruned_loss=0.0454, over 2369929.79 frames. ], batch size: 36, lr: 4.96e-03, grad_scale: 16.0 2023-05-15 19:17:13,586 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.016e+02 3.006e+02 3.545e+02 4.353e+02 1.068e+03, threshold=7.091e+02, percent-clipped=4.0 2023-05-15 19:17:26,363 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=118625.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 19:17:31,485 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=118631.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 19:17:34,136 INFO [finetune.py:992] (0/2) Epoch 2, batch 6400, loss[loss=0.1581, simple_loss=0.24, pruned_loss=0.03806, over 12355.00 frames. ], tot_loss[loss=0.1773, simple_loss=0.2643, pruned_loss=0.04511, over 2375158.70 frames. ], batch size: 30, lr: 4.96e-03, grad_scale: 16.0 2023-05-15 19:17:41,345 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=118645.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 19:17:46,762 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.51 vs. limit=2.0 2023-05-15 19:17:56,059 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 2023-05-15 19:18:09,969 INFO [finetune.py:992] (0/2) Epoch 2, batch 6450, loss[loss=0.1418, simple_loss=0.2273, pruned_loss=0.02814, over 12280.00 frames. ], tot_loss[loss=0.177, simple_loss=0.2643, pruned_loss=0.04484, over 2379101.22 frames. ], batch size: 28, lr: 4.96e-03, grad_scale: 16.0 2023-05-15 19:18:10,052 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=118685.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 19:18:14,246 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.5294, 2.8290, 3.5859, 4.5236, 3.8851, 4.5598, 3.8608, 3.0014], device='cuda:0'), covar=tensor([0.0025, 0.0306, 0.0131, 0.0027, 0.0112, 0.0056, 0.0113, 0.0340], device='cuda:0'), in_proj_covar=tensor([0.0078, 0.0118, 0.0099, 0.0072, 0.0100, 0.0107, 0.0084, 0.0131], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-15 19:18:14,959 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=118692.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 19:18:17,501 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=118696.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 19:18:25,826 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.882e+02 2.933e+02 3.395e+02 3.970e+02 8.348e+02, threshold=6.791e+02, percent-clipped=2.0 2023-05-15 19:18:37,501 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.9393, 5.8728, 5.7013, 5.1381, 5.0974, 5.8644, 5.5008, 5.2845], device='cuda:0'), covar=tensor([0.0616, 0.0868, 0.0576, 0.1387, 0.0596, 0.0626, 0.1265, 0.0859], device='cuda:0'), in_proj_covar=tensor([0.0571, 0.0504, 0.0463, 0.0573, 0.0375, 0.0656, 0.0707, 0.0518], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0004, 0.0004, 0.0003], device='cuda:0') 2023-05-15 19:18:45,813 INFO [finetune.py:992] (0/2) Epoch 2, batch 6500, loss[loss=0.2054, simple_loss=0.2942, pruned_loss=0.0583, over 10502.00 frames. ], tot_loss[loss=0.1777, simple_loss=0.2647, pruned_loss=0.04533, over 2369125.27 frames. ], batch size: 68, lr: 4.96e-03, grad_scale: 16.0 2023-05-15 19:18:52,320 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=118744.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 19:18:54,966 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.33 vs. limit=2.0 2023-05-15 19:19:21,862 INFO [finetune.py:992] (0/2) Epoch 2, batch 6550, loss[loss=0.2466, simple_loss=0.3166, pruned_loss=0.08834, over 8030.00 frames. ], tot_loss[loss=0.1778, simple_loss=0.2651, pruned_loss=0.04527, over 2364427.61 frames. ], batch size: 97, lr: 4.96e-03, grad_scale: 16.0 2023-05-15 19:19:27,281 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.97 vs. limit=2.0 2023-05-15 19:19:37,715 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.936e+02 3.114e+02 3.790e+02 4.411e+02 9.201e+02, threshold=7.579e+02, percent-clipped=1.0 2023-05-15 19:19:49,811 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=118824.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 19:19:57,498 INFO [finetune.py:992] (0/2) Epoch 2, batch 6600, loss[loss=0.1862, simple_loss=0.2718, pruned_loss=0.05033, over 12291.00 frames. ], tot_loss[loss=0.1775, simple_loss=0.2648, pruned_loss=0.0451, over 2367150.03 frames. ], batch size: 33, lr: 4.96e-03, grad_scale: 16.0 2023-05-15 19:20:24,659 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=118872.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 19:20:33,869 INFO [finetune.py:992] (0/2) Epoch 2, batch 6650, loss[loss=0.1647, simple_loss=0.2612, pruned_loss=0.03409, over 12089.00 frames. ], tot_loss[loss=0.1765, simple_loss=0.2639, pruned_loss=0.04452, over 2372331.22 frames. ], batch size: 32, lr: 4.96e-03, grad_scale: 16.0 2023-05-15 19:20:43,678 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=4.95 vs. limit=5.0 2023-05-15 19:20:49,519 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.013e+02 2.777e+02 3.194e+02 3.859e+02 9.715e+02, threshold=6.388e+02, percent-clipped=2.0 2023-05-15 19:21:03,200 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=118925.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 19:21:10,089 INFO [finetune.py:992] (0/2) Epoch 2, batch 6700, loss[loss=0.2156, simple_loss=0.2963, pruned_loss=0.06742, over 12352.00 frames. ], tot_loss[loss=0.1766, simple_loss=0.2644, pruned_loss=0.04443, over 2382178.40 frames. ], batch size: 36, lr: 4.96e-03, grad_scale: 16.0 2023-05-15 19:21:15,190 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.4286, 2.3719, 3.1780, 4.3169, 2.2236, 4.4407, 4.3153, 4.5492], device='cuda:0'), covar=tensor([0.0084, 0.1167, 0.0429, 0.0112, 0.1210, 0.0205, 0.0187, 0.0063], device='cuda:0'), in_proj_covar=tensor([0.0119, 0.0198, 0.0179, 0.0112, 0.0183, 0.0171, 0.0166, 0.0117], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-15 19:21:17,281 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=118945.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 19:21:37,278 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=118973.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 19:21:40,969 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.1507, 5.0701, 4.9329, 4.9831, 4.5757, 5.0991, 5.1170, 5.3742], device='cuda:0'), covar=tensor([0.0216, 0.0134, 0.0168, 0.0297, 0.0726, 0.0266, 0.0142, 0.0150], device='cuda:0'), in_proj_covar=tensor([0.0179, 0.0181, 0.0181, 0.0225, 0.0229, 0.0196, 0.0167, 0.0217], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0005, 0.0005, 0.0005, 0.0004, 0.0005], device='cuda:0') 2023-05-15 19:21:45,617 INFO [finetune.py:992] (0/2) Epoch 2, batch 6750, loss[loss=0.1808, simple_loss=0.2785, pruned_loss=0.0416, over 12280.00 frames. ], tot_loss[loss=0.1764, simple_loss=0.2642, pruned_loss=0.04431, over 2381506.61 frames. ], batch size: 37, lr: 4.96e-03, grad_scale: 16.0 2023-05-15 19:21:45,740 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=118985.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 19:21:47,120 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=118987.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 19:21:51,333 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=118993.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 19:22:02,213 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.416e+02 2.939e+02 3.508e+02 3.989e+02 7.684e+02, threshold=7.016e+02, percent-clipped=3.0 2023-05-15 19:22:04,632 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.9816, 2.1645, 3.4975, 2.8561, 3.3289, 3.0564, 2.1533, 3.4283], device='cuda:0'), covar=tensor([0.0118, 0.0371, 0.0136, 0.0255, 0.0140, 0.0164, 0.0371, 0.0126], device='cuda:0'), in_proj_covar=tensor([0.0156, 0.0191, 0.0168, 0.0172, 0.0191, 0.0146, 0.0182, 0.0161], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-15 19:22:06,305 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.91 vs. limit=2.0 2023-05-15 19:22:20,809 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=119033.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 19:22:22,147 INFO [finetune.py:992] (0/2) Epoch 2, batch 6800, loss[loss=0.2228, simple_loss=0.322, pruned_loss=0.06181, over 11558.00 frames. ], tot_loss[loss=0.1773, simple_loss=0.265, pruned_loss=0.04475, over 2380013.04 frames. ], batch size: 48, lr: 4.96e-03, grad_scale: 16.0 2023-05-15 19:22:50,960 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=119075.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 19:22:57,922 INFO [finetune.py:992] (0/2) Epoch 2, batch 6850, loss[loss=0.1992, simple_loss=0.2922, pruned_loss=0.05307, over 12190.00 frames. ], tot_loss[loss=0.1788, simple_loss=0.2662, pruned_loss=0.04575, over 2367443.65 frames. ], batch size: 35, lr: 4.96e-03, grad_scale: 16.0 2023-05-15 19:23:03,843 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=119093.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 19:23:12,244 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.1070, 2.4661, 3.6474, 3.1499, 3.4116, 3.3301, 2.5802, 3.6075], device='cuda:0'), covar=tensor([0.0100, 0.0265, 0.0111, 0.0180, 0.0140, 0.0116, 0.0268, 0.0103], device='cuda:0'), in_proj_covar=tensor([0.0156, 0.0190, 0.0167, 0.0171, 0.0191, 0.0146, 0.0181, 0.0161], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-15 19:23:13,462 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.191e+02 3.028e+02 3.541e+02 4.177e+02 1.306e+03, threshold=7.081e+02, percent-clipped=3.0 2023-05-15 19:23:32,696 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=3.06 vs. limit=5.0 2023-05-15 19:23:33,012 INFO [finetune.py:992] (0/2) Epoch 2, batch 6900, loss[loss=0.2112, simple_loss=0.3101, pruned_loss=0.05614, over 11805.00 frames. ], tot_loss[loss=0.1787, simple_loss=0.2661, pruned_loss=0.04562, over 2372484.24 frames. ], batch size: 44, lr: 4.96e-03, grad_scale: 16.0 2023-05-15 19:23:33,932 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=119136.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 19:23:47,256 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=119154.0, num_to_drop=1, layers_to_drop={2} 2023-05-15 19:23:57,520 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.37 vs. limit=2.0 2023-05-15 19:24:08,876 INFO [finetune.py:992] (0/2) Epoch 2, batch 6950, loss[loss=0.1598, simple_loss=0.2382, pruned_loss=0.04068, over 11366.00 frames. ], tot_loss[loss=0.1784, simple_loss=0.2661, pruned_loss=0.04536, over 2377085.53 frames. ], batch size: 25, lr: 4.95e-03, grad_scale: 16.0 2023-05-15 19:24:15,493 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=2.58 vs. limit=5.0 2023-05-15 19:24:25,183 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.934e+02 2.934e+02 3.415e+02 4.070e+02 9.985e+02, threshold=6.831e+02, percent-clipped=1.0 2023-05-15 19:24:38,775 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.0330, 3.8584, 3.8417, 4.3451, 2.9616, 3.9284, 2.7054, 3.9555], device='cuda:0'), covar=tensor([0.1629, 0.0706, 0.0894, 0.0533, 0.1041, 0.0541, 0.1600, 0.1176], device='cuda:0'), in_proj_covar=tensor([0.0222, 0.0260, 0.0292, 0.0351, 0.0236, 0.0237, 0.0253, 0.0386], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-15 19:24:44,693 INFO [finetune.py:992] (0/2) Epoch 2, batch 7000, loss[loss=0.1628, simple_loss=0.2459, pruned_loss=0.03989, over 12198.00 frames. ], tot_loss[loss=0.178, simple_loss=0.2656, pruned_loss=0.04518, over 2382881.17 frames. ], batch size: 31, lr: 4.95e-03, grad_scale: 16.0 2023-05-15 19:24:50,553 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.7535, 3.3705, 5.1120, 2.8123, 2.9334, 3.8660, 3.2277, 3.9321], device='cuda:0'), covar=tensor([0.0354, 0.1031, 0.0230, 0.1037, 0.1649, 0.1166, 0.1241, 0.0936], device='cuda:0'), in_proj_covar=tensor([0.0234, 0.0227, 0.0231, 0.0179, 0.0234, 0.0279, 0.0223, 0.0248], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:0') 2023-05-15 19:24:51,959 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.9160, 4.8844, 4.7352, 4.7618, 4.3908, 4.9281, 4.9373, 5.1095], device='cuda:0'), covar=tensor([0.0163, 0.0117, 0.0160, 0.0271, 0.0626, 0.0254, 0.0122, 0.0138], device='cuda:0'), in_proj_covar=tensor([0.0175, 0.0178, 0.0177, 0.0221, 0.0225, 0.0193, 0.0164, 0.0213], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0005, 0.0005, 0.0005, 0.0004, 0.0005], device='cuda:0') 2023-05-15 19:25:02,534 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.2564, 3.6750, 3.5974, 4.1758, 2.9171, 3.6035, 2.6515, 3.6597], device='cuda:0'), covar=tensor([0.1509, 0.0734, 0.0890, 0.0606, 0.1017, 0.0654, 0.1595, 0.1241], device='cuda:0'), in_proj_covar=tensor([0.0222, 0.0261, 0.0293, 0.0352, 0.0236, 0.0238, 0.0253, 0.0388], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-15 19:25:15,395 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.3026, 4.8125, 5.3047, 4.5763, 4.8997, 4.6986, 5.2867, 4.8975], device='cuda:0'), covar=tensor([0.0247, 0.0311, 0.0228, 0.0269, 0.0291, 0.0278, 0.0189, 0.0309], device='cuda:0'), in_proj_covar=tensor([0.0230, 0.0240, 0.0259, 0.0233, 0.0235, 0.0234, 0.0215, 0.0197], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-05-15 19:25:20,154 INFO [finetune.py:992] (0/2) Epoch 2, batch 7050, loss[loss=0.1923, simple_loss=0.2784, pruned_loss=0.05305, over 12196.00 frames. ], tot_loss[loss=0.1762, simple_loss=0.2638, pruned_loss=0.04428, over 2383251.07 frames. ], batch size: 31, lr: 4.95e-03, grad_scale: 8.0 2023-05-15 19:25:21,072 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.3909, 5.1799, 5.2980, 5.3735, 4.9724, 5.0168, 4.8378, 5.3351], device='cuda:0'), covar=tensor([0.0551, 0.0541, 0.0758, 0.0491, 0.1746, 0.1169, 0.0531, 0.0791], device='cuda:0'), in_proj_covar=tensor([0.0481, 0.0627, 0.0530, 0.0587, 0.0762, 0.0697, 0.0512, 0.0460], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0005, 0.0004, 0.0005, 0.0005, 0.0005, 0.0004, 0.0004], device='cuda:0') 2023-05-15 19:25:21,733 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=119287.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 19:25:37,229 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.747e+02 2.828e+02 3.255e+02 4.157e+02 7.416e+02, threshold=6.509e+02, percent-clipped=2.0 2023-05-15 19:25:42,524 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.4935, 4.3229, 4.1147, 4.6077, 3.2388, 4.2390, 2.9643, 4.2899], device='cuda:0'), covar=tensor([0.1377, 0.0510, 0.0782, 0.0532, 0.0934, 0.0463, 0.1388, 0.1164], device='cuda:0'), in_proj_covar=tensor([0.0222, 0.0260, 0.0292, 0.0351, 0.0236, 0.0238, 0.0252, 0.0386], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-15 19:25:54,052 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.92 vs. limit=2.0 2023-05-15 19:25:57,105 INFO [finetune.py:992] (0/2) Epoch 2, batch 7100, loss[loss=0.1627, simple_loss=0.2424, pruned_loss=0.04157, over 12335.00 frames. ], tot_loss[loss=0.1773, simple_loss=0.2652, pruned_loss=0.04474, over 2375573.51 frames. ], batch size: 30, lr: 4.95e-03, grad_scale: 8.0 2023-05-15 19:25:57,167 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=119335.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 19:26:32,154 INFO [finetune.py:992] (0/2) Epoch 2, batch 7150, loss[loss=0.1552, simple_loss=0.235, pruned_loss=0.03764, over 12278.00 frames. ], tot_loss[loss=0.1769, simple_loss=0.2645, pruned_loss=0.04465, over 2373693.28 frames. ], batch size: 28, lr: 4.95e-03, grad_scale: 8.0 2023-05-15 19:26:37,442 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.6053, 2.7647, 3.7820, 4.6455, 3.9881, 4.7061, 4.1732, 3.1350], device='cuda:0'), covar=tensor([0.0027, 0.0326, 0.0106, 0.0029, 0.0112, 0.0053, 0.0077, 0.0315], device='cuda:0'), in_proj_covar=tensor([0.0078, 0.0119, 0.0100, 0.0073, 0.0099, 0.0108, 0.0084, 0.0132], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-15 19:26:48,679 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.77 vs. limit=2.0 2023-05-15 19:26:48,755 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.175e+02 2.928e+02 3.393e+02 4.118e+02 7.925e+02, threshold=6.787e+02, percent-clipped=1.0 2023-05-15 19:26:58,216 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.9890, 2.9340, 4.4461, 2.2893, 2.6150, 3.3873, 2.9574, 3.5242], device='cuda:0'), covar=tensor([0.0530, 0.1169, 0.0317, 0.1317, 0.1817, 0.1324, 0.1330, 0.0989], device='cuda:0'), in_proj_covar=tensor([0.0235, 0.0228, 0.0233, 0.0181, 0.0235, 0.0280, 0.0225, 0.0250], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-15 19:27:04,948 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=119431.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 19:27:07,705 INFO [finetune.py:992] (0/2) Epoch 2, batch 7200, loss[loss=0.1642, simple_loss=0.2578, pruned_loss=0.03535, over 12162.00 frames. ], tot_loss[loss=0.1764, simple_loss=0.2639, pruned_loss=0.04446, over 2373871.74 frames. ], batch size: 29, lr: 4.95e-03, grad_scale: 8.0 2023-05-15 19:27:18,425 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=119449.0, num_to_drop=1, layers_to_drop={3} 2023-05-15 19:27:44,285 INFO [finetune.py:992] (0/2) Epoch 2, batch 7250, loss[loss=0.1543, simple_loss=0.2349, pruned_loss=0.03684, over 11780.00 frames. ], tot_loss[loss=0.1767, simple_loss=0.2637, pruned_loss=0.04485, over 2370592.76 frames. ], batch size: 26, lr: 4.95e-03, grad_scale: 8.0 2023-05-15 19:28:00,672 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.223e+02 2.955e+02 3.322e+02 4.173e+02 9.420e+02, threshold=6.644e+02, percent-clipped=4.0 2023-05-15 19:28:05,100 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=119514.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 19:28:19,663 INFO [finetune.py:992] (0/2) Epoch 2, batch 7300, loss[loss=0.1952, simple_loss=0.2844, pruned_loss=0.05298, over 12135.00 frames. ], tot_loss[loss=0.1763, simple_loss=0.2635, pruned_loss=0.04451, over 2375152.93 frames. ], batch size: 42, lr: 4.95e-03, grad_scale: 8.0 2023-05-15 19:28:38,544 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=4.84 vs. limit=5.0 2023-05-15 19:28:48,189 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=119575.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 19:28:51,000 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.7197, 4.5140, 4.4107, 4.7729, 4.5441, 4.6167, 4.4401, 2.0997], device='cuda:0'), covar=tensor([0.0201, 0.0089, 0.0153, 0.0088, 0.0076, 0.0174, 0.0132, 0.1109], device='cuda:0'), in_proj_covar=tensor([0.0063, 0.0072, 0.0075, 0.0069, 0.0057, 0.0085, 0.0074, 0.0087], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-15 19:28:55,089 INFO [finetune.py:992] (0/2) Epoch 2, batch 7350, loss[loss=0.1933, simple_loss=0.272, pruned_loss=0.0573, over 12003.00 frames. ], tot_loss[loss=0.1764, simple_loss=0.264, pruned_loss=0.04445, over 2379318.78 frames. ], batch size: 40, lr: 4.95e-03, grad_scale: 8.0 2023-05-15 19:29:12,254 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.633e+02 2.761e+02 3.326e+02 4.003e+02 6.449e+02, threshold=6.652e+02, percent-clipped=0.0 2023-05-15 19:29:31,773 INFO [finetune.py:992] (0/2) Epoch 2, batch 7400, loss[loss=0.1985, simple_loss=0.2873, pruned_loss=0.05479, over 12056.00 frames. ], tot_loss[loss=0.1756, simple_loss=0.2631, pruned_loss=0.04399, over 2386207.62 frames. ], batch size: 37, lr: 4.95e-03, grad_scale: 8.0 2023-05-15 19:29:46,089 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.83 vs. limit=2.0 2023-05-15 19:29:53,405 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=119665.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 19:30:07,449 INFO [finetune.py:992] (0/2) Epoch 2, batch 7450, loss[loss=0.1814, simple_loss=0.2649, pruned_loss=0.04898, over 11706.00 frames. ], tot_loss[loss=0.1754, simple_loss=0.2627, pruned_loss=0.04403, over 2384162.86 frames. ], batch size: 48, lr: 4.95e-03, grad_scale: 8.0 2023-05-15 19:30:23,540 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.184e+02 3.159e+02 3.777e+02 4.498e+02 9.403e+02, threshold=7.554e+02, percent-clipped=2.0 2023-05-15 19:30:36,236 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=119726.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 19:30:39,694 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=119731.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 19:30:42,510 INFO [finetune.py:992] (0/2) Epoch 2, batch 7500, loss[loss=0.2037, simple_loss=0.2883, pruned_loss=0.0595, over 12034.00 frames. ], tot_loss[loss=0.1767, simple_loss=0.2641, pruned_loss=0.04469, over 2379427.67 frames. ], batch size: 42, lr: 4.95e-03, grad_scale: 8.0 2023-05-15 19:30:53,336 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=119749.0, num_to_drop=1, layers_to_drop={2} 2023-05-15 19:30:53,996 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.0812, 4.7094, 4.9549, 5.0131, 4.7991, 4.9468, 4.8943, 2.8205], device='cuda:0'), covar=tensor([0.0100, 0.0063, 0.0065, 0.0054, 0.0046, 0.0077, 0.0060, 0.0597], device='cuda:0'), in_proj_covar=tensor([0.0064, 0.0072, 0.0075, 0.0069, 0.0057, 0.0085, 0.0074, 0.0087], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-15 19:31:06,054 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.87 vs. limit=2.0 2023-05-15 19:31:10,130 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.0881, 2.2439, 2.2304, 2.3083, 2.1391, 1.9443, 2.3817, 1.7560], device='cuda:0'), covar=tensor([0.0216, 0.0145, 0.0147, 0.0150, 0.0232, 0.0179, 0.0114, 0.0306], device='cuda:0'), in_proj_covar=tensor([0.0183, 0.0157, 0.0149, 0.0177, 0.0198, 0.0192, 0.0156, 0.0202], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-15 19:31:14,753 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=119779.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 19:31:18,896 INFO [finetune.py:992] (0/2) Epoch 2, batch 7550, loss[loss=0.1981, simple_loss=0.289, pruned_loss=0.05361, over 11990.00 frames. ], tot_loss[loss=0.1776, simple_loss=0.2644, pruned_loss=0.04543, over 2370599.81 frames. ], batch size: 40, lr: 4.95e-03, grad_scale: 8.0 2023-05-15 19:31:27,599 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=119797.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 19:31:35,741 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.735e+02 3.021e+02 3.531e+02 4.316e+02 1.223e+03, threshold=7.062e+02, percent-clipped=7.0 2023-05-15 19:31:44,291 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.6548, 3.7155, 3.2324, 3.2780, 2.8908, 2.8402, 3.7544, 2.2083], device='cuda:0'), covar=tensor([0.0314, 0.0113, 0.0163, 0.0155, 0.0332, 0.0329, 0.0104, 0.0464], device='cuda:0'), in_proj_covar=tensor([0.0184, 0.0158, 0.0149, 0.0178, 0.0199, 0.0192, 0.0158, 0.0203], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-15 19:31:54,855 INFO [finetune.py:992] (0/2) Epoch 2, batch 7600, loss[loss=0.1815, simple_loss=0.2639, pruned_loss=0.04951, over 12304.00 frames. ], tot_loss[loss=0.1782, simple_loss=0.2647, pruned_loss=0.04585, over 2367242.90 frames. ], batch size: 34, lr: 4.95e-03, grad_scale: 8.0 2023-05-15 19:32:19,695 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=119870.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 19:32:30,989 INFO [finetune.py:992] (0/2) Epoch 2, batch 7650, loss[loss=0.1603, simple_loss=0.2573, pruned_loss=0.03159, over 12047.00 frames. ], tot_loss[loss=0.1785, simple_loss=0.2651, pruned_loss=0.046, over 2367595.38 frames. ], batch size: 40, lr: 4.95e-03, grad_scale: 8.0 2023-05-15 19:32:47,881 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.981e+02 2.810e+02 3.505e+02 4.378e+02 1.736e+03, threshold=7.010e+02, percent-clipped=5.0 2023-05-15 19:33:07,245 INFO [finetune.py:992] (0/2) Epoch 2, batch 7700, loss[loss=0.1804, simple_loss=0.2653, pruned_loss=0.04773, over 12177.00 frames. ], tot_loss[loss=0.1771, simple_loss=0.2639, pruned_loss=0.04513, over 2373677.64 frames. ], batch size: 31, lr: 4.95e-03, grad_scale: 8.0 2023-05-15 19:33:16,026 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.2221, 4.5836, 2.6803, 2.3437, 3.9472, 2.3239, 3.9033, 2.9827], device='cuda:0'), covar=tensor([0.0615, 0.0468, 0.1056, 0.1520, 0.0224, 0.1361, 0.0420, 0.0785], device='cuda:0'), in_proj_covar=tensor([0.0187, 0.0247, 0.0173, 0.0196, 0.0138, 0.0178, 0.0191, 0.0172], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-15 19:33:42,685 INFO [finetune.py:992] (0/2) Epoch 2, batch 7750, loss[loss=0.1541, simple_loss=0.2428, pruned_loss=0.03267, over 12124.00 frames. ], tot_loss[loss=0.1769, simple_loss=0.2644, pruned_loss=0.04473, over 2376337.53 frames. ], batch size: 30, lr: 4.95e-03, grad_scale: 8.0 2023-05-15 19:33:53,703 INFO [checkpoint.py:75] (0/2) Saving checkpoint to pruned_transducer_stateless7/exp_giga_finetune/checkpoint-20000.pt 2023-05-15 19:34:02,462 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.783e+02 2.817e+02 3.479e+02 4.094e+02 1.236e+03, threshold=6.958e+02, percent-clipped=2.0 2023-05-15 19:34:11,846 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=120021.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 19:34:22,375 INFO [finetune.py:992] (0/2) Epoch 2, batch 7800, loss[loss=0.1509, simple_loss=0.2325, pruned_loss=0.03467, over 11761.00 frames. ], tot_loss[loss=0.1766, simple_loss=0.2642, pruned_loss=0.04451, over 2381043.48 frames. ], batch size: 26, lr: 4.95e-03, grad_scale: 8.0 2023-05-15 19:34:58,425 INFO [finetune.py:992] (0/2) Epoch 2, batch 7850, loss[loss=0.1696, simple_loss=0.2596, pruned_loss=0.03977, over 12105.00 frames. ], tot_loss[loss=0.1771, simple_loss=0.2646, pruned_loss=0.0448, over 2378303.67 frames. ], batch size: 38, lr: 4.95e-03, grad_scale: 8.0 2023-05-15 19:35:00,971 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=3.39 vs. limit=5.0 2023-05-15 19:35:05,046 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.2456, 5.1842, 5.0519, 5.1629, 4.6978, 5.1798, 5.2339, 5.4082], device='cuda:0'), covar=tensor([0.0169, 0.0132, 0.0164, 0.0239, 0.0658, 0.0243, 0.0135, 0.0141], device='cuda:0'), in_proj_covar=tensor([0.0177, 0.0180, 0.0180, 0.0224, 0.0227, 0.0196, 0.0165, 0.0215], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0005, 0.0005, 0.0005, 0.0004, 0.0005], device='cuda:0') 2023-05-15 19:35:14,932 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.797e+02 2.826e+02 3.357e+02 4.136e+02 8.252e+02, threshold=6.714e+02, percent-clipped=1.0 2023-05-15 19:35:33,897 INFO [finetune.py:992] (0/2) Epoch 2, batch 7900, loss[loss=0.1896, simple_loss=0.2793, pruned_loss=0.04997, over 12290.00 frames. ], tot_loss[loss=0.176, simple_loss=0.2633, pruned_loss=0.04431, over 2378070.15 frames. ], batch size: 34, lr: 4.95e-03, grad_scale: 8.0 2023-05-15 19:35:58,867 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=120170.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 19:36:02,576 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.4247, 4.6981, 4.2277, 5.0982, 4.7554, 2.8879, 4.4083, 3.1755], device='cuda:0'), covar=tensor([0.0702, 0.0683, 0.1221, 0.0330, 0.0907, 0.1519, 0.0862, 0.2935], device='cuda:0'), in_proj_covar=tensor([0.0291, 0.0368, 0.0344, 0.0255, 0.0354, 0.0257, 0.0326, 0.0355], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-15 19:36:10,282 INFO [finetune.py:992] (0/2) Epoch 2, batch 7950, loss[loss=0.1585, simple_loss=0.251, pruned_loss=0.03294, over 12081.00 frames. ], tot_loss[loss=0.1759, simple_loss=0.2636, pruned_loss=0.0441, over 2377537.23 frames. ], batch size: 32, lr: 4.95e-03, grad_scale: 8.0 2023-05-15 19:36:21,289 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.8687, 2.8845, 4.8202, 4.9782, 3.0799, 2.8161, 2.9437, 2.1150], device='cuda:0'), covar=tensor([0.1201, 0.2583, 0.0335, 0.0308, 0.1022, 0.1812, 0.2252, 0.3530], device='cuda:0'), in_proj_covar=tensor([0.0271, 0.0365, 0.0263, 0.0287, 0.0248, 0.0274, 0.0347, 0.0349], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-15 19:36:27,563 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.852e+02 2.839e+02 3.382e+02 4.302e+02 1.919e+03, threshold=6.765e+02, percent-clipped=5.0 2023-05-15 19:36:34,935 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=120218.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 19:36:46,909 INFO [finetune.py:992] (0/2) Epoch 2, batch 8000, loss[loss=0.227, simple_loss=0.3076, pruned_loss=0.07314, over 11785.00 frames. ], tot_loss[loss=0.1776, simple_loss=0.2648, pruned_loss=0.04522, over 2369052.89 frames. ], batch size: 44, lr: 4.95e-03, grad_scale: 8.0 2023-05-15 19:37:22,464 INFO [finetune.py:992] (0/2) Epoch 2, batch 8050, loss[loss=0.198, simple_loss=0.2732, pruned_loss=0.0614, over 12311.00 frames. ], tot_loss[loss=0.1767, simple_loss=0.264, pruned_loss=0.04467, over 2381059.90 frames. ], batch size: 33, lr: 4.95e-03, grad_scale: 8.0 2023-05-15 19:37:31,987 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=4.52 vs. limit=5.0 2023-05-15 19:37:38,857 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.033e+02 2.935e+02 3.464e+02 4.476e+02 1.396e+03, threshold=6.927e+02, percent-clipped=8.0 2023-05-15 19:37:48,287 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=120321.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 19:37:59,593 INFO [finetune.py:992] (0/2) Epoch 2, batch 8100, loss[loss=0.1844, simple_loss=0.2777, pruned_loss=0.04554, over 12060.00 frames. ], tot_loss[loss=0.1766, simple_loss=0.2638, pruned_loss=0.04476, over 2377791.09 frames. ], batch size: 37, lr: 4.95e-03, grad_scale: 8.0 2023-05-15 19:38:23,980 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=120369.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 19:38:35,486 INFO [finetune.py:992] (0/2) Epoch 2, batch 8150, loss[loss=0.1577, simple_loss=0.2443, pruned_loss=0.03552, over 12351.00 frames. ], tot_loss[loss=0.1779, simple_loss=0.2647, pruned_loss=0.0456, over 2368666.23 frames. ], batch size: 31, lr: 4.95e-03, grad_scale: 8.0 2023-05-15 19:38:51,830 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.021e+02 2.925e+02 3.403e+02 4.122e+02 5.583e+02, threshold=6.806e+02, percent-clipped=0.0 2023-05-15 19:39:10,922 INFO [finetune.py:992] (0/2) Epoch 2, batch 8200, loss[loss=0.1777, simple_loss=0.2693, pruned_loss=0.04302, over 12320.00 frames. ], tot_loss[loss=0.1775, simple_loss=0.2644, pruned_loss=0.04532, over 2372751.50 frames. ], batch size: 34, lr: 4.95e-03, grad_scale: 8.0 2023-05-15 19:39:24,577 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.4961, 4.8868, 3.0769, 2.7031, 4.1183, 2.5058, 4.1260, 3.4088], device='cuda:0'), covar=tensor([0.0510, 0.0394, 0.0947, 0.1309, 0.0246, 0.1311, 0.0428, 0.0682], device='cuda:0'), in_proj_covar=tensor([0.0189, 0.0251, 0.0176, 0.0198, 0.0141, 0.0181, 0.0194, 0.0174], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-15 19:39:32,421 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.4488, 2.5510, 3.2293, 4.3709, 2.2021, 4.4537, 4.3228, 4.6717], device='cuda:0'), covar=tensor([0.0110, 0.0966, 0.0413, 0.0099, 0.1073, 0.0159, 0.0169, 0.0065], device='cuda:0'), in_proj_covar=tensor([0.0121, 0.0203, 0.0185, 0.0114, 0.0186, 0.0176, 0.0171, 0.0119], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-15 19:39:47,517 INFO [finetune.py:992] (0/2) Epoch 2, batch 8250, loss[loss=0.1878, simple_loss=0.2783, pruned_loss=0.04865, over 11822.00 frames. ], tot_loss[loss=0.1765, simple_loss=0.2636, pruned_loss=0.04469, over 2381744.07 frames. ], batch size: 44, lr: 4.95e-03, grad_scale: 8.0 2023-05-15 19:40:03,631 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.102e+02 3.007e+02 3.486e+02 4.183e+02 6.928e+02, threshold=6.971e+02, percent-clipped=1.0 2023-05-15 19:40:22,810 INFO [finetune.py:992] (0/2) Epoch 2, batch 8300, loss[loss=0.1934, simple_loss=0.2826, pruned_loss=0.05212, over 12107.00 frames. ], tot_loss[loss=0.1768, simple_loss=0.264, pruned_loss=0.04481, over 2384375.17 frames. ], batch size: 38, lr: 4.95e-03, grad_scale: 8.0 2023-05-15 19:40:26,005 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.52 vs. limit=2.0 2023-05-15 19:40:27,214 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.4672, 4.7248, 2.7236, 2.6197, 4.0337, 2.5392, 3.9288, 3.1188], device='cuda:0'), covar=tensor([0.0533, 0.0385, 0.1055, 0.1367, 0.0268, 0.1179, 0.0441, 0.0729], device='cuda:0'), in_proj_covar=tensor([0.0190, 0.0253, 0.0177, 0.0199, 0.0142, 0.0182, 0.0196, 0.0175], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-15 19:40:57,806 INFO [finetune.py:992] (0/2) Epoch 2, batch 8350, loss[loss=0.1568, simple_loss=0.2407, pruned_loss=0.03646, over 12345.00 frames. ], tot_loss[loss=0.1763, simple_loss=0.2635, pruned_loss=0.04455, over 2388256.96 frames. ], batch size: 30, lr: 4.95e-03, grad_scale: 4.0 2023-05-15 19:41:11,948 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.50 vs. limit=2.0 2023-05-15 19:41:15,162 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.128e+02 3.117e+02 3.632e+02 4.219e+02 7.197e+02, threshold=7.264e+02, percent-clipped=1.0 2023-05-15 19:41:34,975 INFO [finetune.py:992] (0/2) Epoch 2, batch 8400, loss[loss=0.1792, simple_loss=0.2706, pruned_loss=0.04393, over 12115.00 frames. ], tot_loss[loss=0.1776, simple_loss=0.2648, pruned_loss=0.04516, over 2379234.72 frames. ], batch size: 33, lr: 4.95e-03, grad_scale: 8.0 2023-05-15 19:41:47,193 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.0199, 4.7752, 5.0088, 5.0272, 4.7393, 4.9308, 4.7842, 3.0068], device='cuda:0'), covar=tensor([0.0139, 0.0081, 0.0082, 0.0072, 0.0066, 0.0109, 0.0102, 0.0634], device='cuda:0'), in_proj_covar=tensor([0.0062, 0.0070, 0.0074, 0.0068, 0.0056, 0.0084, 0.0073, 0.0085], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-15 19:41:55,739 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.4273, 4.6796, 2.7280, 2.4854, 3.9476, 2.4590, 3.9527, 3.1120], device='cuda:0'), covar=tensor([0.0588, 0.0535, 0.1228, 0.1578, 0.0262, 0.1405, 0.0429, 0.0783], device='cuda:0'), in_proj_covar=tensor([0.0193, 0.0255, 0.0179, 0.0202, 0.0143, 0.0184, 0.0197, 0.0177], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-15 19:42:04,286 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=120676.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 19:42:10,685 INFO [finetune.py:992] (0/2) Epoch 2, batch 8450, loss[loss=0.1805, simple_loss=0.2723, pruned_loss=0.04432, over 11843.00 frames. ], tot_loss[loss=0.1787, simple_loss=0.2659, pruned_loss=0.0457, over 2369149.59 frames. ], batch size: 44, lr: 4.95e-03, grad_scale: 8.0 2023-05-15 19:42:24,136 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=4.79 vs. limit=5.0 2023-05-15 19:42:27,719 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.008e+02 2.780e+02 3.356e+02 3.939e+02 8.101e+02, threshold=6.712e+02, percent-clipped=2.0 2023-05-15 19:42:46,282 INFO [finetune.py:992] (0/2) Epoch 2, batch 8500, loss[loss=0.1724, simple_loss=0.2674, pruned_loss=0.03868, over 12159.00 frames. ], tot_loss[loss=0.1781, simple_loss=0.2654, pruned_loss=0.04545, over 2372241.95 frames. ], batch size: 36, lr: 4.95e-03, grad_scale: 8.0 2023-05-15 19:42:47,983 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=120737.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 19:43:23,297 INFO [finetune.py:992] (0/2) Epoch 2, batch 8550, loss[loss=0.162, simple_loss=0.2417, pruned_loss=0.04109, over 12130.00 frames. ], tot_loss[loss=0.1787, simple_loss=0.2658, pruned_loss=0.0458, over 2365198.53 frames. ], batch size: 30, lr: 4.95e-03, grad_scale: 4.0 2023-05-15 19:43:41,267 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.948e+02 2.862e+02 3.409e+02 3.902e+02 6.283e+02, threshold=6.819e+02, percent-clipped=0.0 2023-05-15 19:43:59,042 INFO [finetune.py:992] (0/2) Epoch 2, batch 8600, loss[loss=0.1537, simple_loss=0.24, pruned_loss=0.03369, over 12126.00 frames. ], tot_loss[loss=0.1791, simple_loss=0.2662, pruned_loss=0.04598, over 2362921.21 frames. ], batch size: 30, lr: 4.95e-03, grad_scale: 4.0 2023-05-15 19:44:34,719 INFO [finetune.py:992] (0/2) Epoch 2, batch 8650, loss[loss=0.1784, simple_loss=0.2684, pruned_loss=0.04423, over 12153.00 frames. ], tot_loss[loss=0.1788, simple_loss=0.2659, pruned_loss=0.04589, over 2366340.12 frames. ], batch size: 34, lr: 4.95e-03, grad_scale: 4.0 2023-05-15 19:44:53,806 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.163e+02 2.876e+02 3.374e+02 4.073e+02 7.321e+02, threshold=6.747e+02, percent-clipped=1.0 2023-05-15 19:44:55,584 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.4340, 4.6126, 3.9384, 5.0751, 4.7287, 3.0571, 4.3478, 3.0768], device='cuda:0'), covar=tensor([0.0788, 0.0811, 0.1525, 0.0352, 0.1087, 0.1526, 0.1090, 0.3132], device='cuda:0'), in_proj_covar=tensor([0.0292, 0.0367, 0.0345, 0.0257, 0.0353, 0.0257, 0.0327, 0.0356], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-15 19:45:11,516 INFO [finetune.py:992] (0/2) Epoch 2, batch 8700, loss[loss=0.1564, simple_loss=0.2411, pruned_loss=0.03584, over 12181.00 frames. ], tot_loss[loss=0.1776, simple_loss=0.2648, pruned_loss=0.04519, over 2376292.11 frames. ], batch size: 31, lr: 4.95e-03, grad_scale: 4.0 2023-05-15 19:45:29,963 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.4114, 3.2657, 4.7923, 2.6013, 2.7064, 3.6055, 3.0391, 3.6582], device='cuda:0'), covar=tensor([0.0430, 0.1032, 0.0325, 0.1097, 0.1796, 0.1388, 0.1349, 0.1217], device='cuda:0'), in_proj_covar=tensor([0.0231, 0.0226, 0.0232, 0.0177, 0.0233, 0.0276, 0.0220, 0.0248], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:0') 2023-05-15 19:45:39,049 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.7318, 2.9739, 4.7739, 4.9853, 3.0227, 2.7468, 3.0347, 2.2280], device='cuda:0'), covar=tensor([0.1396, 0.2790, 0.0369, 0.0326, 0.1128, 0.1958, 0.2562, 0.3764], device='cuda:0'), in_proj_covar=tensor([0.0273, 0.0368, 0.0266, 0.0289, 0.0250, 0.0276, 0.0348, 0.0350], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-15 19:45:46,654 INFO [finetune.py:992] (0/2) Epoch 2, batch 8750, loss[loss=0.1904, simple_loss=0.2836, pruned_loss=0.04857, over 11819.00 frames. ], tot_loss[loss=0.1768, simple_loss=0.2641, pruned_loss=0.04478, over 2376380.31 frames. ], batch size: 44, lr: 4.95e-03, grad_scale: 4.0 2023-05-15 19:45:59,513 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=121003.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 19:46:04,181 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.022e+02 2.962e+02 3.419e+02 4.419e+02 8.057e+02, threshold=6.838e+02, percent-clipped=2.0 2023-05-15 19:46:19,666 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=121032.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 19:46:21,779 INFO [finetune.py:992] (0/2) Epoch 2, batch 8800, loss[loss=0.1557, simple_loss=0.2297, pruned_loss=0.0408, over 12165.00 frames. ], tot_loss[loss=0.1786, simple_loss=0.2657, pruned_loss=0.04571, over 2365579.99 frames. ], batch size: 29, lr: 4.95e-03, grad_scale: 8.0 2023-05-15 19:46:32,543 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.41 vs. limit=2.0 2023-05-15 19:46:32,882 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.3434, 4.8675, 5.3078, 4.6341, 4.9605, 4.6461, 5.3220, 4.9310], device='cuda:0'), covar=tensor([0.0213, 0.0306, 0.0236, 0.0222, 0.0315, 0.0296, 0.0174, 0.0337], device='cuda:0'), in_proj_covar=tensor([0.0225, 0.0233, 0.0252, 0.0227, 0.0231, 0.0228, 0.0208, 0.0195], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-05-15 19:46:43,709 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=121064.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 19:46:52,700 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.65 vs. limit=2.0 2023-05-15 19:46:58,477 INFO [finetune.py:992] (0/2) Epoch 2, batch 8850, loss[loss=0.1967, simple_loss=0.2809, pruned_loss=0.05622, over 11620.00 frames. ], tot_loss[loss=0.1783, simple_loss=0.2654, pruned_loss=0.04558, over 2369005.44 frames. ], batch size: 48, lr: 4.95e-03, grad_scale: 8.0 2023-05-15 19:47:16,519 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.968e+02 3.010e+02 3.613e+02 4.361e+02 8.894e+02, threshold=7.226e+02, percent-clipped=1.0 2023-05-15 19:47:34,449 INFO [finetune.py:992] (0/2) Epoch 2, batch 8900, loss[loss=0.1776, simple_loss=0.2708, pruned_loss=0.04223, over 12292.00 frames. ], tot_loss[loss=0.1784, simple_loss=0.2658, pruned_loss=0.04554, over 2373467.14 frames. ], batch size: 34, lr: 4.95e-03, grad_scale: 8.0 2023-05-15 19:47:55,166 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=121164.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 19:47:55,211 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.3530, 3.1613, 3.2299, 3.5285, 2.6859, 3.2296, 2.5685, 3.0717], device='cuda:0'), covar=tensor([0.1266, 0.0687, 0.0762, 0.0587, 0.0840, 0.0601, 0.1307, 0.1044], device='cuda:0'), in_proj_covar=tensor([0.0225, 0.0262, 0.0295, 0.0352, 0.0237, 0.0239, 0.0256, 0.0390], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-15 19:47:58,256 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.55 vs. limit=2.0 2023-05-15 19:48:05,409 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.36 vs. limit=2.0 2023-05-15 19:48:10,523 INFO [finetune.py:992] (0/2) Epoch 2, batch 8950, loss[loss=0.1713, simple_loss=0.2652, pruned_loss=0.03867, over 12150.00 frames. ], tot_loss[loss=0.1786, simple_loss=0.2662, pruned_loss=0.04554, over 2370628.07 frames. ], batch size: 39, lr: 4.95e-03, grad_scale: 8.0 2023-05-15 19:48:12,741 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.0568, 5.7369, 5.3684, 5.3402, 5.8845, 5.1842, 5.4658, 5.4987], device='cuda:0'), covar=tensor([0.1297, 0.0980, 0.0866, 0.2058, 0.0959, 0.2129, 0.1498, 0.1015], device='cuda:0'), in_proj_covar=tensor([0.0335, 0.0457, 0.0362, 0.0419, 0.0437, 0.0410, 0.0376, 0.0369], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-15 19:48:29,068 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.875e+02 2.756e+02 3.209e+02 4.008e+02 1.639e+03, threshold=6.418e+02, percent-clipped=3.0 2023-05-15 19:48:32,063 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.5278, 2.6544, 3.8063, 4.5478, 4.0991, 4.4896, 4.0336, 3.2301], device='cuda:0'), covar=tensor([0.0024, 0.0306, 0.0086, 0.0026, 0.0077, 0.0060, 0.0061, 0.0283], device='cuda:0'), in_proj_covar=tensor([0.0077, 0.0115, 0.0098, 0.0072, 0.0097, 0.0107, 0.0083, 0.0129], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-15 19:48:39,919 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=121225.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 19:48:46,849 INFO [finetune.py:992] (0/2) Epoch 2, batch 9000, loss[loss=0.1559, simple_loss=0.24, pruned_loss=0.03594, over 12197.00 frames. ], tot_loss[loss=0.1786, simple_loss=0.2657, pruned_loss=0.04572, over 2367726.38 frames. ], batch size: 29, lr: 4.95e-03, grad_scale: 8.0 2023-05-15 19:48:46,850 INFO [finetune.py:1017] (0/2) Computing validation loss 2023-05-15 19:48:59,812 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.7194, 3.8755, 3.5238, 4.1822, 3.8311, 2.6542, 3.6954, 2.8214], device='cuda:0'), covar=tensor([0.0765, 0.0944, 0.1290, 0.0426, 0.1067, 0.1541, 0.0885, 0.3271], device='cuda:0'), in_proj_covar=tensor([0.0290, 0.0364, 0.0342, 0.0253, 0.0350, 0.0255, 0.0325, 0.0353], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-15 19:49:05,158 INFO [finetune.py:1026] (0/2) Epoch 2, validation: loss=0.3439, simple_loss=0.4128, pruned_loss=0.1375, over 1020973.00 frames. 2023-05-15 19:49:05,159 INFO [finetune.py:1027] (0/2) Maximum memory allocated so far is 12745MB 2023-05-15 19:49:12,146 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.2947, 6.0791, 5.6377, 5.6891, 6.1489, 5.3802, 5.8336, 5.6796], device='cuda:0'), covar=tensor([0.1323, 0.0819, 0.0768, 0.1849, 0.0873, 0.2023, 0.1295, 0.1135], device='cuda:0'), in_proj_covar=tensor([0.0333, 0.0453, 0.0359, 0.0416, 0.0433, 0.0407, 0.0373, 0.0367], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-15 19:49:17,887 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.4927, 3.6166, 3.3744, 3.7627, 3.5949, 2.6731, 3.4244, 2.8717], device='cuda:0'), covar=tensor([0.0725, 0.0885, 0.1124, 0.0484, 0.0900, 0.1221, 0.0896, 0.2262], device='cuda:0'), in_proj_covar=tensor([0.0290, 0.0363, 0.0342, 0.0253, 0.0350, 0.0255, 0.0325, 0.0352], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-15 19:49:40,797 INFO [finetune.py:992] (0/2) Epoch 2, batch 9050, loss[loss=0.1829, simple_loss=0.2877, pruned_loss=0.03904, over 12164.00 frames. ], tot_loss[loss=0.1793, simple_loss=0.2663, pruned_loss=0.04615, over 2365292.15 frames. ], batch size: 36, lr: 4.94e-03, grad_scale: 8.0 2023-05-15 19:49:50,012 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.9215, 4.8698, 4.6971, 4.7638, 4.3631, 4.8467, 4.9091, 5.0971], device='cuda:0'), covar=tensor([0.0265, 0.0149, 0.0201, 0.0279, 0.0755, 0.0258, 0.0145, 0.0179], device='cuda:0'), in_proj_covar=tensor([0.0178, 0.0181, 0.0181, 0.0226, 0.0228, 0.0199, 0.0167, 0.0216], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0005, 0.0005, 0.0005, 0.0004, 0.0005], device='cuda:0') 2023-05-15 19:49:58,849 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.083e+02 2.921e+02 3.503e+02 4.414e+02 9.528e+02, threshold=7.007e+02, percent-clipped=6.0 2023-05-15 19:49:59,825 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=121311.0, num_to_drop=1, layers_to_drop={0} 2023-05-15 19:50:14,653 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=121332.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 19:50:16,266 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.6000, 2.7817, 4.2982, 4.5996, 3.0226, 2.7210, 2.9555, 2.0628], device='cuda:0'), covar=tensor([0.1371, 0.2605, 0.0496, 0.0380, 0.1031, 0.1885, 0.2254, 0.3641], device='cuda:0'), in_proj_covar=tensor([0.0273, 0.0368, 0.0266, 0.0289, 0.0250, 0.0275, 0.0348, 0.0349], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-15 19:50:16,682 INFO [finetune.py:992] (0/2) Epoch 2, batch 9100, loss[loss=0.1978, simple_loss=0.2903, pruned_loss=0.05268, over 12027.00 frames. ], tot_loss[loss=0.1781, simple_loss=0.2651, pruned_loss=0.04554, over 2373228.38 frames. ], batch size: 42, lr: 4.94e-03, grad_scale: 4.0 2023-05-15 19:50:33,427 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=121359.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 19:50:42,873 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=121372.0, num_to_drop=1, layers_to_drop={3} 2023-05-15 19:50:48,304 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=121380.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 19:50:51,920 INFO [finetune.py:992] (0/2) Epoch 2, batch 9150, loss[loss=0.1572, simple_loss=0.244, pruned_loss=0.03523, over 12331.00 frames. ], tot_loss[loss=0.1782, simple_loss=0.2652, pruned_loss=0.04565, over 2375817.43 frames. ], batch size: 31, lr: 4.94e-03, grad_scale: 4.0 2023-05-15 19:51:07,977 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=121407.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 19:51:10,589 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.922e+02 3.099e+02 3.650e+02 4.725e+02 1.105e+03, threshold=7.299e+02, percent-clipped=6.0 2023-05-15 19:51:27,885 INFO [finetune.py:992] (0/2) Epoch 2, batch 9200, loss[loss=0.1553, simple_loss=0.2304, pruned_loss=0.04011, over 12354.00 frames. ], tot_loss[loss=0.178, simple_loss=0.2647, pruned_loss=0.04564, over 2376828.41 frames. ], batch size: 30, lr: 4.94e-03, grad_scale: 8.0 2023-05-15 19:51:52,010 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=121468.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 19:51:59,023 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.4269, 4.9829, 5.4113, 4.7395, 5.0827, 4.7717, 5.3968, 5.0440], device='cuda:0'), covar=tensor([0.0193, 0.0260, 0.0183, 0.0192, 0.0241, 0.0280, 0.0190, 0.0241], device='cuda:0'), in_proj_covar=tensor([0.0225, 0.0232, 0.0251, 0.0226, 0.0229, 0.0228, 0.0209, 0.0193], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-05-15 19:52:02,715 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.6806, 2.8041, 3.4230, 4.6658, 2.8808, 4.6800, 4.6266, 4.9033], device='cuda:0'), covar=tensor([0.0123, 0.0995, 0.0364, 0.0118, 0.0906, 0.0173, 0.0143, 0.0057], device='cuda:0'), in_proj_covar=tensor([0.0118, 0.0198, 0.0182, 0.0111, 0.0180, 0.0170, 0.0165, 0.0116], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-15 19:52:03,946 INFO [finetune.py:992] (0/2) Epoch 2, batch 9250, loss[loss=0.1615, simple_loss=0.2423, pruned_loss=0.0403, over 12289.00 frames. ], tot_loss[loss=0.1781, simple_loss=0.2647, pruned_loss=0.04574, over 2371797.43 frames. ], batch size: 33, lr: 4.94e-03, grad_scale: 8.0 2023-05-15 19:52:22,382 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.873e+02 2.914e+02 3.402e+02 4.168e+02 8.635e+02, threshold=6.805e+02, percent-clipped=2.0 2023-05-15 19:52:28,782 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=121520.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 19:52:39,428 INFO [finetune.py:992] (0/2) Epoch 2, batch 9300, loss[loss=0.2559, simple_loss=0.3209, pruned_loss=0.09548, over 8124.00 frames. ], tot_loss[loss=0.1782, simple_loss=0.2649, pruned_loss=0.04571, over 2367577.14 frames. ], batch size: 99, lr: 4.94e-03, grad_scale: 8.0 2023-05-15 19:52:48,972 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=4.32 vs. limit=5.0 2023-05-15 19:52:52,232 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.6281, 4.3011, 4.6450, 4.1502, 4.3721, 4.1465, 4.6054, 4.2338], device='cuda:0'), covar=tensor([0.0267, 0.0314, 0.0236, 0.0221, 0.0291, 0.0292, 0.0223, 0.0477], device='cuda:0'), in_proj_covar=tensor([0.0226, 0.0233, 0.0251, 0.0227, 0.0230, 0.0228, 0.0209, 0.0194], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-05-15 19:53:15,969 INFO [finetune.py:992] (0/2) Epoch 2, batch 9350, loss[loss=0.2011, simple_loss=0.2762, pruned_loss=0.06294, over 12249.00 frames. ], tot_loss[loss=0.1783, simple_loss=0.2655, pruned_loss=0.04553, over 2364274.48 frames. ], batch size: 32, lr: 4.94e-03, grad_scale: 8.0 2023-05-15 19:53:27,471 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([6.0410, 6.0099, 5.7553, 5.3298, 5.1823, 5.9314, 5.5011, 5.2554], device='cuda:0'), covar=tensor([0.0626, 0.0850, 0.0629, 0.1422, 0.0617, 0.0690, 0.1437, 0.0891], device='cuda:0'), in_proj_covar=tensor([0.0572, 0.0514, 0.0473, 0.0590, 0.0384, 0.0665, 0.0725, 0.0528], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0004, 0.0004, 0.0003], device='cuda:0') 2023-05-15 19:53:34,340 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.715e+02 2.845e+02 3.346e+02 4.203e+02 7.244e+02, threshold=6.692e+02, percent-clipped=1.0 2023-05-15 19:53:51,426 INFO [finetune.py:992] (0/2) Epoch 2, batch 9400, loss[loss=0.1895, simple_loss=0.2791, pruned_loss=0.04995, over 10784.00 frames. ], tot_loss[loss=0.1772, simple_loss=0.2645, pruned_loss=0.04494, over 2370222.79 frames. ], batch size: 68, lr: 4.94e-03, grad_scale: 8.0 2023-05-15 19:54:08,416 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=121659.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 19:54:13,995 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=121667.0, num_to_drop=1, layers_to_drop={2} 2023-05-15 19:54:14,105 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.4901, 2.6913, 3.9201, 4.6359, 4.0932, 4.5787, 4.2062, 3.2078], device='cuda:0'), covar=tensor([0.0038, 0.0345, 0.0090, 0.0028, 0.0100, 0.0064, 0.0068, 0.0310], device='cuda:0'), in_proj_covar=tensor([0.0077, 0.0115, 0.0097, 0.0072, 0.0097, 0.0107, 0.0083, 0.0128], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-15 19:54:22,019 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.8338, 2.8805, 4.4383, 4.5940, 2.9378, 2.7533, 2.9312, 2.0840], device='cuda:0'), covar=tensor([0.1188, 0.2518, 0.0464, 0.0370, 0.1049, 0.1750, 0.2257, 0.3471], device='cuda:0'), in_proj_covar=tensor([0.0274, 0.0370, 0.0268, 0.0291, 0.0252, 0.0277, 0.0350, 0.0350], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-15 19:54:26,750 INFO [finetune.py:992] (0/2) Epoch 2, batch 9450, loss[loss=0.1536, simple_loss=0.2342, pruned_loss=0.03651, over 12182.00 frames. ], tot_loss[loss=0.176, simple_loss=0.2637, pruned_loss=0.04412, over 2377357.11 frames. ], batch size: 31, lr: 4.94e-03, grad_scale: 8.0 2023-05-15 19:54:42,343 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=121707.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 19:54:45,053 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.011e+02 2.877e+02 3.491e+02 4.071e+02 9.031e+02, threshold=6.982e+02, percent-clipped=1.0 2023-05-15 19:55:03,144 INFO [finetune.py:992] (0/2) Epoch 2, batch 9500, loss[loss=0.214, simple_loss=0.3201, pruned_loss=0.05398, over 12370.00 frames. ], tot_loss[loss=0.176, simple_loss=0.2634, pruned_loss=0.04432, over 2371861.41 frames. ], batch size: 35, lr: 4.94e-03, grad_scale: 8.0 2023-05-15 19:55:23,190 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=121763.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 19:55:38,593 INFO [finetune.py:992] (0/2) Epoch 2, batch 9550, loss[loss=0.1604, simple_loss=0.2457, pruned_loss=0.03759, over 12356.00 frames. ], tot_loss[loss=0.1772, simple_loss=0.2646, pruned_loss=0.04497, over 2355702.57 frames. ], batch size: 30, lr: 4.94e-03, grad_scale: 8.0 2023-05-15 19:55:54,130 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.5451, 4.7955, 4.2457, 5.2469, 4.8222, 2.5454, 4.1806, 2.9975], device='cuda:0'), covar=tensor([0.0590, 0.0652, 0.1131, 0.0320, 0.0870, 0.1657, 0.1047, 0.2955], device='cuda:0'), in_proj_covar=tensor([0.0294, 0.0365, 0.0345, 0.0255, 0.0353, 0.0256, 0.0329, 0.0355], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-15 19:55:57,234 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.830e+02 2.787e+02 3.335e+02 3.979e+02 8.470e+02, threshold=6.670e+02, percent-clipped=2.0 2023-05-15 19:56:03,098 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.3388, 5.1026, 5.2338, 5.2964, 4.6848, 4.7878, 4.7993, 5.1295], device='cuda:0'), covar=tensor([0.0693, 0.0739, 0.0763, 0.0688, 0.2478, 0.1795, 0.0667, 0.1281], device='cuda:0'), in_proj_covar=tensor([0.0482, 0.0623, 0.0525, 0.0578, 0.0766, 0.0699, 0.0512, 0.0459], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0005, 0.0004, 0.0004, 0.0005, 0.0005, 0.0004, 0.0004], device='cuda:0') 2023-05-15 19:56:03,826 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=121820.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 19:56:04,458 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.0704, 5.7154, 5.4888, 5.3500, 5.8941, 5.2432, 5.3652, 5.3992], device='cuda:0'), covar=tensor([0.1424, 0.0950, 0.0765, 0.2064, 0.0897, 0.1966, 0.1860, 0.1018], device='cuda:0'), in_proj_covar=tensor([0.0331, 0.0452, 0.0359, 0.0414, 0.0431, 0.0407, 0.0374, 0.0361], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-15 19:56:14,065 INFO [finetune.py:992] (0/2) Epoch 2, batch 9600, loss[loss=0.1858, simple_loss=0.2773, pruned_loss=0.04714, over 12135.00 frames. ], tot_loss[loss=0.1757, simple_loss=0.2631, pruned_loss=0.04415, over 2368810.91 frames. ], batch size: 38, lr: 4.94e-03, grad_scale: 8.0 2023-05-15 19:56:37,784 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=121868.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 19:56:42,197 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.1724, 4.9301, 5.1469, 5.0216, 4.9826, 5.1406, 4.9731, 3.2823], device='cuda:0'), covar=tensor([0.0092, 0.0047, 0.0054, 0.0050, 0.0035, 0.0065, 0.0051, 0.0434], device='cuda:0'), in_proj_covar=tensor([0.0063, 0.0071, 0.0075, 0.0069, 0.0056, 0.0085, 0.0074, 0.0087], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-15 19:56:51,186 INFO [finetune.py:992] (0/2) Epoch 2, batch 9650, loss[loss=0.2497, simple_loss=0.3237, pruned_loss=0.08789, over 8034.00 frames. ], tot_loss[loss=0.1757, simple_loss=0.2632, pruned_loss=0.04407, over 2374039.65 frames. ], batch size: 98, lr: 4.94e-03, grad_scale: 8.0 2023-05-15 19:57:03,982 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.3969, 4.7782, 3.0096, 2.7618, 4.0060, 2.4529, 4.0153, 3.3256], device='cuda:0'), covar=tensor([0.0648, 0.0449, 0.0980, 0.1409, 0.0287, 0.1378, 0.0416, 0.0762], device='cuda:0'), in_proj_covar=tensor([0.0192, 0.0252, 0.0175, 0.0198, 0.0141, 0.0181, 0.0194, 0.0173], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-15 19:57:06,091 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.6872, 4.4862, 4.5193, 4.5835, 4.4767, 4.6212, 4.5496, 2.6190], device='cuda:0'), covar=tensor([0.0108, 0.0065, 0.0100, 0.0072, 0.0054, 0.0094, 0.0095, 0.0704], device='cuda:0'), in_proj_covar=tensor([0.0064, 0.0071, 0.0075, 0.0069, 0.0056, 0.0085, 0.0074, 0.0087], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-15 19:57:09,351 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.125e+02 3.093e+02 3.583e+02 4.110e+02 1.152e+03, threshold=7.167e+02, percent-clipped=4.0 2023-05-15 19:57:26,313 INFO [finetune.py:992] (0/2) Epoch 2, batch 9700, loss[loss=0.1677, simple_loss=0.2645, pruned_loss=0.0355, over 12352.00 frames. ], tot_loss[loss=0.1767, simple_loss=0.2647, pruned_loss=0.04437, over 2374150.11 frames. ], batch size: 35, lr: 4.94e-03, grad_scale: 8.0 2023-05-15 19:57:37,156 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=121950.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 19:57:49,235 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=121967.0, num_to_drop=1, layers_to_drop={2} 2023-05-15 19:57:50,658 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.3690, 2.3052, 3.6381, 4.4144, 3.8446, 4.3489, 3.8736, 2.9524], device='cuda:0'), covar=tensor([0.0027, 0.0375, 0.0101, 0.0032, 0.0107, 0.0059, 0.0073, 0.0330], device='cuda:0'), in_proj_covar=tensor([0.0078, 0.0117, 0.0098, 0.0073, 0.0098, 0.0108, 0.0084, 0.0130], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-15 19:58:00,195 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.39 vs. limit=2.0 2023-05-15 19:58:01,735 INFO [finetune.py:992] (0/2) Epoch 2, batch 9750, loss[loss=0.1702, simple_loss=0.2576, pruned_loss=0.04138, over 12336.00 frames. ], tot_loss[loss=0.1764, simple_loss=0.2641, pruned_loss=0.04433, over 2381128.17 frames. ], batch size: 31, lr: 4.94e-03, grad_scale: 8.0 2023-05-15 19:58:12,517 INFO [checkpoint.py:75] (0/2) Saving checkpoint to pruned_transducer_stateless7/exp_giga_finetune/checkpoint-22000.pt 2023-05-15 19:58:19,896 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.2110, 2.4740, 3.8059, 3.1914, 3.6316, 3.2770, 2.5162, 3.6506], device='cuda:0'), covar=tensor([0.0102, 0.0288, 0.0092, 0.0175, 0.0093, 0.0145, 0.0306, 0.0094], device='cuda:0'), in_proj_covar=tensor([0.0157, 0.0192, 0.0168, 0.0173, 0.0191, 0.0149, 0.0181, 0.0162], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-15 19:58:23,192 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.892e+02 2.816e+02 3.449e+02 4.237e+02 2.383e+03, threshold=6.898e+02, percent-clipped=6.0 2023-05-15 19:58:23,451 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=122011.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 19:58:26,901 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=122015.0, num_to_drop=1, layers_to_drop={1} 2023-05-15 19:58:33,283 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.3927, 3.4887, 3.2077, 3.0822, 2.7225, 2.4907, 3.4494, 2.0904], device='cuda:0'), covar=tensor([0.0326, 0.0108, 0.0130, 0.0154, 0.0376, 0.0351, 0.0096, 0.0427], device='cuda:0'), in_proj_covar=tensor([0.0185, 0.0156, 0.0147, 0.0178, 0.0198, 0.0191, 0.0158, 0.0198], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-15 19:58:41,412 INFO [finetune.py:992] (0/2) Epoch 2, batch 9800, loss[loss=0.1885, simple_loss=0.2751, pruned_loss=0.05093, over 12105.00 frames. ], tot_loss[loss=0.177, simple_loss=0.2644, pruned_loss=0.0448, over 2383498.29 frames. ], batch size: 38, lr: 4.94e-03, grad_scale: 8.0 2023-05-15 19:58:42,483 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=2.60 vs. limit=5.0 2023-05-15 19:59:01,449 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=122063.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 19:59:16,884 INFO [finetune.py:992] (0/2) Epoch 2, batch 9850, loss[loss=0.1473, simple_loss=0.2329, pruned_loss=0.03079, over 12021.00 frames. ], tot_loss[loss=0.1764, simple_loss=0.2638, pruned_loss=0.04449, over 2383067.90 frames. ], batch size: 28, lr: 4.94e-03, grad_scale: 8.0 2023-05-15 19:59:35,149 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.931e+02 2.944e+02 3.381e+02 4.045e+02 6.935e+02, threshold=6.762e+02, percent-clipped=1.0 2023-05-15 19:59:35,236 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=122111.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 19:59:39,092 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.6658, 2.8038, 4.4189, 4.6317, 3.0642, 2.5777, 2.7880, 2.0147], device='cuda:0'), covar=tensor([0.1310, 0.2682, 0.0460, 0.0373, 0.0973, 0.1917, 0.2450, 0.3589], device='cuda:0'), in_proj_covar=tensor([0.0270, 0.0364, 0.0264, 0.0286, 0.0247, 0.0273, 0.0344, 0.0344], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-15 19:59:39,743 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=122117.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 19:59:42,918 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.43 vs. limit=2.0 2023-05-15 19:59:52,106 INFO [finetune.py:992] (0/2) Epoch 2, batch 9900, loss[loss=0.1735, simple_loss=0.2627, pruned_loss=0.04209, over 11548.00 frames. ], tot_loss[loss=0.1778, simple_loss=0.2654, pruned_loss=0.04512, over 2385323.23 frames. ], batch size: 48, lr: 4.94e-03, grad_scale: 8.0 2023-05-15 20:00:24,039 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=122178.0, num_to_drop=1, layers_to_drop={0} 2023-05-15 20:00:28,802 INFO [finetune.py:992] (0/2) Epoch 2, batch 9950, loss[loss=0.1828, simple_loss=0.2715, pruned_loss=0.04707, over 12022.00 frames. ], tot_loss[loss=0.1781, simple_loss=0.2658, pruned_loss=0.04521, over 2374602.39 frames. ], batch size: 42, lr: 4.94e-03, grad_scale: 8.0 2023-05-15 20:00:35,813 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=122195.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 20:00:47,441 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.059e+02 2.921e+02 3.481e+02 4.108e+02 1.157e+03, threshold=6.963e+02, percent-clipped=3.0 2023-05-15 20:01:04,626 INFO [finetune.py:992] (0/2) Epoch 2, batch 10000, loss[loss=0.1784, simple_loss=0.2713, pruned_loss=0.04275, over 12185.00 frames. ], tot_loss[loss=0.1784, simple_loss=0.2662, pruned_loss=0.04524, over 2380160.87 frames. ], batch size: 35, lr: 4.94e-03, grad_scale: 8.0 2023-05-15 20:01:09,735 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.2745, 2.2342, 3.5449, 4.2819, 3.8733, 4.3121, 3.8319, 2.9459], device='cuda:0'), covar=tensor([0.0033, 0.0378, 0.0111, 0.0032, 0.0094, 0.0061, 0.0084, 0.0312], device='cuda:0'), in_proj_covar=tensor([0.0078, 0.0117, 0.0097, 0.0073, 0.0098, 0.0108, 0.0086, 0.0130], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-15 20:01:19,686 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=122256.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 20:01:40,205 INFO [finetune.py:992] (0/2) Epoch 2, batch 10050, loss[loss=0.151, simple_loss=0.2371, pruned_loss=0.03247, over 12130.00 frames. ], tot_loss[loss=0.178, simple_loss=0.2659, pruned_loss=0.04508, over 2379473.17 frames. ], batch size: 30, lr: 4.94e-03, grad_scale: 8.0 2023-05-15 20:01:56,595 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=122306.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 20:01:59,984 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.891e+02 2.902e+02 3.287e+02 3.882e+02 8.143e+02, threshold=6.574e+02, percent-clipped=1.0 2023-05-15 20:02:05,964 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=122319.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 20:02:06,730 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.1813, 4.0092, 4.1158, 4.5416, 3.2309, 3.8624, 2.6100, 4.1937], device='cuda:0'), covar=tensor([0.1559, 0.0704, 0.0836, 0.0638, 0.0977, 0.0629, 0.1772, 0.1193], device='cuda:0'), in_proj_covar=tensor([0.0223, 0.0261, 0.0296, 0.0353, 0.0238, 0.0238, 0.0256, 0.0388], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-15 20:02:17,043 INFO [finetune.py:992] (0/2) Epoch 2, batch 10100, loss[loss=0.1887, simple_loss=0.2762, pruned_loss=0.05058, over 12155.00 frames. ], tot_loss[loss=0.1786, simple_loss=0.2661, pruned_loss=0.04553, over 2370330.86 frames. ], batch size: 36, lr: 4.94e-03, grad_scale: 4.0 2023-05-15 20:02:48,789 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=122380.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 20:02:52,160 INFO [finetune.py:992] (0/2) Epoch 2, batch 10150, loss[loss=0.1467, simple_loss=0.2282, pruned_loss=0.03262, over 11329.00 frames. ], tot_loss[loss=0.1782, simple_loss=0.2657, pruned_loss=0.04542, over 2371872.19 frames. ], batch size: 25, lr: 4.94e-03, grad_scale: 4.0 2023-05-15 20:03:06,773 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.4679, 2.2480, 3.7507, 4.4355, 4.0270, 4.4850, 4.0287, 3.0267], device='cuda:0'), covar=tensor([0.0030, 0.0399, 0.0099, 0.0039, 0.0078, 0.0056, 0.0075, 0.0318], device='cuda:0'), in_proj_covar=tensor([0.0079, 0.0119, 0.0099, 0.0074, 0.0099, 0.0109, 0.0087, 0.0131], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-15 20:03:09,308 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.39 vs. limit=2.0 2023-05-15 20:03:11,566 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.036e+02 2.956e+02 3.518e+02 4.206e+02 6.663e+02, threshold=7.036e+02, percent-clipped=1.0 2023-05-15 20:03:22,722 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.9209, 2.3713, 3.3169, 2.8723, 3.2542, 2.8845, 2.3774, 3.3085], device='cuda:0'), covar=tensor([0.0108, 0.0289, 0.0141, 0.0209, 0.0127, 0.0206, 0.0298, 0.0115], device='cuda:0'), in_proj_covar=tensor([0.0156, 0.0191, 0.0167, 0.0172, 0.0191, 0.0148, 0.0181, 0.0162], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-15 20:03:28,114 INFO [finetune.py:992] (0/2) Epoch 2, batch 10200, loss[loss=0.189, simple_loss=0.2789, pruned_loss=0.04954, over 12115.00 frames. ], tot_loss[loss=0.1778, simple_loss=0.265, pruned_loss=0.04527, over 2380870.50 frames. ], batch size: 38, lr: 4.94e-03, grad_scale: 4.0 2023-05-15 20:03:52,376 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=122468.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 20:03:55,752 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=122473.0, num_to_drop=1, layers_to_drop={0} 2023-05-15 20:04:00,204 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.50 vs. limit=2.0 2023-05-15 20:04:04,163 INFO [finetune.py:992] (0/2) Epoch 2, batch 10250, loss[loss=0.1595, simple_loss=0.2422, pruned_loss=0.03843, over 12305.00 frames. ], tot_loss[loss=0.1786, simple_loss=0.266, pruned_loss=0.04557, over 2375895.95 frames. ], batch size: 34, lr: 4.94e-03, grad_scale: 4.0 2023-05-15 20:04:23,102 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.148e+02 2.763e+02 3.392e+02 4.054e+02 1.109e+03, threshold=6.785e+02, percent-clipped=4.0 2023-05-15 20:04:26,169 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.2775, 2.8594, 2.8135, 2.7852, 2.5112, 2.3445, 2.8472, 1.9872], device='cuda:0'), covar=tensor([0.0293, 0.0134, 0.0128, 0.0144, 0.0300, 0.0235, 0.0114, 0.0374], device='cuda:0'), in_proj_covar=tensor([0.0183, 0.0156, 0.0146, 0.0177, 0.0197, 0.0191, 0.0157, 0.0197], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-15 20:04:35,520 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=122529.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 20:04:39,627 INFO [finetune.py:992] (0/2) Epoch 2, batch 10300, loss[loss=0.183, simple_loss=0.2703, pruned_loss=0.04779, over 12363.00 frames. ], tot_loss[loss=0.1788, simple_loss=0.2664, pruned_loss=0.04555, over 2369545.75 frames. ], batch size: 38, lr: 4.94e-03, grad_scale: 4.0 2023-05-15 20:04:51,362 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=122551.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 20:05:16,315 INFO [finetune.py:992] (0/2) Epoch 2, batch 10350, loss[loss=0.1572, simple_loss=0.2402, pruned_loss=0.03715, over 12019.00 frames. ], tot_loss[loss=0.1771, simple_loss=0.2647, pruned_loss=0.04477, over 2375064.52 frames. ], batch size: 31, lr: 4.94e-03, grad_scale: 4.0 2023-05-15 20:05:26,379 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.4687, 4.9620, 5.3953, 4.7481, 4.9888, 4.7704, 5.4292, 5.1056], device='cuda:0'), covar=tensor([0.0214, 0.0318, 0.0244, 0.0213, 0.0330, 0.0317, 0.0246, 0.0212], device='cuda:0'), in_proj_covar=tensor([0.0228, 0.0236, 0.0252, 0.0229, 0.0232, 0.0229, 0.0210, 0.0195], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-05-15 20:05:32,342 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=122606.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 20:05:36,471 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.681e+02 2.820e+02 3.400e+02 4.097e+02 7.939e+02, threshold=6.800e+02, percent-clipped=1.0 2023-05-15 20:05:45,265 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.81 vs. limit=2.0 2023-05-15 20:05:52,630 INFO [finetune.py:992] (0/2) Epoch 2, batch 10400, loss[loss=0.1671, simple_loss=0.2483, pruned_loss=0.04294, over 11781.00 frames. ], tot_loss[loss=0.1768, simple_loss=0.2645, pruned_loss=0.04457, over 2381315.78 frames. ], batch size: 26, lr: 4.94e-03, grad_scale: 8.0 2023-05-15 20:06:02,822 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.5588, 2.7533, 4.4031, 4.7771, 3.2067, 2.6115, 2.8674, 2.0043], device='cuda:0'), covar=tensor([0.1300, 0.2707, 0.0481, 0.0292, 0.0904, 0.1813, 0.2267, 0.3466], device='cuda:0'), in_proj_covar=tensor([0.0272, 0.0368, 0.0266, 0.0287, 0.0249, 0.0275, 0.0347, 0.0348], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-15 20:06:06,078 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=122654.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 20:06:20,942 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=122675.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 20:06:25,359 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.6485, 3.0334, 5.0131, 2.7582, 2.6671, 3.8097, 3.0541, 3.8185], device='cuda:0'), covar=tensor([0.0453, 0.1288, 0.0351, 0.1164, 0.1941, 0.1259, 0.1402, 0.1048], device='cuda:0'), in_proj_covar=tensor([0.0232, 0.0227, 0.0233, 0.0178, 0.0233, 0.0278, 0.0223, 0.0250], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-15 20:06:27,901 INFO [finetune.py:992] (0/2) Epoch 2, batch 10450, loss[loss=0.1542, simple_loss=0.2385, pruned_loss=0.03493, over 12340.00 frames. ], tot_loss[loss=0.1771, simple_loss=0.2649, pruned_loss=0.04463, over 2380510.15 frames. ], batch size: 30, lr: 4.94e-03, grad_scale: 8.0 2023-05-15 20:06:46,906 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.858e+02 2.972e+02 3.516e+02 4.240e+02 1.276e+03, threshold=7.032e+02, percent-clipped=3.0 2023-05-15 20:06:57,732 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.2605, 4.5634, 2.7907, 2.4659, 3.7690, 2.5031, 3.8805, 3.0849], device='cuda:0'), covar=tensor([0.0649, 0.0500, 0.0968, 0.1472, 0.0353, 0.1228, 0.0411, 0.0772], device='cuda:0'), in_proj_covar=tensor([0.0196, 0.0255, 0.0178, 0.0203, 0.0142, 0.0184, 0.0198, 0.0178], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-15 20:07:03,785 INFO [finetune.py:992] (0/2) Epoch 2, batch 10500, loss[loss=0.2332, simple_loss=0.3293, pruned_loss=0.06854, over 10449.00 frames. ], tot_loss[loss=0.1762, simple_loss=0.2638, pruned_loss=0.04432, over 2384443.82 frames. ], batch size: 68, lr: 4.94e-03, grad_scale: 8.0 2023-05-15 20:07:10,506 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.61 vs. limit=2.0 2023-05-15 20:07:16,545 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.4283, 4.6935, 4.2406, 4.9672, 4.5691, 2.7339, 4.3113, 3.1110], device='cuda:0'), covar=tensor([0.0628, 0.0626, 0.1089, 0.0327, 0.0902, 0.1502, 0.0893, 0.2787], device='cuda:0'), in_proj_covar=tensor([0.0290, 0.0364, 0.0344, 0.0256, 0.0354, 0.0256, 0.0329, 0.0356], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-15 20:07:21,070 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=4.71 vs. limit=5.0 2023-05-15 20:07:31,043 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=122773.0, num_to_drop=1, layers_to_drop={2} 2023-05-15 20:07:39,212 INFO [finetune.py:992] (0/2) Epoch 2, batch 10550, loss[loss=0.1948, simple_loss=0.2862, pruned_loss=0.05165, over 11889.00 frames. ], tot_loss[loss=0.1775, simple_loss=0.2655, pruned_loss=0.04472, over 2380111.68 frames. ], batch size: 44, lr: 4.94e-03, grad_scale: 8.0 2023-05-15 20:07:56,330 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 2023-05-15 20:07:58,686 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.095e+02 2.924e+02 3.382e+02 3.985e+02 8.472e+02, threshold=6.763e+02, percent-clipped=1.0 2023-05-15 20:08:05,221 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=122821.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 20:08:06,800 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.5716, 2.4140, 3.6148, 4.5627, 4.1504, 4.4598, 4.0665, 2.9759], device='cuda:0'), covar=tensor([0.0028, 0.0356, 0.0121, 0.0033, 0.0078, 0.0074, 0.0081, 0.0325], device='cuda:0'), in_proj_covar=tensor([0.0078, 0.0117, 0.0098, 0.0073, 0.0097, 0.0107, 0.0086, 0.0130], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-15 20:08:07,410 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=122824.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 20:08:15,042 INFO [finetune.py:992] (0/2) Epoch 2, batch 10600, loss[loss=0.1662, simple_loss=0.2575, pruned_loss=0.03748, over 11048.00 frames. ], tot_loss[loss=0.1763, simple_loss=0.264, pruned_loss=0.04432, over 2376255.82 frames. ], batch size: 55, lr: 4.94e-03, grad_scale: 8.0 2023-05-15 20:08:26,872 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=122851.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 20:08:42,431 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=122872.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 20:08:51,534 INFO [finetune.py:992] (0/2) Epoch 2, batch 10650, loss[loss=0.1817, simple_loss=0.2778, pruned_loss=0.04274, over 12155.00 frames. ], tot_loss[loss=0.1757, simple_loss=0.2632, pruned_loss=0.04407, over 2382836.37 frames. ], batch size: 34, lr: 4.94e-03, grad_scale: 8.0 2023-05-15 20:08:57,201 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 2023-05-15 20:09:02,341 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=122899.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 20:09:11,589 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.792e+02 2.882e+02 3.250e+02 4.087e+02 1.166e+03, threshold=6.501e+02, percent-clipped=2.0 2023-05-15 20:09:26,419 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=122933.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 20:09:27,614 INFO [finetune.py:992] (0/2) Epoch 2, batch 10700, loss[loss=0.1725, simple_loss=0.2513, pruned_loss=0.0469, over 12354.00 frames. ], tot_loss[loss=0.1756, simple_loss=0.263, pruned_loss=0.04411, over 2381759.31 frames. ], batch size: 30, lr: 4.94e-03, grad_scale: 8.0 2023-05-15 20:09:45,731 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.8496, 2.6421, 4.7565, 5.0259, 3.4975, 2.5875, 3.0380, 2.0844], device='cuda:0'), covar=tensor([0.1295, 0.3196, 0.0365, 0.0286, 0.0792, 0.2088, 0.2388, 0.4271], device='cuda:0'), in_proj_covar=tensor([0.0268, 0.0364, 0.0262, 0.0284, 0.0246, 0.0272, 0.0344, 0.0345], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-15 20:09:56,051 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=122975.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 20:10:03,050 INFO [finetune.py:992] (0/2) Epoch 2, batch 10750, loss[loss=0.2043, simple_loss=0.2898, pruned_loss=0.05943, over 10602.00 frames. ], tot_loss[loss=0.1765, simple_loss=0.2635, pruned_loss=0.04477, over 2378847.65 frames. ], batch size: 69, lr: 4.94e-03, grad_scale: 8.0 2023-05-15 20:10:23,398 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.957e+02 2.914e+02 3.371e+02 4.358e+02 9.799e+02, threshold=6.743e+02, percent-clipped=2.0 2023-05-15 20:10:31,371 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=123023.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 20:10:39,763 INFO [finetune.py:992] (0/2) Epoch 2, batch 10800, loss[loss=0.1614, simple_loss=0.2469, pruned_loss=0.03799, over 11990.00 frames. ], tot_loss[loss=0.1754, simple_loss=0.2627, pruned_loss=0.04403, over 2389042.42 frames. ], batch size: 28, lr: 4.94e-03, grad_scale: 8.0 2023-05-15 20:11:15,868 INFO [finetune.py:992] (0/2) Epoch 2, batch 10850, loss[loss=0.1724, simple_loss=0.2503, pruned_loss=0.04721, over 12134.00 frames. ], tot_loss[loss=0.1764, simple_loss=0.2636, pruned_loss=0.04464, over 2379122.84 frames. ], batch size: 30, lr: 4.94e-03, grad_scale: 8.0 2023-05-15 20:11:36,886 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.974e+02 3.070e+02 3.614e+02 4.156e+02 8.622e+02, threshold=7.228e+02, percent-clipped=2.0 2023-05-15 20:11:43,847 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.88 vs. limit=2.0 2023-05-15 20:11:44,127 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=123124.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 20:11:51,923 INFO [finetune.py:992] (0/2) Epoch 2, batch 10900, loss[loss=0.1881, simple_loss=0.2651, pruned_loss=0.05557, over 12139.00 frames. ], tot_loss[loss=0.1776, simple_loss=0.2648, pruned_loss=0.04524, over 2375883.00 frames. ], batch size: 30, lr: 4.94e-03, grad_scale: 4.0 2023-05-15 20:11:54,336 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.50 vs. limit=2.0 2023-05-15 20:12:18,587 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=123172.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 20:12:28,239 INFO [finetune.py:992] (0/2) Epoch 2, batch 10950, loss[loss=0.1622, simple_loss=0.2554, pruned_loss=0.03455, over 12347.00 frames. ], tot_loss[loss=0.1791, simple_loss=0.2661, pruned_loss=0.04604, over 2368273.21 frames. ], batch size: 36, lr: 4.93e-03, grad_scale: 4.0 2023-05-15 20:12:32,418 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=3.14 vs. limit=5.0 2023-05-15 20:12:49,144 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.014e+02 2.930e+02 3.524e+02 4.092e+02 6.133e+02, threshold=7.048e+02, percent-clipped=1.0 2023-05-15 20:12:49,632 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=4.83 vs. limit=5.0 2023-05-15 20:12:59,131 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=123228.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 20:13:03,914 INFO [finetune.py:992] (0/2) Epoch 2, batch 11000, loss[loss=0.1714, simple_loss=0.2518, pruned_loss=0.04554, over 12178.00 frames. ], tot_loss[loss=0.1818, simple_loss=0.2686, pruned_loss=0.04744, over 2352465.12 frames. ], batch size: 29, lr: 4.93e-03, grad_scale: 4.0 2023-05-15 20:13:39,339 INFO [finetune.py:992] (0/2) Epoch 2, batch 11050, loss[loss=0.2897, simple_loss=0.3646, pruned_loss=0.1074, over 10441.00 frames. ], tot_loss[loss=0.1857, simple_loss=0.2722, pruned_loss=0.04959, over 2317240.25 frames. ], batch size: 68, lr: 4.93e-03, grad_scale: 4.0 2023-05-15 20:14:00,282 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.041e+02 3.317e+02 4.016e+02 4.800e+02 7.007e+02, threshold=8.032e+02, percent-clipped=0.0 2023-05-15 20:14:00,524 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=123314.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 20:14:12,427 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=2.05 vs. limit=2.0 2023-05-15 20:14:14,843 INFO [finetune.py:992] (0/2) Epoch 2, batch 11100, loss[loss=0.231, simple_loss=0.3069, pruned_loss=0.07755, over 7995.00 frames. ], tot_loss[loss=0.1899, simple_loss=0.2754, pruned_loss=0.05218, over 2267094.49 frames. ], batch size: 98, lr: 4.93e-03, grad_scale: 4.0 2023-05-15 20:14:15,401 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.33 vs. limit=2.0 2023-05-15 20:14:39,709 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.7176, 2.3151, 2.9321, 3.7407, 2.0658, 3.8851, 3.7398, 3.9131], device='cuda:0'), covar=tensor([0.0137, 0.1088, 0.0394, 0.0097, 0.1184, 0.0155, 0.0215, 0.0079], device='cuda:0'), in_proj_covar=tensor([0.0119, 0.0199, 0.0183, 0.0113, 0.0183, 0.0171, 0.0165, 0.0116], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-15 20:14:43,718 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=123375.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 20:14:50,127 INFO [finetune.py:992] (0/2) Epoch 2, batch 11150, loss[loss=0.2166, simple_loss=0.3, pruned_loss=0.06659, over 12023.00 frames. ], tot_loss[loss=0.1982, simple_loss=0.2822, pruned_loss=0.05713, over 2204607.51 frames. ], batch size: 42, lr: 4.93e-03, grad_scale: 4.0 2023-05-15 20:15:05,875 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=5.64 vs. limit=5.0 2023-05-15 20:15:11,055 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.003e+02 3.729e+02 4.554e+02 5.583e+02 9.300e+02, threshold=9.109e+02, percent-clipped=4.0 2023-05-15 20:15:20,926 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=123428.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 20:15:25,758 INFO [finetune.py:992] (0/2) Epoch 2, batch 11200, loss[loss=0.2643, simple_loss=0.3454, pruned_loss=0.09159, over 10486.00 frames. ], tot_loss[loss=0.2075, simple_loss=0.2896, pruned_loss=0.06268, over 2136114.83 frames. ], batch size: 68, lr: 4.93e-03, grad_scale: 8.0 2023-05-15 20:16:01,415 INFO [finetune.py:992] (0/2) Epoch 2, batch 11250, loss[loss=0.2477, simple_loss=0.3417, pruned_loss=0.07679, over 10377.00 frames. ], tot_loss[loss=0.2165, simple_loss=0.2973, pruned_loss=0.06787, over 2083708.38 frames. ], batch size: 68, lr: 4.93e-03, grad_scale: 8.0 2023-05-15 20:16:04,423 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=123489.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 20:16:21,093 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.415e+02 3.857e+02 4.379e+02 5.227e+02 1.226e+03, threshold=8.757e+02, percent-clipped=1.0 2023-05-15 20:16:31,190 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=123528.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 20:16:36,541 INFO [finetune.py:992] (0/2) Epoch 2, batch 11300, loss[loss=0.2872, simple_loss=0.3499, pruned_loss=0.1123, over 6516.00 frames. ], tot_loss[loss=0.2241, simple_loss=0.3038, pruned_loss=0.0722, over 2028967.37 frames. ], batch size: 99, lr: 4.93e-03, grad_scale: 8.0 2023-05-15 20:16:45,145 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.48 vs. limit=2.0 2023-05-15 20:17:04,532 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=123576.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 20:17:10,545 INFO [finetune.py:992] (0/2) Epoch 2, batch 11350, loss[loss=0.288, simple_loss=0.3687, pruned_loss=0.1037, over 6782.00 frames. ], tot_loss[loss=0.2296, simple_loss=0.3084, pruned_loss=0.07537, over 1966737.86 frames. ], batch size: 99, lr: 4.93e-03, grad_scale: 8.0 2023-05-15 20:17:31,981 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.677e+02 3.491e+02 4.298e+02 4.901e+02 8.727e+02, threshold=8.596e+02, percent-clipped=0.0 2023-05-15 20:17:40,514 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=5.18 vs. limit=5.0 2023-05-15 20:17:46,098 INFO [finetune.py:992] (0/2) Epoch 2, batch 11400, loss[loss=0.258, simple_loss=0.3373, pruned_loss=0.08935, over 10583.00 frames. ], tot_loss[loss=0.2343, simple_loss=0.3128, pruned_loss=0.07792, over 1945205.21 frames. ], batch size: 69, lr: 4.93e-03, grad_scale: 8.0 2023-05-15 20:18:10,405 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=123670.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 20:18:21,021 INFO [finetune.py:992] (0/2) Epoch 2, batch 11450, loss[loss=0.2837, simple_loss=0.3485, pruned_loss=0.1095, over 7022.00 frames. ], tot_loss[loss=0.2399, simple_loss=0.3166, pruned_loss=0.08155, over 1893916.77 frames. ], batch size: 97, lr: 4.93e-03, grad_scale: 8.0 2023-05-15 20:18:30,717 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.6490, 1.9776, 2.9850, 2.5459, 2.8957, 2.8233, 1.9430, 2.9034], device='cuda:0'), covar=tensor([0.0088, 0.0325, 0.0085, 0.0183, 0.0110, 0.0141, 0.0342, 0.0088], device='cuda:0'), in_proj_covar=tensor([0.0150, 0.0183, 0.0158, 0.0164, 0.0180, 0.0141, 0.0174, 0.0152], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-15 20:18:41,102 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.333e+02 3.687e+02 4.414e+02 5.331e+02 1.195e+03, threshold=8.827e+02, percent-clipped=1.0 2023-05-15 20:18:51,378 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.39 vs. limit=2.0 2023-05-15 20:18:54,874 INFO [finetune.py:992] (0/2) Epoch 2, batch 11500, loss[loss=0.3216, simple_loss=0.3749, pruned_loss=0.1341, over 6959.00 frames. ], tot_loss[loss=0.245, simple_loss=0.3202, pruned_loss=0.0849, over 1853308.98 frames. ], batch size: 99, lr: 4.93e-03, grad_scale: 8.0 2023-05-15 20:19:29,359 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=123784.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 20:19:29,945 INFO [finetune.py:992] (0/2) Epoch 2, batch 11550, loss[loss=0.2709, simple_loss=0.3424, pruned_loss=0.09971, over 6688.00 frames. ], tot_loss[loss=0.247, simple_loss=0.3215, pruned_loss=0.08625, over 1824621.31 frames. ], batch size: 98, lr: 4.93e-03, grad_scale: 8.0 2023-05-15 20:19:50,520 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.512e+02 3.528e+02 4.263e+02 5.111e+02 8.080e+02, threshold=8.525e+02, percent-clipped=0.0 2023-05-15 20:20:03,632 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 2023-05-15 20:20:04,555 INFO [finetune.py:992] (0/2) Epoch 2, batch 11600, loss[loss=0.305, simple_loss=0.3506, pruned_loss=0.1297, over 7252.00 frames. ], tot_loss[loss=0.248, simple_loss=0.3217, pruned_loss=0.08716, over 1817957.55 frames. ], batch size: 100, lr: 4.93e-03, grad_scale: 8.0 2023-05-15 20:20:29,888 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.7529, 3.0898, 2.3412, 2.1500, 2.7990, 2.2435, 2.9790, 2.5526], device='cuda:0'), covar=tensor([0.0507, 0.0644, 0.0807, 0.1321, 0.0209, 0.1127, 0.0418, 0.0741], device='cuda:0'), in_proj_covar=tensor([0.0184, 0.0234, 0.0169, 0.0192, 0.0131, 0.0175, 0.0185, 0.0169], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-15 20:20:32,261 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=123872.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 20:20:41,340 INFO [finetune.py:992] (0/2) Epoch 2, batch 11650, loss[loss=0.232, simple_loss=0.3154, pruned_loss=0.07431, over 11458.00 frames. ], tot_loss[loss=0.248, simple_loss=0.3215, pruned_loss=0.08719, over 1806103.34 frames. ], batch size: 48, lr: 4.93e-03, grad_scale: 8.0 2023-05-15 20:21:01,764 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.751e+02 3.666e+02 4.379e+02 5.316e+02 1.010e+03, threshold=8.758e+02, percent-clipped=1.0 2023-05-15 20:21:14,929 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=4.78 vs. limit=5.0 2023-05-15 20:21:15,354 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=123933.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 20:21:16,387 INFO [finetune.py:992] (0/2) Epoch 2, batch 11700, loss[loss=0.3161, simple_loss=0.3629, pruned_loss=0.1346, over 6552.00 frames. ], tot_loss[loss=0.2489, simple_loss=0.3217, pruned_loss=0.08805, over 1773312.96 frames. ], batch size: 98, lr: 4.93e-03, grad_scale: 8.0 2023-05-15 20:21:38,846 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.8906, 2.1165, 2.6233, 2.8903, 2.8699, 2.9326, 2.8199, 2.4045], device='cuda:0'), covar=tensor([0.0051, 0.0334, 0.0152, 0.0054, 0.0099, 0.0092, 0.0081, 0.0312], device='cuda:0'), in_proj_covar=tensor([0.0075, 0.0115, 0.0094, 0.0071, 0.0093, 0.0104, 0.0082, 0.0128], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-15 20:21:40,827 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=123970.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 20:21:50,544 INFO [finetune.py:992] (0/2) Epoch 2, batch 11750, loss[loss=0.2704, simple_loss=0.3225, pruned_loss=0.1091, over 6556.00 frames. ], tot_loss[loss=0.2504, simple_loss=0.3224, pruned_loss=0.08919, over 1742252.07 frames. ], batch size: 97, lr: 4.93e-03, grad_scale: 8.0 2023-05-15 20:22:01,704 INFO [checkpoint.py:75] (0/2) Saving checkpoint to pruned_transducer_stateless7/exp_giga_finetune/checkpoint-24000.pt 2023-05-15 20:22:06,386 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.4758, 4.4519, 4.3822, 4.0236, 4.1315, 4.4873, 4.2058, 4.1236], device='cuda:0'), covar=tensor([0.0868, 0.0933, 0.0624, 0.1327, 0.2474, 0.0799, 0.1396, 0.1032], device='cuda:0'), in_proj_covar=tensor([0.0533, 0.0479, 0.0434, 0.0541, 0.0360, 0.0603, 0.0658, 0.0483], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-05-15 20:22:14,223 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.778e+02 3.583e+02 4.299e+02 5.119e+02 9.573e+02, threshold=8.597e+02, percent-clipped=1.0 2023-05-15 20:22:17,074 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=124018.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 20:22:29,144 INFO [finetune.py:992] (0/2) Epoch 2, batch 11800, loss[loss=0.2701, simple_loss=0.3373, pruned_loss=0.1014, over 6451.00 frames. ], tot_loss[loss=0.2527, simple_loss=0.3243, pruned_loss=0.09059, over 1728115.22 frames. ], batch size: 100, lr: 4.93e-03, grad_scale: 8.0 2023-05-15 20:22:30,098 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=124036.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 20:22:37,026 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=3.03 vs. limit=5.0 2023-05-15 20:22:48,143 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=124062.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 20:23:03,284 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=124084.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 20:23:03,786 INFO [finetune.py:992] (0/2) Epoch 2, batch 11850, loss[loss=0.2127, simple_loss=0.2962, pruned_loss=0.06456, over 11139.00 frames. ], tot_loss[loss=0.2541, simple_loss=0.3259, pruned_loss=0.09121, over 1716330.26 frames. ], batch size: 55, lr: 4.93e-03, grad_scale: 8.0 2023-05-15 20:23:12,234 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=124097.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 20:23:22,636 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.77 vs. limit=2.0 2023-05-15 20:23:24,100 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.378e+02 3.700e+02 4.417e+02 5.163e+02 1.393e+03, threshold=8.834e+02, percent-clipped=2.0 2023-05-15 20:23:30,443 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=124123.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 20:23:36,440 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=124132.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 20:23:38,333 INFO [finetune.py:992] (0/2) Epoch 2, batch 11900, loss[loss=0.251, simple_loss=0.3246, pruned_loss=0.08863, over 7142.00 frames. ], tot_loss[loss=0.2522, simple_loss=0.325, pruned_loss=0.0897, over 1709687.83 frames. ], batch size: 98, lr: 4.93e-03, grad_scale: 8.0 2023-05-15 20:24:13,588 INFO [finetune.py:992] (0/2) Epoch 2, batch 11950, loss[loss=0.1881, simple_loss=0.2798, pruned_loss=0.04822, over 11116.00 frames. ], tot_loss[loss=0.2475, simple_loss=0.3215, pruned_loss=0.08675, over 1701501.46 frames. ], batch size: 55, lr: 4.93e-03, grad_scale: 8.0 2023-05-15 20:24:34,132 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.275e+02 3.145e+02 3.876e+02 4.654e+02 9.978e+02, threshold=7.752e+02, percent-clipped=1.0 2023-05-15 20:24:43,696 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=124228.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 20:24:48,922 INFO [finetune.py:992] (0/2) Epoch 2, batch 12000, loss[loss=0.2388, simple_loss=0.3037, pruned_loss=0.08698, over 6745.00 frames. ], tot_loss[loss=0.2418, simple_loss=0.3173, pruned_loss=0.08318, over 1699246.75 frames. ], batch size: 98, lr: 4.93e-03, grad_scale: 8.0 2023-05-15 20:24:48,923 INFO [finetune.py:1017] (0/2) Computing validation loss 2023-05-15 20:24:55,538 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.6467, 1.8680, 2.9969, 2.4617, 2.8054, 2.8038, 1.9539, 2.8012], device='cuda:0'), covar=tensor([0.0080, 0.0321, 0.0075, 0.0174, 0.0124, 0.0109, 0.0284, 0.0081], device='cuda:0'), in_proj_covar=tensor([0.0141, 0.0176, 0.0147, 0.0157, 0.0170, 0.0134, 0.0166, 0.0143], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-15 20:25:06,676 INFO [finetune.py:1026] (0/2) Epoch 2, validation: loss=0.2947, simple_loss=0.3701, pruned_loss=0.1097, over 1020973.00 frames. 2023-05-15 20:25:06,677 INFO [finetune.py:1027] (0/2) Maximum memory allocated so far is 12745MB 2023-05-15 20:25:32,584 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.9239, 3.8790, 3.8776, 3.9636, 3.7561, 3.7801, 3.7608, 3.8875], device='cuda:0'), covar=tensor([0.0803, 0.0610, 0.1068, 0.0682, 0.1496, 0.1153, 0.0543, 0.0877], device='cuda:0'), in_proj_covar=tensor([0.0430, 0.0551, 0.0480, 0.0517, 0.0664, 0.0622, 0.0459, 0.0417], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0004, 0.0004, 0.0005, 0.0004, 0.0003, 0.0003], device='cuda:0') 2023-05-15 20:25:32,626 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=124272.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 20:25:41,014 INFO [finetune.py:992] (0/2) Epoch 2, batch 12050, loss[loss=0.2016, simple_loss=0.2897, pruned_loss=0.05676, over 10361.00 frames. ], tot_loss[loss=0.2363, simple_loss=0.3132, pruned_loss=0.07969, over 1710311.45 frames. ], batch size: 68, lr: 4.93e-03, grad_scale: 8.0 2023-05-15 20:25:43,326 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.0555, 3.0818, 4.4564, 2.4893, 2.6637, 3.5094, 2.9582, 3.6329], device='cuda:0'), covar=tensor([0.0412, 0.1108, 0.0216, 0.1275, 0.1866, 0.1112, 0.1446, 0.0911], device='cuda:0'), in_proj_covar=tensor([0.0212, 0.0213, 0.0207, 0.0166, 0.0218, 0.0253, 0.0207, 0.0225], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:0') 2023-05-15 20:25:46,043 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.5003, 2.9920, 3.8197, 2.3265, 2.6456, 3.0951, 2.8570, 3.2234], device='cuda:0'), covar=tensor([0.0419, 0.1025, 0.0296, 0.1282, 0.1680, 0.1112, 0.1190, 0.1032], device='cuda:0'), in_proj_covar=tensor([0.0212, 0.0212, 0.0207, 0.0166, 0.0217, 0.0253, 0.0207, 0.0224], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:0') 2023-05-15 20:26:00,646 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.298e+02 3.069e+02 3.589e+02 4.447e+02 1.626e+03, threshold=7.177e+02, percent-clipped=4.0 2023-05-15 20:26:12,924 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=124333.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 20:26:14,012 INFO [finetune.py:992] (0/2) Epoch 2, batch 12100, loss[loss=0.2088, simple_loss=0.2923, pruned_loss=0.06259, over 10358.00 frames. ], tot_loss[loss=0.2333, simple_loss=0.3108, pruned_loss=0.07789, over 1706701.80 frames. ], batch size: 69, lr: 4.93e-03, grad_scale: 8.0 2023-05-15 20:26:22,597 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=124348.0, num_to_drop=1, layers_to_drop={1} 2023-05-15 20:26:25,249 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.3395, 3.5060, 3.2506, 3.6246, 3.4468, 2.5054, 3.2324, 2.8607], device='cuda:0'), covar=tensor([0.0856, 0.0957, 0.1398, 0.0453, 0.1130, 0.1666, 0.1013, 0.2749], device='cuda:0'), in_proj_covar=tensor([0.0275, 0.0343, 0.0322, 0.0232, 0.0331, 0.0249, 0.0312, 0.0337], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-15 20:26:35,063 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.47 vs. limit=2.0 2023-05-15 20:26:46,133 INFO [finetune.py:992] (0/2) Epoch 2, batch 12150, loss[loss=0.229, simple_loss=0.3177, pruned_loss=0.07015, over 10420.00 frames. ], tot_loss[loss=0.2349, simple_loss=0.3119, pruned_loss=0.07898, over 1697983.97 frames. ], batch size: 68, lr: 4.93e-03, grad_scale: 8.0 2023-05-15 20:26:50,649 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=124392.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 20:27:02,238 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=124409.0, num_to_drop=1, layers_to_drop={3} 2023-05-15 20:27:05,399 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.374e+02 3.308e+02 3.785e+02 4.469e+02 7.285e+02, threshold=7.570e+02, percent-clipped=1.0 2023-05-15 20:27:08,027 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=124418.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 20:27:16,864 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=124432.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 20:27:18,576 INFO [finetune.py:992] (0/2) Epoch 2, batch 12200, loss[loss=0.2543, simple_loss=0.3299, pruned_loss=0.08937, over 7289.00 frames. ], tot_loss[loss=0.2371, simple_loss=0.3135, pruned_loss=0.08038, over 1678222.43 frames. ], batch size: 98, lr: 4.93e-03, grad_scale: 8.0 2023-05-15 20:27:29,243 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.3145, 2.9913, 3.6671, 2.2208, 2.6340, 3.0335, 2.7884, 3.1531], device='cuda:0'), covar=tensor([0.0466, 0.0965, 0.0310, 0.1478, 0.1772, 0.1152, 0.1219, 0.1062], device='cuda:0'), in_proj_covar=tensor([0.0212, 0.0213, 0.0206, 0.0166, 0.0219, 0.0253, 0.0208, 0.0224], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:0') 2023-05-15 20:27:39,874 INFO [checkpoint.py:75] (0/2) Saving checkpoint to pruned_transducer_stateless7/exp_giga_finetune/epoch-2.pt 2023-05-15 20:28:02,416 INFO [finetune.py:992] (0/2) Epoch 3, batch 0, loss[loss=0.1772, simple_loss=0.2635, pruned_loss=0.04546, over 12021.00 frames. ], tot_loss[loss=0.1772, simple_loss=0.2635, pruned_loss=0.04546, over 12021.00 frames. ], batch size: 28, lr: 4.93e-03, grad_scale: 8.0 2023-05-15 20:28:02,416 INFO [finetune.py:1017] (0/2) Computing validation loss 2023-05-15 20:28:19,572 INFO [finetune.py:1026] (0/2) Epoch 3, validation: loss=0.2957, simple_loss=0.3691, pruned_loss=0.1112, over 1020973.00 frames. 2023-05-15 20:28:19,573 INFO [finetune.py:1027] (0/2) Maximum memory allocated so far is 12745MB 2023-05-15 20:28:37,738 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=124493.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 20:28:52,937 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.479e+02 3.480e+02 3.921e+02 4.591e+02 7.457e+02, threshold=7.842e+02, percent-clipped=0.0 2023-05-15 20:28:55,732 INFO [finetune.py:992] (0/2) Epoch 3, batch 50, loss[loss=0.1616, simple_loss=0.2448, pruned_loss=0.03924, over 12288.00 frames. ], tot_loss[loss=0.1873, simple_loss=0.2751, pruned_loss=0.04976, over 546004.99 frames. ], batch size: 28, lr: 4.93e-03, grad_scale: 4.0 2023-05-15 20:29:02,939 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=124528.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 20:29:31,893 INFO [finetune.py:992] (0/2) Epoch 3, batch 100, loss[loss=0.2261, simple_loss=0.3035, pruned_loss=0.07432, over 8482.00 frames. ], tot_loss[loss=0.1855, simple_loss=0.2721, pruned_loss=0.04943, over 943115.07 frames. ], batch size: 98, lr: 4.93e-03, grad_scale: 4.0 2023-05-15 20:29:37,634 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=124576.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 20:29:55,855 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.9093, 4.5419, 4.6298, 4.7239, 4.5099, 4.7959, 4.6904, 2.4342], device='cuda:0'), covar=tensor([0.0102, 0.0063, 0.0093, 0.0067, 0.0061, 0.0089, 0.0078, 0.0828], device='cuda:0'), in_proj_covar=tensor([0.0059, 0.0067, 0.0071, 0.0065, 0.0053, 0.0079, 0.0069, 0.0085], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-15 20:30:05,512 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.093e+02 2.969e+02 3.564e+02 3.993e+02 8.882e+02, threshold=7.127e+02, percent-clipped=1.0 2023-05-15 20:30:07,670 INFO [finetune.py:992] (0/2) Epoch 3, batch 150, loss[loss=0.1765, simple_loss=0.2588, pruned_loss=0.04711, over 12172.00 frames. ], tot_loss[loss=0.1846, simple_loss=0.2711, pruned_loss=0.04902, over 1252360.58 frames. ], batch size: 29, lr: 4.93e-03, grad_scale: 4.0 2023-05-15 20:30:14,808 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=124628.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 20:30:22,058 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=124638.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 20:30:40,556 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=124663.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 20:30:43,768 INFO [finetune.py:992] (0/2) Epoch 3, batch 200, loss[loss=0.1798, simple_loss=0.2819, pruned_loss=0.03885, over 12162.00 frames. ], tot_loss[loss=0.185, simple_loss=0.2715, pruned_loss=0.04921, over 1492267.01 frames. ], batch size: 36, lr: 4.93e-03, grad_scale: 4.0 2023-05-15 20:31:01,523 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=124692.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 20:31:06,573 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=124699.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 20:31:09,986 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=124704.0, num_to_drop=1, layers_to_drop={3} 2023-05-15 20:31:17,439 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.886e+02 2.889e+02 3.397e+02 4.137e+02 7.935e+02, threshold=6.795e+02, percent-clipped=1.0 2023-05-15 20:31:19,626 INFO [finetune.py:992] (0/2) Epoch 3, batch 250, loss[loss=0.1934, simple_loss=0.2867, pruned_loss=0.05002, over 12268.00 frames. ], tot_loss[loss=0.1831, simple_loss=0.27, pruned_loss=0.04807, over 1694870.45 frames. ], batch size: 37, lr: 4.93e-03, grad_scale: 4.0 2023-05-15 20:31:19,813 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=124718.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 20:31:24,119 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=124724.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 20:31:35,432 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=124740.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 20:31:53,870 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=124766.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 20:31:55,247 INFO [finetune.py:992] (0/2) Epoch 3, batch 300, loss[loss=0.1784, simple_loss=0.2729, pruned_loss=0.042, over 12013.00 frames. ], tot_loss[loss=0.1814, simple_loss=0.2684, pruned_loss=0.04724, over 1847286.04 frames. ], batch size: 40, lr: 4.93e-03, grad_scale: 4.0 2023-05-15 20:32:09,669 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=124788.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 20:32:29,653 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.929e+02 2.943e+02 3.581e+02 4.310e+02 1.397e+03, threshold=7.162e+02, percent-clipped=2.0 2023-05-15 20:32:31,899 INFO [finetune.py:992] (0/2) Epoch 3, batch 350, loss[loss=0.1506, simple_loss=0.2405, pruned_loss=0.03039, over 12264.00 frames. ], tot_loss[loss=0.181, simple_loss=0.268, pruned_loss=0.04696, over 1971148.29 frames. ], batch size: 32, lr: 4.93e-03, grad_scale: 4.0 2023-05-15 20:32:33,539 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=124820.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 20:32:43,369 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.91 vs. limit=2.0 2023-05-15 20:33:05,861 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.5683, 5.2858, 5.4888, 5.4756, 5.1164, 5.1260, 4.9643, 5.4808], device='cuda:0'), covar=tensor([0.0652, 0.0591, 0.0667, 0.0612, 0.1942, 0.1410, 0.0508, 0.0886], device='cuda:0'), in_proj_covar=tensor([0.0453, 0.0583, 0.0505, 0.0545, 0.0711, 0.0655, 0.0479, 0.0437], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0005, 0.0005, 0.0003, 0.0004], device='cuda:0') 2023-05-15 20:33:07,893 INFO [finetune.py:992] (0/2) Epoch 3, batch 400, loss[loss=0.1552, simple_loss=0.2364, pruned_loss=0.03702, over 12129.00 frames. ], tot_loss[loss=0.1798, simple_loss=0.2668, pruned_loss=0.04638, over 2068107.23 frames. ], batch size: 30, lr: 4.93e-03, grad_scale: 8.0 2023-05-15 20:33:17,341 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=124881.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 20:33:41,351 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.028e+02 2.875e+02 3.470e+02 4.271e+02 8.495e+02, threshold=6.940e+02, percent-clipped=2.0 2023-05-15 20:33:43,533 INFO [finetune.py:992] (0/2) Epoch 3, batch 450, loss[loss=0.1686, simple_loss=0.2633, pruned_loss=0.03694, over 12160.00 frames. ], tot_loss[loss=0.1798, simple_loss=0.2675, pruned_loss=0.04605, over 2139267.85 frames. ], batch size: 34, lr: 4.92e-03, grad_scale: 8.0 2023-05-15 20:33:50,799 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=124928.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 20:34:19,822 INFO [finetune.py:992] (0/2) Epoch 3, batch 500, loss[loss=0.1674, simple_loss=0.2611, pruned_loss=0.03682, over 12093.00 frames. ], tot_loss[loss=0.1794, simple_loss=0.2676, pruned_loss=0.04563, over 2190970.95 frames. ], batch size: 32, lr: 4.92e-03, grad_scale: 8.0 2023-05-15 20:34:25,341 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=124976.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 20:34:34,950 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 2023-05-15 20:34:38,979 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=124994.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 20:34:46,468 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=125004.0, num_to_drop=1, layers_to_drop={0} 2023-05-15 20:34:54,250 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.166e+02 2.875e+02 3.419e+02 4.165e+02 9.341e+02, threshold=6.838e+02, percent-clipped=3.0 2023-05-15 20:34:56,420 INFO [finetune.py:992] (0/2) Epoch 3, batch 550, loss[loss=0.1558, simple_loss=0.2351, pruned_loss=0.03831, over 12255.00 frames. ], tot_loss[loss=0.1784, simple_loss=0.2665, pruned_loss=0.04519, over 2234730.23 frames. ], batch size: 28, lr: 4.92e-03, grad_scale: 8.0 2023-05-15 20:34:57,252 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=125019.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 20:35:00,247 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.6116, 2.4501, 4.0271, 4.1621, 2.8565, 2.5881, 2.6153, 2.1066], device='cuda:0'), covar=tensor([0.1343, 0.2943, 0.0505, 0.0438, 0.1032, 0.1908, 0.2650, 0.3753], device='cuda:0'), in_proj_covar=tensor([0.0266, 0.0359, 0.0257, 0.0278, 0.0242, 0.0273, 0.0343, 0.0350], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-15 20:35:16,567 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.8043, 4.7489, 4.5597, 4.6674, 4.3075, 4.8766, 4.7604, 5.1128], device='cuda:0'), covar=tensor([0.0250, 0.0184, 0.0248, 0.0354, 0.0829, 0.0277, 0.0191, 0.0188], device='cuda:0'), in_proj_covar=tensor([0.0166, 0.0168, 0.0168, 0.0210, 0.0212, 0.0183, 0.0154, 0.0202], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0005, 0.0004, 0.0004, 0.0003, 0.0005], device='cuda:0') 2023-05-15 20:35:20,763 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=125052.0, num_to_drop=1, layers_to_drop={0} 2023-05-15 20:35:26,332 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.43 vs. limit=2.0 2023-05-15 20:35:32,272 INFO [finetune.py:992] (0/2) Epoch 3, batch 600, loss[loss=0.1841, simple_loss=0.2788, pruned_loss=0.04464, over 12159.00 frames. ], tot_loss[loss=0.1784, simple_loss=0.2663, pruned_loss=0.04528, over 2264495.04 frames. ], batch size: 34, lr: 4.92e-03, grad_scale: 8.0 2023-05-15 20:35:46,520 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=125088.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 20:35:54,420 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.1064, 3.8774, 3.9568, 4.4433, 2.9894, 3.9075, 2.5200, 4.0288], device='cuda:0'), covar=tensor([0.1718, 0.0787, 0.0931, 0.0527, 0.1106, 0.0605, 0.1949, 0.1191], device='cuda:0'), in_proj_covar=tensor([0.0225, 0.0259, 0.0290, 0.0343, 0.0236, 0.0235, 0.0257, 0.0379], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-15 20:36:06,018 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.348e+02 2.876e+02 3.279e+02 4.102e+02 1.110e+03, threshold=6.559e+02, percent-clipped=4.0 2023-05-15 20:36:06,905 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.5885, 4.5703, 4.3941, 4.0486, 4.1613, 4.5607, 4.2200, 4.0697], device='cuda:0'), covar=tensor([0.0868, 0.0986, 0.0796, 0.1359, 0.2004, 0.0921, 0.1643, 0.1215], device='cuda:0'), in_proj_covar=tensor([0.0548, 0.0495, 0.0456, 0.0566, 0.0369, 0.0630, 0.0686, 0.0510], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0004, 0.0004, 0.0003], device='cuda:0') 2023-05-15 20:36:08,150 INFO [finetune.py:992] (0/2) Epoch 3, batch 650, loss[loss=0.1789, simple_loss=0.2743, pruned_loss=0.04174, over 12299.00 frames. ], tot_loss[loss=0.1777, simple_loss=0.2653, pruned_loss=0.04503, over 2292261.95 frames. ], batch size: 34, lr: 4.92e-03, grad_scale: 8.0 2023-05-15 20:36:15,342 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=125128.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 20:36:15,441 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.5190, 3.1503, 4.9069, 2.4939, 2.5106, 3.7204, 3.0262, 3.8215], device='cuda:0'), covar=tensor([0.0470, 0.1184, 0.0326, 0.1225, 0.2093, 0.1304, 0.1402, 0.1044], device='cuda:0'), in_proj_covar=tensor([0.0226, 0.0226, 0.0221, 0.0175, 0.0232, 0.0268, 0.0220, 0.0240], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:0') 2023-05-15 20:36:17,121 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.43 vs. limit=2.0 2023-05-15 20:36:21,002 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=125136.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 20:36:21,077 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.0321, 4.8363, 4.9618, 5.0017, 4.6209, 4.6701, 4.5178, 4.9562], device='cuda:0'), covar=tensor([0.0665, 0.0574, 0.0844, 0.0589, 0.1949, 0.1313, 0.0557, 0.0848], device='cuda:0'), in_proj_covar=tensor([0.0463, 0.0596, 0.0512, 0.0557, 0.0727, 0.0668, 0.0490, 0.0445], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0005, 0.0005, 0.0004, 0.0004], device='cuda:0') 2023-05-15 20:36:44,322 INFO [finetune.py:992] (0/2) Epoch 3, batch 700, loss[loss=0.1942, simple_loss=0.2794, pruned_loss=0.05447, over 11263.00 frames. ], tot_loss[loss=0.1779, simple_loss=0.2656, pruned_loss=0.04507, over 2309473.93 frames. ], batch size: 55, lr: 4.92e-03, grad_scale: 8.0 2023-05-15 20:36:49,941 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=125176.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 20:36:59,364 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=125189.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 20:37:17,938 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.041e+02 2.879e+02 3.347e+02 4.098e+02 6.743e+02, threshold=6.693e+02, percent-clipped=1.0 2023-05-15 20:37:20,164 INFO [finetune.py:992] (0/2) Epoch 3, batch 750, loss[loss=0.1797, simple_loss=0.2774, pruned_loss=0.04105, over 12264.00 frames. ], tot_loss[loss=0.1782, simple_loss=0.2659, pruned_loss=0.04524, over 2323601.24 frames. ], batch size: 37, lr: 4.92e-03, grad_scale: 8.0 2023-05-15 20:37:56,259 INFO [finetune.py:992] (0/2) Epoch 3, batch 800, loss[loss=0.205, simple_loss=0.2898, pruned_loss=0.06007, over 12348.00 frames. ], tot_loss[loss=0.1773, simple_loss=0.265, pruned_loss=0.04484, over 2341261.40 frames. ], batch size: 36, lr: 4.92e-03, grad_scale: 8.0 2023-05-15 20:38:15,523 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=125294.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 20:38:30,188 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.122e+02 2.921e+02 3.420e+02 4.045e+02 6.574e+02, threshold=6.840e+02, percent-clipped=0.0 2023-05-15 20:38:31,469 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=4.72 vs. limit=5.0 2023-05-15 20:38:32,347 INFO [finetune.py:992] (0/2) Epoch 3, batch 850, loss[loss=0.2021, simple_loss=0.2917, pruned_loss=0.05624, over 11653.00 frames. ], tot_loss[loss=0.178, simple_loss=0.2655, pruned_loss=0.0452, over 2348431.30 frames. ], batch size: 48, lr: 4.92e-03, grad_scale: 8.0 2023-05-15 20:38:33,086 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=125319.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 20:38:37,376 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([6.0334, 5.9970, 5.7200, 5.2597, 5.0333, 5.9481, 5.5059, 5.2458], device='cuda:0'), covar=tensor([0.0764, 0.0973, 0.0713, 0.1622, 0.0655, 0.0729, 0.1603, 0.1166], device='cuda:0'), in_proj_covar=tensor([0.0553, 0.0501, 0.0459, 0.0570, 0.0371, 0.0634, 0.0693, 0.0515], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0004, 0.0004, 0.0003], device='cuda:0') 2023-05-15 20:38:49,458 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=125342.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 20:38:59,372 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=125356.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 20:39:06,980 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=125367.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 20:39:07,611 INFO [finetune.py:992] (0/2) Epoch 3, batch 900, loss[loss=0.1826, simple_loss=0.2738, pruned_loss=0.04566, over 12360.00 frames. ], tot_loss[loss=0.1773, simple_loss=0.2649, pruned_loss=0.04488, over 2357871.45 frames. ], batch size: 38, lr: 4.92e-03, grad_scale: 8.0 2023-05-15 20:39:13,546 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=125376.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 20:39:14,331 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.3134, 3.1797, 4.6362, 2.4209, 2.5492, 3.5835, 3.0086, 3.7412], device='cuda:0'), covar=tensor([0.0431, 0.1050, 0.0337, 0.1271, 0.1996, 0.1328, 0.1352, 0.0936], device='cuda:0'), in_proj_covar=tensor([0.0227, 0.0226, 0.0223, 0.0174, 0.0232, 0.0269, 0.0220, 0.0241], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:0') 2023-05-15 20:39:42,149 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.949e+02 2.973e+02 3.414e+02 3.855e+02 6.034e+02, threshold=6.827e+02, percent-clipped=0.0 2023-05-15 20:39:43,461 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.36 vs. limit=2.0 2023-05-15 20:39:43,756 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=125417.0, num_to_drop=1, layers_to_drop={3} 2023-05-15 20:39:44,254 INFO [finetune.py:992] (0/2) Epoch 3, batch 950, loss[loss=0.1811, simple_loss=0.274, pruned_loss=0.04406, over 12296.00 frames. ], tot_loss[loss=0.1766, simple_loss=0.2643, pruned_loss=0.04443, over 2366770.94 frames. ], batch size: 33, lr: 4.92e-03, grad_scale: 8.0 2023-05-15 20:39:46,445 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=125421.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 20:39:58,390 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=125437.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 20:40:09,874 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.4461, 4.2664, 4.2940, 4.4971, 3.1974, 4.1794, 3.0195, 4.0566], device='cuda:0'), covar=tensor([0.1561, 0.0594, 0.0729, 0.0529, 0.1060, 0.0536, 0.1503, 0.1482], device='cuda:0'), in_proj_covar=tensor([0.0224, 0.0259, 0.0290, 0.0344, 0.0236, 0.0234, 0.0257, 0.0380], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-15 20:40:18,580 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.62 vs. limit=2.0 2023-05-15 20:40:20,372 INFO [finetune.py:992] (0/2) Epoch 3, batch 1000, loss[loss=0.1738, simple_loss=0.2751, pruned_loss=0.03624, over 12354.00 frames. ], tot_loss[loss=0.1766, simple_loss=0.2644, pruned_loss=0.04445, over 2369976.21 frames. ], batch size: 36, lr: 4.92e-03, grad_scale: 4.0 2023-05-15 20:40:26,198 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=125476.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 20:40:29,065 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.6335, 3.6774, 3.3444, 3.4260, 3.0412, 2.7939, 3.6240, 2.4413], device='cuda:0'), covar=tensor([0.0288, 0.0117, 0.0123, 0.0113, 0.0311, 0.0264, 0.0096, 0.0378], device='cuda:0'), in_proj_covar=tensor([0.0178, 0.0149, 0.0140, 0.0170, 0.0191, 0.0184, 0.0149, 0.0192], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], device='cuda:0') 2023-05-15 20:40:30,400 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=125482.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 20:40:31,754 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=125484.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 20:40:44,632 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.4362, 2.4077, 3.2633, 4.3567, 2.3449, 4.4590, 4.3085, 4.6021], device='cuda:0'), covar=tensor([0.0149, 0.1155, 0.0388, 0.0136, 0.1232, 0.0183, 0.0165, 0.0074], device='cuda:0'), in_proj_covar=tensor([0.0118, 0.0202, 0.0181, 0.0110, 0.0185, 0.0167, 0.0164, 0.0115], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-15 20:40:54,381 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.710e+02 2.897e+02 3.324e+02 3.887e+02 6.575e+02, threshold=6.648e+02, percent-clipped=0.0 2023-05-15 20:40:55,818 INFO [finetune.py:992] (0/2) Epoch 3, batch 1050, loss[loss=0.1749, simple_loss=0.2587, pruned_loss=0.04557, over 12172.00 frames. ], tot_loss[loss=0.1764, simple_loss=0.2641, pruned_loss=0.04438, over 2371395.72 frames. ], batch size: 31, lr: 4.92e-03, grad_scale: 4.0 2023-05-15 20:41:00,156 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=125524.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 20:41:31,871 INFO [finetune.py:992] (0/2) Epoch 3, batch 1100, loss[loss=0.2077, simple_loss=0.2919, pruned_loss=0.06177, over 12066.00 frames. ], tot_loss[loss=0.1775, simple_loss=0.2648, pruned_loss=0.04507, over 2353767.39 frames. ], batch size: 40, lr: 4.92e-03, grad_scale: 4.0 2023-05-15 20:41:41,237 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.6424, 2.5177, 3.2688, 4.5221, 2.4259, 4.6699, 4.5142, 4.7990], device='cuda:0'), covar=tensor([0.0111, 0.1232, 0.0440, 0.0133, 0.1310, 0.0152, 0.0143, 0.0064], device='cuda:0'), in_proj_covar=tensor([0.0119, 0.0202, 0.0182, 0.0111, 0.0185, 0.0167, 0.0164, 0.0115], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-15 20:41:51,077 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.0027, 3.4956, 5.2599, 2.6723, 2.9266, 3.8423, 3.4358, 4.0026], device='cuda:0'), covar=tensor([0.0325, 0.1026, 0.0202, 0.1221, 0.1892, 0.1275, 0.1203, 0.0883], device='cuda:0'), in_proj_covar=tensor([0.0225, 0.0224, 0.0222, 0.0173, 0.0230, 0.0268, 0.0218, 0.0240], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:0') 2023-05-15 20:41:57,557 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=125603.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 20:42:06,360 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.695e+02 3.110e+02 3.758e+02 4.115e+02 6.117e+02, threshold=7.515e+02, percent-clipped=0.0 2023-05-15 20:42:07,853 INFO [finetune.py:992] (0/2) Epoch 3, batch 1150, loss[loss=0.1943, simple_loss=0.2867, pruned_loss=0.05093, over 12136.00 frames. ], tot_loss[loss=0.1772, simple_loss=0.2647, pruned_loss=0.04487, over 2361058.86 frames. ], batch size: 38, lr: 4.92e-03, grad_scale: 4.0 2023-05-15 20:42:20,741 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.2301, 4.8087, 5.1252, 4.5565, 4.8047, 4.6182, 5.2038, 4.8954], device='cuda:0'), covar=tensor([0.0258, 0.0354, 0.0299, 0.0231, 0.0320, 0.0296, 0.0244, 0.0293], device='cuda:0'), in_proj_covar=tensor([0.0224, 0.0231, 0.0248, 0.0223, 0.0225, 0.0223, 0.0203, 0.0192], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-15 20:42:40,320 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=125664.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 20:42:42,962 INFO [finetune.py:992] (0/2) Epoch 3, batch 1200, loss[loss=0.1781, simple_loss=0.2665, pruned_loss=0.04485, over 12367.00 frames. ], tot_loss[loss=0.1775, simple_loss=0.2647, pruned_loss=0.04519, over 2352981.04 frames. ], batch size: 35, lr: 4.92e-03, grad_scale: 4.0 2023-05-15 20:42:53,326 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.6050, 2.5720, 4.3720, 4.6855, 2.8189, 2.6840, 2.7609, 1.9864], device='cuda:0'), covar=tensor([0.1383, 0.3064, 0.0435, 0.0322, 0.1071, 0.1896, 0.2483, 0.3868], device='cuda:0'), in_proj_covar=tensor([0.0264, 0.0361, 0.0256, 0.0277, 0.0242, 0.0271, 0.0340, 0.0348], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-15 20:43:06,715 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.0702, 4.6754, 4.6986, 4.8581, 4.6430, 4.9273, 4.7606, 2.7768], device='cuda:0'), covar=tensor([0.0107, 0.0057, 0.0091, 0.0055, 0.0044, 0.0085, 0.0079, 0.0672], device='cuda:0'), in_proj_covar=tensor([0.0061, 0.0070, 0.0074, 0.0067, 0.0055, 0.0083, 0.0073, 0.0088], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-15 20:43:15,055 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=125712.0, num_to_drop=1, layers_to_drop={1} 2023-05-15 20:43:18,353 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.879e+02 2.924e+02 3.346e+02 4.080e+02 7.279e+02, threshold=6.692e+02, percent-clipped=0.0 2023-05-15 20:43:18,966 INFO [finetune.py:992] (0/2) Epoch 3, batch 1250, loss[loss=0.2078, simple_loss=0.2839, pruned_loss=0.06583, over 12281.00 frames. ], tot_loss[loss=0.1774, simple_loss=0.2646, pruned_loss=0.04515, over 2357383.62 frames. ], batch size: 34, lr: 4.92e-03, grad_scale: 4.0 2023-05-15 20:43:24,726 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=125725.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 20:43:26,795 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.7239, 2.5284, 3.4158, 4.6396, 2.6500, 4.7142, 4.5782, 4.8391], device='cuda:0'), covar=tensor([0.0104, 0.1158, 0.0358, 0.0104, 0.1126, 0.0138, 0.0119, 0.0057], device='cuda:0'), in_proj_covar=tensor([0.0118, 0.0201, 0.0180, 0.0110, 0.0184, 0.0166, 0.0163, 0.0115], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-15 20:43:29,621 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=125732.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 20:43:41,298 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=4.62 vs. limit=5.0 2023-05-15 20:43:55,052 INFO [finetune.py:992] (0/2) Epoch 3, batch 1300, loss[loss=0.1611, simple_loss=0.2394, pruned_loss=0.04137, over 12188.00 frames. ], tot_loss[loss=0.1772, simple_loss=0.2644, pruned_loss=0.04495, over 2359089.94 frames. ], batch size: 29, lr: 4.92e-03, grad_scale: 4.0 2023-05-15 20:44:01,533 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=125777.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 20:44:06,485 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=125784.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 20:44:07,885 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=125786.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 20:44:29,861 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.093e+02 3.119e+02 3.684e+02 4.625e+02 9.118e+02, threshold=7.367e+02, percent-clipped=4.0 2023-05-15 20:44:30,595 INFO [finetune.py:992] (0/2) Epoch 3, batch 1350, loss[loss=0.1574, simple_loss=0.2479, pruned_loss=0.03347, over 12247.00 frames. ], tot_loss[loss=0.1773, simple_loss=0.2647, pruned_loss=0.045, over 2360053.89 frames. ], batch size: 32, lr: 4.92e-03, grad_scale: 4.0 2023-05-15 20:44:40,908 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=125832.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 20:45:06,290 INFO [finetune.py:992] (0/2) Epoch 3, batch 1400, loss[loss=0.2059, simple_loss=0.2921, pruned_loss=0.05984, over 11608.00 frames. ], tot_loss[loss=0.1778, simple_loss=0.2653, pruned_loss=0.04516, over 2355519.93 frames. ], batch size: 48, lr: 4.92e-03, grad_scale: 4.0 2023-05-15 20:45:41,508 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.155e+02 2.857e+02 3.363e+02 4.209e+02 6.773e+02, threshold=6.726e+02, percent-clipped=0.0 2023-05-15 20:45:42,193 INFO [finetune.py:992] (0/2) Epoch 3, batch 1450, loss[loss=0.1982, simple_loss=0.2952, pruned_loss=0.05061, over 12354.00 frames. ], tot_loss[loss=0.1767, simple_loss=0.2644, pruned_loss=0.04454, over 2364490.76 frames. ], batch size: 36, lr: 4.92e-03, grad_scale: 4.0 2023-05-15 20:46:11,337 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=125959.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 20:46:17,785 INFO [finetune.py:992] (0/2) Epoch 3, batch 1500, loss[loss=0.1469, simple_loss=0.2328, pruned_loss=0.03049, over 12122.00 frames. ], tot_loss[loss=0.1751, simple_loss=0.2632, pruned_loss=0.04353, over 2372351.32 frames. ], batch size: 30, lr: 4.92e-03, grad_scale: 4.0 2023-05-15 20:46:21,440 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.1373, 5.0870, 4.9392, 4.9644, 4.6448, 5.1104, 5.0931, 5.3986], device='cuda:0'), covar=tensor([0.0215, 0.0130, 0.0172, 0.0279, 0.0698, 0.0300, 0.0142, 0.0153], device='cuda:0'), in_proj_covar=tensor([0.0175, 0.0176, 0.0177, 0.0220, 0.0223, 0.0193, 0.0163, 0.0211], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0005, 0.0005, 0.0005, 0.0004, 0.0005], device='cuda:0') 2023-05-15 20:46:39,566 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=125998.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 20:46:41,147 INFO [checkpoint.py:75] (0/2) Saving checkpoint to pruned_transducer_stateless7/exp_giga_finetune/checkpoint-26000.pt 2023-05-15 20:46:52,449 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=126012.0, num_to_drop=1, layers_to_drop={2} 2023-05-15 20:46:55,826 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.152e+02 2.827e+02 3.591e+02 4.249e+02 9.010e+02, threshold=7.182e+02, percent-clipped=2.0 2023-05-15 20:46:56,543 INFO [finetune.py:992] (0/2) Epoch 3, batch 1550, loss[loss=0.1908, simple_loss=0.2787, pruned_loss=0.05144, over 11722.00 frames. ], tot_loss[loss=0.1754, simple_loss=0.2634, pruned_loss=0.0437, over 2374227.34 frames. ], batch size: 48, lr: 4.92e-03, grad_scale: 4.0 2023-05-15 20:47:07,369 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=126032.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 20:47:26,519 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=126059.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 20:47:27,038 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=126060.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 20:47:31,583 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.6762, 2.7833, 4.7114, 5.1325, 3.5084, 2.7923, 2.9727, 2.1318], device='cuda:0'), covar=tensor([0.1395, 0.3041, 0.0384, 0.0297, 0.0832, 0.1952, 0.2565, 0.4154], device='cuda:0'), in_proj_covar=tensor([0.0265, 0.0361, 0.0256, 0.0280, 0.0243, 0.0272, 0.0341, 0.0348], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-15 20:47:32,751 INFO [finetune.py:992] (0/2) Epoch 3, batch 1600, loss[loss=0.1838, simple_loss=0.2841, pruned_loss=0.04175, over 12366.00 frames. ], tot_loss[loss=0.1759, simple_loss=0.2638, pruned_loss=0.04401, over 2368624.65 frames. ], batch size: 36, lr: 4.92e-03, grad_scale: 8.0 2023-05-15 20:47:39,326 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=126077.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 20:47:40,788 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.2410, 5.2070, 5.0302, 4.5876, 4.6482, 5.1772, 4.8008, 4.6282], device='cuda:0'), covar=tensor([0.0745, 0.0896, 0.0675, 0.1358, 0.1130, 0.0767, 0.1460, 0.1017], device='cuda:0'), in_proj_covar=tensor([0.0568, 0.0506, 0.0469, 0.0582, 0.0378, 0.0652, 0.0709, 0.0525], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0004, 0.0004, 0.0003], device='cuda:0') 2023-05-15 20:47:41,406 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=126080.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 20:47:42,147 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=126081.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 20:48:07,484 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.864e+02 2.888e+02 3.272e+02 4.126e+02 7.200e+02, threshold=6.545e+02, percent-clipped=1.0 2023-05-15 20:48:08,163 INFO [finetune.py:992] (0/2) Epoch 3, batch 1650, loss[loss=0.1798, simple_loss=0.2745, pruned_loss=0.0425, over 12148.00 frames. ], tot_loss[loss=0.1751, simple_loss=0.2631, pruned_loss=0.04358, over 2373160.03 frames. ], batch size: 34, lr: 4.92e-03, grad_scale: 8.0 2023-05-15 20:48:12,510 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.2529, 5.1659, 5.0252, 5.1389, 4.6923, 5.2301, 5.1948, 5.4795], device='cuda:0'), covar=tensor([0.0161, 0.0126, 0.0171, 0.0241, 0.0677, 0.0213, 0.0130, 0.0125], device='cuda:0'), in_proj_covar=tensor([0.0176, 0.0177, 0.0179, 0.0223, 0.0224, 0.0195, 0.0165, 0.0212], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0005, 0.0005, 0.0005, 0.0004, 0.0005], device='cuda:0') 2023-05-15 20:48:13,763 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=126125.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 20:48:28,842 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.0446, 5.9112, 5.4328, 5.3964, 5.9404, 5.2966, 5.6140, 5.4668], device='cuda:0'), covar=tensor([0.1319, 0.0875, 0.0925, 0.1962, 0.0915, 0.2261, 0.1428, 0.1109], device='cuda:0'), in_proj_covar=tensor([0.0325, 0.0446, 0.0355, 0.0408, 0.0425, 0.0409, 0.0360, 0.0348], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-15 20:48:44,856 INFO [finetune.py:992] (0/2) Epoch 3, batch 1700, loss[loss=0.1603, simple_loss=0.2514, pruned_loss=0.03462, over 12153.00 frames. ], tot_loss[loss=0.1753, simple_loss=0.2629, pruned_loss=0.04382, over 2372264.49 frames. ], batch size: 36, lr: 4.92e-03, grad_scale: 8.0 2023-05-15 20:49:14,826 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.3245, 4.7234, 2.8526, 2.3618, 4.0083, 2.4835, 4.1095, 3.2808], device='cuda:0'), covar=tensor([0.0584, 0.0443, 0.1064, 0.1681, 0.0253, 0.1309, 0.0362, 0.0706], device='cuda:0'), in_proj_covar=tensor([0.0187, 0.0243, 0.0173, 0.0196, 0.0134, 0.0178, 0.0188, 0.0172], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-15 20:49:20,340 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.950e+02 3.003e+02 3.423e+02 3.859e+02 6.319e+02, threshold=6.845e+02, percent-clipped=0.0 2023-05-15 20:49:21,065 INFO [finetune.py:992] (0/2) Epoch 3, batch 1750, loss[loss=0.1739, simple_loss=0.2646, pruned_loss=0.04165, over 12285.00 frames. ], tot_loss[loss=0.1746, simple_loss=0.2621, pruned_loss=0.04354, over 2372331.49 frames. ], batch size: 33, lr: 4.92e-03, grad_scale: 8.0 2023-05-15 20:49:42,231 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([6.1396, 6.0972, 5.9452, 5.4196, 5.2080, 6.0911, 5.6177, 5.4613], device='cuda:0'), covar=tensor([0.0671, 0.0913, 0.0621, 0.1382, 0.0547, 0.0677, 0.1385, 0.0933], device='cuda:0'), in_proj_covar=tensor([0.0568, 0.0507, 0.0467, 0.0579, 0.0377, 0.0652, 0.0707, 0.0524], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0004, 0.0004, 0.0003], device='cuda:0') 2023-05-15 20:49:47,446 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.1922, 3.6625, 3.3605, 3.2011, 2.8931, 2.6552, 3.5612, 2.1314], device='cuda:0'), covar=tensor([0.0376, 0.0102, 0.0132, 0.0158, 0.0295, 0.0309, 0.0110, 0.0444], device='cuda:0'), in_proj_covar=tensor([0.0180, 0.0152, 0.0143, 0.0172, 0.0193, 0.0187, 0.0151, 0.0193], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], device='cuda:0') 2023-05-15 20:49:50,132 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=126259.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 20:49:56,497 INFO [finetune.py:992] (0/2) Epoch 3, batch 1800, loss[loss=0.2522, simple_loss=0.3307, pruned_loss=0.0868, over 8518.00 frames. ], tot_loss[loss=0.175, simple_loss=0.2625, pruned_loss=0.04371, over 2374568.34 frames. ], batch size: 98, lr: 4.92e-03, grad_scale: 8.0 2023-05-15 20:50:24,801 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=126307.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 20:50:31,951 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.025e+02 2.885e+02 3.436e+02 3.955e+02 8.404e+02, threshold=6.872e+02, percent-clipped=1.0 2023-05-15 20:50:32,657 INFO [finetune.py:992] (0/2) Epoch 3, batch 1850, loss[loss=0.15, simple_loss=0.2327, pruned_loss=0.03363, over 12122.00 frames. ], tot_loss[loss=0.1748, simple_loss=0.2626, pruned_loss=0.0435, over 2376073.95 frames. ], batch size: 30, lr: 4.92e-03, grad_scale: 8.0 2023-05-15 20:50:58,758 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=126354.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 20:51:08,433 INFO [finetune.py:992] (0/2) Epoch 3, batch 1900, loss[loss=0.1625, simple_loss=0.2485, pruned_loss=0.03827, over 12360.00 frames. ], tot_loss[loss=0.1743, simple_loss=0.2617, pruned_loss=0.0435, over 2381385.20 frames. ], batch size: 35, lr: 4.92e-03, grad_scale: 8.0 2023-05-15 20:51:17,165 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=126380.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 20:51:17,829 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=126381.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 20:51:43,430 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.785e+02 2.796e+02 3.344e+02 4.119e+02 6.290e+02, threshold=6.688e+02, percent-clipped=0.0 2023-05-15 20:51:44,199 INFO [finetune.py:992] (0/2) Epoch 3, batch 1950, loss[loss=0.1739, simple_loss=0.2591, pruned_loss=0.04433, over 12364.00 frames. ], tot_loss[loss=0.1754, simple_loss=0.2629, pruned_loss=0.04393, over 2385565.99 frames. ], batch size: 38, lr: 4.92e-03, grad_scale: 8.0 2023-05-15 20:51:52,633 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=126429.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 20:52:00,641 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.6080, 2.6201, 4.1072, 4.3576, 2.8750, 2.6082, 2.7796, 2.1373], device='cuda:0'), covar=tensor([0.1366, 0.2933, 0.0534, 0.0438, 0.1119, 0.1973, 0.2567, 0.3683], device='cuda:0'), in_proj_covar=tensor([0.0268, 0.0367, 0.0259, 0.0285, 0.0247, 0.0275, 0.0346, 0.0353], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-15 20:52:01,303 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=126441.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 20:52:03,074 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=4.78 vs. limit=5.0 2023-05-15 20:52:07,602 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=126450.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 20:52:21,045 INFO [finetune.py:992] (0/2) Epoch 3, batch 2000, loss[loss=0.1984, simple_loss=0.2835, pruned_loss=0.0566, over 10347.00 frames. ], tot_loss[loss=0.1751, simple_loss=0.2625, pruned_loss=0.04387, over 2383266.52 frames. ], batch size: 68, lr: 4.92e-03, grad_scale: 8.0 2023-05-15 20:52:51,838 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=126511.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 20:52:55,392 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.4049, 1.9646, 3.2088, 4.2692, 2.2718, 4.3882, 4.3155, 4.5520], device='cuda:0'), covar=tensor([0.0131, 0.1392, 0.0417, 0.0134, 0.1219, 0.0174, 0.0162, 0.0072], device='cuda:0'), in_proj_covar=tensor([0.0117, 0.0198, 0.0179, 0.0109, 0.0181, 0.0167, 0.0161, 0.0113], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-15 20:52:55,870 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.690e+02 2.900e+02 3.495e+02 4.173e+02 1.166e+03, threshold=6.989e+02, percent-clipped=2.0 2023-05-15 20:52:56,572 INFO [finetune.py:992] (0/2) Epoch 3, batch 2050, loss[loss=0.1894, simple_loss=0.2851, pruned_loss=0.04683, over 11638.00 frames. ], tot_loss[loss=0.1747, simple_loss=0.2622, pruned_loss=0.04364, over 2383385.13 frames. ], batch size: 48, lr: 4.92e-03, grad_scale: 8.0 2023-05-15 20:53:31,066 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.44 vs. limit=2.0 2023-05-15 20:53:32,059 INFO [finetune.py:992] (0/2) Epoch 3, batch 2100, loss[loss=0.1811, simple_loss=0.2671, pruned_loss=0.04749, over 12305.00 frames. ], tot_loss[loss=0.174, simple_loss=0.2613, pruned_loss=0.04335, over 2385562.74 frames. ], batch size: 34, lr: 4.91e-03, grad_scale: 8.0 2023-05-15 20:54:08,263 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.812e+02 2.866e+02 3.267e+02 3.861e+02 7.719e+02, threshold=6.534e+02, percent-clipped=2.0 2023-05-15 20:54:08,945 INFO [finetune.py:992] (0/2) Epoch 3, batch 2150, loss[loss=0.201, simple_loss=0.2836, pruned_loss=0.05918, over 11674.00 frames. ], tot_loss[loss=0.1736, simple_loss=0.261, pruned_loss=0.04316, over 2387744.72 frames. ], batch size: 48, lr: 4.91e-03, grad_scale: 4.0 2023-05-15 20:54:20,317 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.8268, 5.5281, 5.1173, 5.0657, 5.6494, 5.0397, 5.1163, 5.0800], device='cuda:0'), covar=tensor([0.1323, 0.1018, 0.0979, 0.1951, 0.0937, 0.1906, 0.1880, 0.1231], device='cuda:0'), in_proj_covar=tensor([0.0326, 0.0446, 0.0352, 0.0404, 0.0421, 0.0406, 0.0361, 0.0346], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-15 20:54:21,073 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([6.0052, 6.0617, 5.8380, 5.3476, 5.1994, 5.9689, 5.5648, 5.3131], device='cuda:0'), covar=tensor([0.0716, 0.0748, 0.0573, 0.1265, 0.0581, 0.0626, 0.1233, 0.0986], device='cuda:0'), in_proj_covar=tensor([0.0569, 0.0502, 0.0465, 0.0579, 0.0376, 0.0650, 0.0708, 0.0524], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0004, 0.0004, 0.0003], device='cuda:0') 2023-05-15 20:54:22,200 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.64 vs. limit=2.0 2023-05-15 20:54:34,735 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=126654.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 20:54:44,576 INFO [finetune.py:992] (0/2) Epoch 3, batch 2200, loss[loss=0.1607, simple_loss=0.252, pruned_loss=0.03474, over 12364.00 frames. ], tot_loss[loss=0.1739, simple_loss=0.2613, pruned_loss=0.04325, over 2381198.46 frames. ], batch size: 35, lr: 4.91e-03, grad_scale: 4.0 2023-05-15 20:55:08,786 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=126702.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 20:55:20,247 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.091e+02 2.895e+02 3.584e+02 4.587e+02 8.110e+02, threshold=7.168e+02, percent-clipped=5.0 2023-05-15 20:55:20,265 INFO [finetune.py:992] (0/2) Epoch 3, batch 2250, loss[loss=0.1972, simple_loss=0.2843, pruned_loss=0.05502, over 12138.00 frames. ], tot_loss[loss=0.1735, simple_loss=0.2609, pruned_loss=0.04305, over 2373848.13 frames. ], batch size: 38, lr: 4.91e-03, grad_scale: 4.0 2023-05-15 20:55:30,252 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.5667, 2.3908, 3.2878, 4.3442, 2.4887, 4.5206, 4.5543, 4.7262], device='cuda:0'), covar=tensor([0.0119, 0.1150, 0.0407, 0.0147, 0.1135, 0.0176, 0.0115, 0.0062], device='cuda:0'), in_proj_covar=tensor([0.0117, 0.0198, 0.0179, 0.0110, 0.0181, 0.0167, 0.0161, 0.0114], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-15 20:55:33,807 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=126736.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 20:55:49,568 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.7334, 3.3153, 5.0645, 2.5373, 2.7445, 3.8653, 3.1555, 3.9177], device='cuda:0'), covar=tensor([0.0382, 0.1082, 0.0219, 0.1159, 0.1905, 0.1236, 0.1369, 0.0994], device='cuda:0'), in_proj_covar=tensor([0.0228, 0.0228, 0.0227, 0.0175, 0.0233, 0.0272, 0.0222, 0.0245], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:0') 2023-05-15 20:55:56,828 INFO [finetune.py:992] (0/2) Epoch 3, batch 2300, loss[loss=0.2159, simple_loss=0.3049, pruned_loss=0.06341, over 11248.00 frames. ], tot_loss[loss=0.1744, simple_loss=0.2618, pruned_loss=0.0435, over 2375430.06 frames. ], batch size: 55, lr: 4.91e-03, grad_scale: 4.0 2023-05-15 20:56:24,175 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=126806.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 20:56:25,781 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.3634, 3.2011, 2.9975, 2.9354, 2.5505, 2.5010, 3.1682, 2.0238], device='cuda:0'), covar=tensor([0.0306, 0.0121, 0.0143, 0.0158, 0.0341, 0.0275, 0.0109, 0.0397], device='cuda:0'), in_proj_covar=tensor([0.0181, 0.0153, 0.0145, 0.0173, 0.0197, 0.0189, 0.0152, 0.0195], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], device='cuda:0') 2023-05-15 20:56:31,300 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=126816.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 20:56:32,499 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.251e+02 3.050e+02 3.373e+02 4.203e+02 1.390e+03, threshold=6.745e+02, percent-clipped=2.0 2023-05-15 20:56:32,517 INFO [finetune.py:992] (0/2) Epoch 3, batch 2350, loss[loss=0.2217, simple_loss=0.3003, pruned_loss=0.07153, over 8480.00 frames. ], tot_loss[loss=0.1737, simple_loss=0.2612, pruned_loss=0.04309, over 2377672.41 frames. ], batch size: 100, lr: 4.91e-03, grad_scale: 4.0 2023-05-15 20:56:41,123 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.4347, 5.2307, 5.3871, 5.4034, 4.9782, 4.9643, 4.7915, 5.3240], device='cuda:0'), covar=tensor([0.0610, 0.0507, 0.0620, 0.0548, 0.2047, 0.1408, 0.0514, 0.1001], device='cuda:0'), in_proj_covar=tensor([0.0476, 0.0609, 0.0522, 0.0570, 0.0748, 0.0683, 0.0497, 0.0451], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0005, 0.0004, 0.0004, 0.0005, 0.0005, 0.0004, 0.0004], device='cuda:0') 2023-05-15 20:57:07,764 INFO [finetune.py:992] (0/2) Epoch 3, batch 2400, loss[loss=0.1815, simple_loss=0.2711, pruned_loss=0.04598, over 11267.00 frames. ], tot_loss[loss=0.1749, simple_loss=0.2628, pruned_loss=0.04353, over 2365373.45 frames. ], batch size: 55, lr: 4.91e-03, grad_scale: 8.0 2023-05-15 20:57:10,924 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=2.97 vs. limit=5.0 2023-05-15 20:57:15,057 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=126877.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 20:57:44,838 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.791e+02 2.886e+02 3.409e+02 3.999e+02 2.480e+03, threshold=6.818e+02, percent-clipped=0.0 2023-05-15 20:57:44,858 INFO [finetune.py:992] (0/2) Epoch 3, batch 2450, loss[loss=0.1942, simple_loss=0.283, pruned_loss=0.05266, over 12100.00 frames. ], tot_loss[loss=0.1756, simple_loss=0.263, pruned_loss=0.04409, over 2366699.11 frames. ], batch size: 38, lr: 4.91e-03, grad_scale: 8.0 2023-05-15 20:58:20,051 INFO [finetune.py:992] (0/2) Epoch 3, batch 2500, loss[loss=0.1932, simple_loss=0.2765, pruned_loss=0.05492, over 11346.00 frames. ], tot_loss[loss=0.1762, simple_loss=0.2636, pruned_loss=0.04441, over 2368712.97 frames. ], batch size: 55, lr: 4.91e-03, grad_scale: 8.0 2023-05-15 20:58:55,962 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.205e+02 2.860e+02 3.434e+02 4.041e+02 7.879e+02, threshold=6.868e+02, percent-clipped=5.0 2023-05-15 20:58:55,980 INFO [finetune.py:992] (0/2) Epoch 3, batch 2550, loss[loss=0.1828, simple_loss=0.2771, pruned_loss=0.04424, over 11806.00 frames. ], tot_loss[loss=0.1756, simple_loss=0.2631, pruned_loss=0.04409, over 2372075.41 frames. ], batch size: 44, lr: 4.91e-03, grad_scale: 8.0 2023-05-15 20:59:09,792 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=127036.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 20:59:29,787 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-05-15 20:59:32,988 INFO [finetune.py:992] (0/2) Epoch 3, batch 2600, loss[loss=0.1828, simple_loss=0.2769, pruned_loss=0.04434, over 12275.00 frames. ], tot_loss[loss=0.1755, simple_loss=0.2629, pruned_loss=0.04405, over 2371623.28 frames. ], batch size: 37, lr: 4.91e-03, grad_scale: 8.0 2023-05-15 20:59:43,855 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.9889, 4.1943, 3.7333, 4.5472, 4.2221, 2.7458, 3.9383, 2.9789], device='cuda:0'), covar=tensor([0.0775, 0.0779, 0.1307, 0.0402, 0.0940, 0.1477, 0.0860, 0.2715], device='cuda:0'), in_proj_covar=tensor([0.0289, 0.0363, 0.0340, 0.0252, 0.0347, 0.0257, 0.0325, 0.0349], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-15 20:59:44,377 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=127084.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 20:59:47,641 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-05-15 21:00:00,116 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=127106.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 21:00:07,467 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.4742, 5.0746, 5.4601, 4.8080, 5.1261, 4.8229, 5.4544, 5.0712], device='cuda:0'), covar=tensor([0.0207, 0.0255, 0.0202, 0.0187, 0.0259, 0.0211, 0.0165, 0.0231], device='cuda:0'), in_proj_covar=tensor([0.0231, 0.0236, 0.0255, 0.0228, 0.0231, 0.0229, 0.0209, 0.0197], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-05-15 21:00:08,664 INFO [finetune.py:992] (0/2) Epoch 3, batch 2650, loss[loss=0.1873, simple_loss=0.2752, pruned_loss=0.04967, over 12182.00 frames. ], tot_loss[loss=0.1747, simple_loss=0.2624, pruned_loss=0.04346, over 2376390.20 frames. ], batch size: 35, lr: 4.91e-03, grad_scale: 4.0 2023-05-15 21:00:09,358 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.927e+02 2.855e+02 3.326e+02 4.115e+02 6.905e+02, threshold=6.652e+02, percent-clipped=2.0 2023-05-15 21:00:30,279 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-05-15 21:00:34,234 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=127154.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 21:00:44,084 INFO [finetune.py:992] (0/2) Epoch 3, batch 2700, loss[loss=0.166, simple_loss=0.2576, pruned_loss=0.0372, over 12142.00 frames. ], tot_loss[loss=0.1749, simple_loss=0.2624, pruned_loss=0.04366, over 2370627.03 frames. ], batch size: 36, lr: 4.91e-03, grad_scale: 4.0 2023-05-15 21:00:47,088 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=127172.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 21:00:52,194 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.1375, 3.7699, 3.9997, 4.3809, 2.7511, 3.5514, 2.4048, 3.8465], device='cuda:0'), covar=tensor([0.1767, 0.0917, 0.0960, 0.0595, 0.1335, 0.0849, 0.2188, 0.1451], device='cuda:0'), in_proj_covar=tensor([0.0226, 0.0264, 0.0293, 0.0348, 0.0240, 0.0239, 0.0259, 0.0386], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-15 21:00:52,561 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=4.62 vs. limit=5.0 2023-05-15 21:01:21,318 INFO [finetune.py:992] (0/2) Epoch 3, batch 2750, loss[loss=0.2216, simple_loss=0.2956, pruned_loss=0.07382, over 8024.00 frames. ], tot_loss[loss=0.1754, simple_loss=0.2629, pruned_loss=0.04391, over 2364458.91 frames. ], batch size: 98, lr: 4.91e-03, grad_scale: 4.0 2023-05-15 21:01:21,935 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.156e+02 3.036e+02 3.627e+02 4.430e+02 3.094e+03, threshold=7.254e+02, percent-clipped=6.0 2023-05-15 21:01:22,240 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.3750, 4.5062, 4.0886, 4.9463, 4.5959, 2.9120, 4.3257, 3.1869], device='cuda:0'), covar=tensor([0.0670, 0.0765, 0.1156, 0.0339, 0.0904, 0.1474, 0.0826, 0.2569], device='cuda:0'), in_proj_covar=tensor([0.0294, 0.0368, 0.0345, 0.0256, 0.0353, 0.0260, 0.0331, 0.0354], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-15 21:01:43,126 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=127249.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 21:01:56,095 INFO [finetune.py:992] (0/2) Epoch 3, batch 2800, loss[loss=0.1897, simple_loss=0.2805, pruned_loss=0.04941, over 11804.00 frames. ], tot_loss[loss=0.1765, simple_loss=0.2636, pruned_loss=0.04469, over 2358629.83 frames. ], batch size: 44, lr: 4.91e-03, grad_scale: 8.0 2023-05-15 21:02:09,023 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.4817, 4.6340, 4.0888, 4.9703, 4.5700, 2.8813, 4.2600, 3.0983], device='cuda:0'), covar=tensor([0.0699, 0.0736, 0.1380, 0.0436, 0.1106, 0.1535, 0.0973, 0.3279], device='cuda:0'), in_proj_covar=tensor([0.0292, 0.0365, 0.0343, 0.0255, 0.0351, 0.0258, 0.0328, 0.0351], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-15 21:02:22,635 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-05-15 21:02:24,718 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.2450, 4.2795, 3.9527, 4.7715, 4.4990, 2.7767, 4.1712, 2.9081], device='cuda:0'), covar=tensor([0.0803, 0.0962, 0.1347, 0.0506, 0.0979, 0.1586, 0.1103, 0.3444], device='cuda:0'), in_proj_covar=tensor([0.0293, 0.0366, 0.0344, 0.0255, 0.0352, 0.0260, 0.0330, 0.0353], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-15 21:02:26,086 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=127310.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 21:02:32,265 INFO [finetune.py:992] (0/2) Epoch 3, batch 2850, loss[loss=0.1636, simple_loss=0.2463, pruned_loss=0.04043, over 12322.00 frames. ], tot_loss[loss=0.1752, simple_loss=0.2624, pruned_loss=0.04403, over 2368449.56 frames. ], batch size: 30, lr: 4.91e-03, grad_scale: 8.0 2023-05-15 21:02:32,983 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.250e+02 2.893e+02 3.266e+02 4.085e+02 1.029e+03, threshold=6.532e+02, percent-clipped=2.0 2023-05-15 21:03:08,417 INFO [finetune.py:992] (0/2) Epoch 3, batch 2900, loss[loss=0.1818, simple_loss=0.2699, pruned_loss=0.04683, over 12307.00 frames. ], tot_loss[loss=0.1755, simple_loss=0.2623, pruned_loss=0.04435, over 2362478.67 frames. ], batch size: 34, lr: 4.91e-03, grad_scale: 8.0 2023-05-15 21:03:29,173 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.4513, 2.6274, 3.6684, 4.4840, 3.8624, 4.3874, 3.9150, 3.0655], device='cuda:0'), covar=tensor([0.0023, 0.0291, 0.0121, 0.0024, 0.0117, 0.0059, 0.0080, 0.0295], device='cuda:0'), in_proj_covar=tensor([0.0079, 0.0121, 0.0099, 0.0073, 0.0099, 0.0109, 0.0087, 0.0134], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-15 21:03:44,257 INFO [finetune.py:992] (0/2) Epoch 3, batch 2950, loss[loss=0.1591, simple_loss=0.2393, pruned_loss=0.03944, over 12120.00 frames. ], tot_loss[loss=0.1748, simple_loss=0.2619, pruned_loss=0.0439, over 2374438.97 frames. ], batch size: 30, lr: 4.91e-03, grad_scale: 8.0 2023-05-15 21:03:44,894 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.888e+02 2.714e+02 3.322e+02 3.797e+02 1.040e+03, threshold=6.645e+02, percent-clipped=2.0 2023-05-15 21:03:50,820 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=127427.0, num_to_drop=1, layers_to_drop={0} 2023-05-15 21:04:20,315 INFO [finetune.py:992] (0/2) Epoch 3, batch 3000, loss[loss=0.1546, simple_loss=0.2354, pruned_loss=0.03693, over 12197.00 frames. ], tot_loss[loss=0.1755, simple_loss=0.2626, pruned_loss=0.04424, over 2366517.54 frames. ], batch size: 29, lr: 4.91e-03, grad_scale: 8.0 2023-05-15 21:04:20,316 INFO [finetune.py:1017] (0/2) Computing validation loss 2023-05-15 21:04:38,161 INFO [finetune.py:1026] (0/2) Epoch 3, validation: loss=0.3349, simple_loss=0.4074, pruned_loss=0.1313, over 1020973.00 frames. 2023-05-15 21:04:38,162 INFO [finetune.py:1027] (0/2) Maximum memory allocated so far is 12745MB 2023-05-15 21:04:41,108 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=127472.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 21:04:52,474 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=127488.0, num_to_drop=1, layers_to_drop={2} 2023-05-15 21:05:03,833 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.0418, 2.3905, 3.6850, 3.1007, 3.4448, 3.2686, 2.4288, 3.5773], device='cuda:0'), covar=tensor([0.0112, 0.0308, 0.0112, 0.0207, 0.0150, 0.0135, 0.0342, 0.0095], device='cuda:0'), in_proj_covar=tensor([0.0159, 0.0192, 0.0169, 0.0174, 0.0189, 0.0149, 0.0183, 0.0162], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-15 21:05:13,613 INFO [finetune.py:992] (0/2) Epoch 3, batch 3050, loss[loss=0.1624, simple_loss=0.2349, pruned_loss=0.045, over 11996.00 frames. ], tot_loss[loss=0.1756, simple_loss=0.263, pruned_loss=0.04411, over 2375411.19 frames. ], batch size: 28, lr: 4.91e-03, grad_scale: 8.0 2023-05-15 21:05:14,240 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.940e+02 2.817e+02 3.491e+02 4.167e+02 6.780e+02, threshold=6.983e+02, percent-clipped=2.0 2023-05-15 21:05:15,041 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=127520.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 21:05:21,534 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=127529.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 21:05:35,317 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.4429, 4.1147, 4.3329, 4.7651, 3.3228, 4.1565, 2.6959, 4.2954], device='cuda:0'), covar=tensor([0.1540, 0.0708, 0.0725, 0.0581, 0.1009, 0.0583, 0.1715, 0.1344], device='cuda:0'), in_proj_covar=tensor([0.0223, 0.0260, 0.0289, 0.0344, 0.0236, 0.0235, 0.0255, 0.0381], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-15 21:05:49,170 INFO [finetune.py:992] (0/2) Epoch 3, batch 3100, loss[loss=0.1591, simple_loss=0.2454, pruned_loss=0.03637, over 12182.00 frames. ], tot_loss[loss=0.175, simple_loss=0.2623, pruned_loss=0.0439, over 2376524.01 frames. ], batch size: 31, lr: 4.91e-03, grad_scale: 8.0 2023-05-15 21:06:02,852 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.0825, 5.1002, 4.8627, 4.8878, 4.5328, 5.0204, 5.0353, 5.3084], device='cuda:0'), covar=tensor([0.0220, 0.0112, 0.0166, 0.0269, 0.0730, 0.0251, 0.0168, 0.0143], device='cuda:0'), in_proj_covar=tensor([0.0181, 0.0179, 0.0179, 0.0224, 0.0228, 0.0199, 0.0167, 0.0213], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0005, 0.0005, 0.0005, 0.0004, 0.0005], device='cuda:0') 2023-05-15 21:06:06,087 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=127590.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 21:06:16,984 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=127605.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 21:06:18,539 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.3461, 4.1930, 4.2535, 4.7599, 3.4163, 4.0691, 2.4950, 4.2431], device='cuda:0'), covar=tensor([0.1700, 0.0683, 0.0828, 0.0490, 0.1000, 0.0628, 0.2008, 0.1248], device='cuda:0'), in_proj_covar=tensor([0.0224, 0.0262, 0.0292, 0.0346, 0.0237, 0.0237, 0.0256, 0.0384], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-15 21:06:22,707 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.9551, 4.9532, 4.7836, 4.8004, 4.4308, 4.9682, 4.9439, 5.1804], device='cuda:0'), covar=tensor([0.0243, 0.0122, 0.0140, 0.0279, 0.0779, 0.0222, 0.0150, 0.0166], device='cuda:0'), in_proj_covar=tensor([0.0181, 0.0180, 0.0179, 0.0226, 0.0229, 0.0199, 0.0168, 0.0214], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0005, 0.0005, 0.0005, 0.0004, 0.0005], device='cuda:0') 2023-05-15 21:06:26,003 INFO [finetune.py:992] (0/2) Epoch 3, batch 3150, loss[loss=0.1747, simple_loss=0.2674, pruned_loss=0.04102, over 12351.00 frames. ], tot_loss[loss=0.1754, simple_loss=0.2629, pruned_loss=0.04398, over 2376349.17 frames. ], batch size: 35, lr: 4.91e-03, grad_scale: 4.0 2023-05-15 21:06:27,410 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.019e+02 2.701e+02 3.163e+02 3.791e+02 6.573e+02, threshold=6.327e+02, percent-clipped=0.0 2023-05-15 21:06:57,501 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=127662.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 21:07:01,613 INFO [finetune.py:992] (0/2) Epoch 3, batch 3200, loss[loss=0.1652, simple_loss=0.2668, pruned_loss=0.03176, over 12306.00 frames. ], tot_loss[loss=0.1759, simple_loss=0.2633, pruned_loss=0.04425, over 2373998.27 frames. ], batch size: 34, lr: 4.91e-03, grad_scale: 8.0 2023-05-15 21:07:37,611 INFO [finetune.py:992] (0/2) Epoch 3, batch 3250, loss[loss=0.1599, simple_loss=0.2563, pruned_loss=0.03178, over 12322.00 frames. ], tot_loss[loss=0.1758, simple_loss=0.2632, pruned_loss=0.04419, over 2379349.53 frames. ], batch size: 34, lr: 4.91e-03, grad_scale: 8.0 2023-05-15 21:07:39,049 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.130e+02 2.912e+02 3.431e+02 4.040e+02 1.283e+03, threshold=6.861e+02, percent-clipped=3.0 2023-05-15 21:07:41,425 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=127723.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 21:07:45,782 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.9062, 2.9167, 5.2799, 2.4063, 2.5373, 4.2148, 3.1065, 4.2350], device='cuda:0'), covar=tensor([0.0367, 0.1371, 0.0179, 0.1262, 0.1947, 0.1088, 0.1366, 0.0813], device='cuda:0'), in_proj_covar=tensor([0.0227, 0.0226, 0.0228, 0.0176, 0.0232, 0.0273, 0.0221, 0.0246], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:0') 2023-05-15 21:07:47,981 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.5339, 4.2718, 4.2552, 4.6578, 3.3916, 4.1793, 2.6080, 4.2756], device='cuda:0'), covar=tensor([0.1360, 0.0580, 0.0871, 0.0556, 0.0910, 0.0496, 0.1659, 0.1493], device='cuda:0'), in_proj_covar=tensor([0.0224, 0.0261, 0.0292, 0.0347, 0.0238, 0.0236, 0.0256, 0.0385], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-15 21:08:13,970 INFO [finetune.py:992] (0/2) Epoch 3, batch 3300, loss[loss=0.1779, simple_loss=0.2678, pruned_loss=0.04404, over 12363.00 frames. ], tot_loss[loss=0.1751, simple_loss=0.2626, pruned_loss=0.04379, over 2385957.94 frames. ], batch size: 35, lr: 4.91e-03, grad_scale: 8.0 2023-05-15 21:08:18,646 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.4005, 4.6073, 3.8262, 4.9845, 4.5840, 2.8030, 4.1231, 3.0030], device='cuda:0'), covar=tensor([0.0675, 0.0730, 0.1527, 0.0329, 0.0929, 0.1583, 0.0974, 0.3058], device='cuda:0'), in_proj_covar=tensor([0.0293, 0.0368, 0.0346, 0.0257, 0.0353, 0.0260, 0.0330, 0.0354], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-15 21:08:22,129 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=127779.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 21:08:24,332 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.2598, 5.1723, 5.0438, 4.6007, 4.6924, 5.1525, 4.7893, 4.6299], device='cuda:0'), covar=tensor([0.0706, 0.0886, 0.0597, 0.1344, 0.1120, 0.0731, 0.1550, 0.1029], device='cuda:0'), in_proj_covar=tensor([0.0564, 0.0501, 0.0468, 0.0579, 0.0375, 0.0649, 0.0709, 0.0523], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0004, 0.0004, 0.0003], device='cuda:0') 2023-05-15 21:08:24,994 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=127783.0, num_to_drop=1, layers_to_drop={3} 2023-05-15 21:08:50,074 INFO [finetune.py:992] (0/2) Epoch 3, batch 3350, loss[loss=0.1661, simple_loss=0.2523, pruned_loss=0.03996, over 12089.00 frames. ], tot_loss[loss=0.1754, simple_loss=0.2632, pruned_loss=0.04384, over 2380064.82 frames. ], batch size: 32, lr: 4.91e-03, grad_scale: 8.0 2023-05-15 21:08:51,512 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.857e+02 2.744e+02 3.185e+02 3.891e+02 6.795e+02, threshold=6.371e+02, percent-clipped=0.0 2023-05-15 21:09:02,582 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=3.22 vs. limit=5.0 2023-05-15 21:09:05,902 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=127840.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 21:09:13,055 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=127850.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 21:09:20,981 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.6641, 2.8160, 4.4809, 4.7799, 3.0859, 2.7993, 2.9561, 2.2460], device='cuda:0'), covar=tensor([0.1352, 0.2947, 0.0463, 0.0364, 0.0987, 0.1836, 0.2398, 0.3456], device='cuda:0'), in_proj_covar=tensor([0.0269, 0.0366, 0.0258, 0.0284, 0.0249, 0.0276, 0.0344, 0.0350], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-15 21:09:26,325 INFO [finetune.py:992] (0/2) Epoch 3, batch 3400, loss[loss=0.1587, simple_loss=0.235, pruned_loss=0.04114, over 12344.00 frames. ], tot_loss[loss=0.1748, simple_loss=0.2624, pruned_loss=0.04355, over 2384139.39 frames. ], batch size: 30, lr: 4.91e-03, grad_scale: 8.0 2023-05-15 21:09:29,164 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=127872.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 21:09:38,829 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=127885.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 21:09:46,605 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.2934, 4.8265, 2.9252, 2.5822, 4.1357, 2.5167, 4.1473, 3.4135], device='cuda:0'), covar=tensor([0.0714, 0.0494, 0.1094, 0.1430, 0.0246, 0.1275, 0.0426, 0.0702], device='cuda:0'), in_proj_covar=tensor([0.0188, 0.0244, 0.0173, 0.0195, 0.0134, 0.0177, 0.0191, 0.0171], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-15 21:09:52,900 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=127905.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 21:09:57,177 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=127911.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 21:10:01,953 INFO [finetune.py:992] (0/2) Epoch 3, batch 3450, loss[loss=0.1731, simple_loss=0.2568, pruned_loss=0.04468, over 11797.00 frames. ], tot_loss[loss=0.1746, simple_loss=0.2616, pruned_loss=0.04378, over 2381671.41 frames. ], batch size: 26, lr: 4.91e-03, grad_scale: 8.0 2023-05-15 21:10:03,381 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.048e+02 3.039e+02 3.509e+02 4.157e+02 8.157e+02, threshold=7.017e+02, percent-clipped=2.0 2023-05-15 21:10:12,762 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=127933.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 21:10:26,956 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=127953.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 21:10:28,459 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.2090, 2.1028, 3.0632, 4.0941, 2.1819, 4.2015, 4.1038, 4.3277], device='cuda:0'), covar=tensor([0.0107, 0.1220, 0.0432, 0.0128, 0.1210, 0.0185, 0.0140, 0.0060], device='cuda:0'), in_proj_covar=tensor([0.0117, 0.0202, 0.0182, 0.0112, 0.0185, 0.0171, 0.0164, 0.0117], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-15 21:10:30,142 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.40 vs. limit=2.0 2023-05-15 21:10:37,296 INFO [finetune.py:992] (0/2) Epoch 3, batch 3500, loss[loss=0.1762, simple_loss=0.2635, pruned_loss=0.04447, over 12127.00 frames. ], tot_loss[loss=0.1741, simple_loss=0.261, pruned_loss=0.04363, over 2374117.45 frames. ], batch size: 38, lr: 4.91e-03, grad_scale: 8.0 2023-05-15 21:11:00,505 INFO [checkpoint.py:75] (0/2) Saving checkpoint to pruned_transducer_stateless7/exp_giga_finetune/checkpoint-28000.pt 2023-05-15 21:11:17,006 INFO [finetune.py:992] (0/2) Epoch 3, batch 3550, loss[loss=0.147, simple_loss=0.2232, pruned_loss=0.03544, over 11792.00 frames. ], tot_loss[loss=0.1731, simple_loss=0.2597, pruned_loss=0.04324, over 2367768.06 frames. ], batch size: 26, lr: 4.91e-03, grad_scale: 8.0 2023-05-15 21:11:17,097 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=128018.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 21:11:18,461 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.920e+02 2.802e+02 3.301e+02 3.996e+02 6.263e+02, threshold=6.602e+02, percent-clipped=0.0 2023-05-15 21:11:53,309 INFO [finetune.py:992] (0/2) Epoch 3, batch 3600, loss[loss=0.1519, simple_loss=0.2408, pruned_loss=0.03151, over 12340.00 frames. ], tot_loss[loss=0.1737, simple_loss=0.2605, pruned_loss=0.04341, over 2367625.33 frames. ], batch size: 31, lr: 4.91e-03, grad_scale: 8.0 2023-05-15 21:12:04,015 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=128083.0, num_to_drop=1, layers_to_drop={2} 2023-05-15 21:12:19,402 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.7016, 2.7979, 4.6553, 4.8498, 2.8342, 2.7244, 2.8866, 2.0736], device='cuda:0'), covar=tensor([0.1310, 0.2920, 0.0372, 0.0347, 0.1107, 0.1845, 0.2523, 0.3573], device='cuda:0'), in_proj_covar=tensor([0.0268, 0.0365, 0.0257, 0.0283, 0.0249, 0.0275, 0.0342, 0.0348], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-15 21:12:21,494 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2023-05-15 21:12:28,278 INFO [finetune.py:992] (0/2) Epoch 3, batch 3650, loss[loss=0.1703, simple_loss=0.2505, pruned_loss=0.04502, over 12338.00 frames. ], tot_loss[loss=0.1727, simple_loss=0.2596, pruned_loss=0.04288, over 2371538.64 frames. ], batch size: 30, lr: 4.91e-03, grad_scale: 8.0 2023-05-15 21:12:29,749 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.114e+02 2.958e+02 3.551e+02 4.110e+02 1.232e+03, threshold=7.103e+02, percent-clipped=2.0 2023-05-15 21:12:37,715 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=128131.0, num_to_drop=1, layers_to_drop={0} 2023-05-15 21:12:40,573 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=128135.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 21:12:44,338 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.9678, 3.5677, 5.3185, 2.8844, 3.0349, 3.9948, 3.5312, 4.0638], device='cuda:0'), covar=tensor([0.0402, 0.1015, 0.0199, 0.1053, 0.1746, 0.1294, 0.1151, 0.1054], device='cuda:0'), in_proj_covar=tensor([0.0229, 0.0227, 0.0229, 0.0177, 0.0233, 0.0274, 0.0223, 0.0248], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:0') 2023-05-15 21:13:02,064 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=3.45 vs. limit=5.0 2023-05-15 21:13:05,092 INFO [finetune.py:992] (0/2) Epoch 3, batch 3700, loss[loss=0.1534, simple_loss=0.2358, pruned_loss=0.03553, over 12183.00 frames. ], tot_loss[loss=0.1733, simple_loss=0.2607, pruned_loss=0.04297, over 2371274.54 frames. ], batch size: 29, lr: 4.90e-03, grad_scale: 8.0 2023-05-15 21:13:08,967 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2023-05-15 21:13:17,189 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=128185.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 21:13:32,287 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=128206.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 21:13:34,646 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.4344, 4.6796, 3.9961, 5.0586, 4.6076, 3.0023, 4.2367, 3.0563], device='cuda:0'), covar=tensor([0.0727, 0.0760, 0.1406, 0.0312, 0.0971, 0.1508, 0.0904, 0.3018], device='cuda:0'), in_proj_covar=tensor([0.0289, 0.0362, 0.0339, 0.0252, 0.0347, 0.0256, 0.0324, 0.0348], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-15 21:13:40,823 INFO [finetune.py:992] (0/2) Epoch 3, batch 3750, loss[loss=0.2003, simple_loss=0.2882, pruned_loss=0.05624, over 12138.00 frames. ], tot_loss[loss=0.1746, simple_loss=0.2619, pruned_loss=0.04366, over 2367755.65 frames. ], batch size: 38, lr: 4.90e-03, grad_scale: 8.0 2023-05-15 21:13:42,275 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.982e+02 2.975e+02 3.523e+02 4.137e+02 1.386e+03, threshold=7.045e+02, percent-clipped=3.0 2023-05-15 21:13:47,979 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=128228.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 21:13:51,409 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=128233.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 21:13:59,611 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 2023-05-15 21:14:15,970 INFO [finetune.py:992] (0/2) Epoch 3, batch 3800, loss[loss=0.1976, simple_loss=0.2926, pruned_loss=0.05131, over 12111.00 frames. ], tot_loss[loss=0.1754, simple_loss=0.263, pruned_loss=0.04391, over 2367342.92 frames. ], batch size: 38, lr: 4.90e-03, grad_scale: 8.0 2023-05-15 21:14:38,262 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.3789, 4.7718, 2.8777, 2.7360, 4.1918, 2.6345, 4.0865, 3.2888], device='cuda:0'), covar=tensor([0.0649, 0.0603, 0.1228, 0.1458, 0.0223, 0.1253, 0.0510, 0.0796], device='cuda:0'), in_proj_covar=tensor([0.0189, 0.0248, 0.0176, 0.0198, 0.0136, 0.0178, 0.0194, 0.0174], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-15 21:14:53,009 INFO [finetune.py:992] (0/2) Epoch 3, batch 3850, loss[loss=0.144, simple_loss=0.231, pruned_loss=0.02854, over 12195.00 frames. ], tot_loss[loss=0.1746, simple_loss=0.2622, pruned_loss=0.04354, over 2376539.94 frames. ], batch size: 29, lr: 4.90e-03, grad_scale: 8.0 2023-05-15 21:14:53,167 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=128318.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 21:14:54,347 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.143e+02 3.113e+02 3.561e+02 4.163e+02 1.278e+03, threshold=7.122e+02, percent-clipped=0.0 2023-05-15 21:14:54,522 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=128320.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 21:15:27,210 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=128366.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 21:15:28,614 INFO [finetune.py:992] (0/2) Epoch 3, batch 3900, loss[loss=0.1696, simple_loss=0.2586, pruned_loss=0.04025, over 12149.00 frames. ], tot_loss[loss=0.1755, simple_loss=0.2632, pruned_loss=0.04396, over 2375309.94 frames. ], batch size: 39, lr: 4.90e-03, grad_scale: 8.0 2023-05-15 21:15:38,055 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.0833, 4.7413, 4.7641, 4.9794, 4.6765, 4.9259, 4.8399, 3.0882], device='cuda:0'), covar=tensor([0.0102, 0.0066, 0.0093, 0.0062, 0.0053, 0.0082, 0.0074, 0.0592], device='cuda:0'), in_proj_covar=tensor([0.0064, 0.0072, 0.0076, 0.0069, 0.0056, 0.0085, 0.0075, 0.0089], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-15 21:15:38,106 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=128381.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 21:15:55,350 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=2.55 vs. limit=5.0 2023-05-15 21:16:00,959 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=128413.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 21:16:04,426 INFO [finetune.py:992] (0/2) Epoch 3, batch 3950, loss[loss=0.2044, simple_loss=0.274, pruned_loss=0.06741, over 8112.00 frames. ], tot_loss[loss=0.1757, simple_loss=0.2635, pruned_loss=0.04392, over 2378623.66 frames. ], batch size: 97, lr: 4.90e-03, grad_scale: 8.0 2023-05-15 21:16:05,834 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.091e+02 2.899e+02 3.271e+02 3.921e+02 6.752e+02, threshold=6.543e+02, percent-clipped=1.0 2023-05-15 21:16:16,739 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=128435.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 21:16:37,833 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.53 vs. limit=2.0 2023-05-15 21:16:41,042 INFO [finetune.py:992] (0/2) Epoch 3, batch 4000, loss[loss=0.1744, simple_loss=0.2597, pruned_loss=0.04458, over 12348.00 frames. ], tot_loss[loss=0.1756, simple_loss=0.2634, pruned_loss=0.04395, over 2373863.92 frames. ], batch size: 31, lr: 4.90e-03, grad_scale: 8.0 2023-05-15 21:16:45,559 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=128474.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 21:16:51,896 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=128483.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 21:17:08,230 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=128506.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 21:17:16,403 INFO [finetune.py:992] (0/2) Epoch 3, batch 4050, loss[loss=0.1539, simple_loss=0.2372, pruned_loss=0.03536, over 12255.00 frames. ], tot_loss[loss=0.1765, simple_loss=0.2641, pruned_loss=0.04448, over 2375045.26 frames. ], batch size: 32, lr: 4.90e-03, grad_scale: 8.0 2023-05-15 21:17:17,795 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.064e+02 2.990e+02 3.571e+02 4.197e+02 7.894e+02, threshold=7.143e+02, percent-clipped=2.0 2023-05-15 21:17:18,075 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.3199, 4.2971, 4.2510, 4.5742, 3.3963, 4.0430, 2.8373, 4.2136], device='cuda:0'), covar=tensor([0.1419, 0.0499, 0.0711, 0.0440, 0.0841, 0.0527, 0.1473, 0.1090], device='cuda:0'), in_proj_covar=tensor([0.0221, 0.0259, 0.0290, 0.0346, 0.0235, 0.0235, 0.0253, 0.0383], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-15 21:17:23,680 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=128528.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 21:17:29,951 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.4709, 2.7019, 3.5633, 4.4082, 3.8787, 4.4524, 3.7965, 3.1792], device='cuda:0'), covar=tensor([0.0026, 0.0318, 0.0143, 0.0043, 0.0114, 0.0063, 0.0097, 0.0311], device='cuda:0'), in_proj_covar=tensor([0.0082, 0.0124, 0.0102, 0.0076, 0.0101, 0.0111, 0.0090, 0.0138], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-15 21:17:37,996 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=4.58 vs. limit=5.0 2023-05-15 21:17:41,880 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=128554.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 21:17:51,683 INFO [finetune.py:992] (0/2) Epoch 3, batch 4100, loss[loss=0.225, simple_loss=0.3067, pruned_loss=0.0716, over 11536.00 frames. ], tot_loss[loss=0.1758, simple_loss=0.2632, pruned_loss=0.04416, over 2377196.01 frames. ], batch size: 48, lr: 4.90e-03, grad_scale: 8.0 2023-05-15 21:17:57,528 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=128576.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 21:18:01,170 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.1491, 4.7114, 4.9205, 5.0567, 4.6991, 5.0346, 4.9191, 2.9866], device='cuda:0'), covar=tensor([0.0097, 0.0070, 0.0074, 0.0054, 0.0050, 0.0081, 0.0069, 0.0571], device='cuda:0'), in_proj_covar=tensor([0.0063, 0.0072, 0.0075, 0.0068, 0.0056, 0.0084, 0.0075, 0.0088], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-15 21:18:29,039 INFO [finetune.py:992] (0/2) Epoch 3, batch 4150, loss[loss=0.168, simple_loss=0.2597, pruned_loss=0.03815, over 12361.00 frames. ], tot_loss[loss=0.1753, simple_loss=0.263, pruned_loss=0.0438, over 2371774.46 frames. ], batch size: 36, lr: 4.90e-03, grad_scale: 4.0 2023-05-15 21:18:31,152 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.169e+02 2.865e+02 3.448e+02 4.142e+02 7.233e+02, threshold=6.896e+02, percent-clipped=1.0 2023-05-15 21:18:53,599 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 2023-05-15 21:19:04,689 INFO [finetune.py:992] (0/2) Epoch 3, batch 4200, loss[loss=0.1592, simple_loss=0.2422, pruned_loss=0.03812, over 12346.00 frames. ], tot_loss[loss=0.175, simple_loss=0.2626, pruned_loss=0.04373, over 2375509.20 frames. ], batch size: 30, lr: 4.90e-03, grad_scale: 4.0 2023-05-15 21:19:10,573 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=128676.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 21:19:39,284 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.7231, 2.9646, 4.8061, 5.1386, 3.0387, 2.9277, 2.9933, 2.1778], device='cuda:0'), covar=tensor([0.1356, 0.2634, 0.0383, 0.0288, 0.1059, 0.1812, 0.2437, 0.3676], device='cuda:0'), in_proj_covar=tensor([0.0268, 0.0366, 0.0260, 0.0283, 0.0249, 0.0276, 0.0344, 0.0350], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-15 21:19:40,410 INFO [finetune.py:992] (0/2) Epoch 3, batch 4250, loss[loss=0.2231, simple_loss=0.2924, pruned_loss=0.07691, over 7951.00 frames. ], tot_loss[loss=0.1754, simple_loss=0.2628, pruned_loss=0.04396, over 2365963.02 frames. ], batch size: 98, lr: 4.90e-03, grad_scale: 4.0 2023-05-15 21:19:42,608 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.169e+02 2.988e+02 3.450e+02 4.097e+02 6.207e+02, threshold=6.901e+02, percent-clipped=0.0 2023-05-15 21:20:02,672 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.1811, 6.0583, 5.5842, 5.6438, 6.1683, 5.6212, 5.7680, 5.5908], device='cuda:0'), covar=tensor([0.1193, 0.0756, 0.0886, 0.1509, 0.0822, 0.1745, 0.1311, 0.1103], device='cuda:0'), in_proj_covar=tensor([0.0327, 0.0450, 0.0354, 0.0403, 0.0432, 0.0404, 0.0367, 0.0350], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-15 21:20:17,476 INFO [finetune.py:992] (0/2) Epoch 3, batch 4300, loss[loss=0.168, simple_loss=0.2576, pruned_loss=0.03923, over 12205.00 frames. ], tot_loss[loss=0.1757, simple_loss=0.2632, pruned_loss=0.04406, over 2370271.73 frames. ], batch size: 35, lr: 4.90e-03, grad_scale: 4.0 2023-05-15 21:20:18,272 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=128769.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 21:20:26,809 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=128781.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 21:20:52,967 INFO [finetune.py:992] (0/2) Epoch 3, batch 4350, loss[loss=0.1995, simple_loss=0.2895, pruned_loss=0.05474, over 12048.00 frames. ], tot_loss[loss=0.1758, simple_loss=0.2638, pruned_loss=0.04395, over 2382770.39 frames. ], batch size: 37, lr: 4.90e-03, grad_scale: 4.0 2023-05-15 21:20:55,187 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.239e+02 2.702e+02 3.202e+02 3.965e+02 7.883e+02, threshold=6.404e+02, percent-clipped=2.0 2023-05-15 21:21:10,256 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=128842.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 21:21:28,675 INFO [finetune.py:992] (0/2) Epoch 3, batch 4400, loss[loss=0.1821, simple_loss=0.2707, pruned_loss=0.04672, over 12347.00 frames. ], tot_loss[loss=0.1752, simple_loss=0.2632, pruned_loss=0.04358, over 2382214.99 frames. ], batch size: 35, lr: 4.90e-03, grad_scale: 8.0 2023-05-15 21:21:33,895 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.6657, 4.7696, 4.2804, 5.1659, 4.8369, 3.1942, 4.4628, 3.3826], device='cuda:0'), covar=tensor([0.0567, 0.0777, 0.1216, 0.0340, 0.0914, 0.1301, 0.0931, 0.2577], device='cuda:0'), in_proj_covar=tensor([0.0293, 0.0365, 0.0343, 0.0257, 0.0351, 0.0259, 0.0328, 0.0352], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-15 21:21:45,187 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=128890.0, num_to_drop=1, layers_to_drop={1} 2023-05-15 21:21:56,669 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.6960, 2.9053, 4.7788, 4.9095, 3.2130, 2.7731, 2.9995, 2.1523], device='cuda:0'), covar=tensor([0.1299, 0.2702, 0.0350, 0.0325, 0.0935, 0.1777, 0.2394, 0.3512], device='cuda:0'), in_proj_covar=tensor([0.0267, 0.0364, 0.0259, 0.0281, 0.0248, 0.0274, 0.0343, 0.0348], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-15 21:22:05,701 INFO [finetune.py:992] (0/2) Epoch 3, batch 4450, loss[loss=0.1546, simple_loss=0.2349, pruned_loss=0.03721, over 11764.00 frames. ], tot_loss[loss=0.1753, simple_loss=0.2633, pruned_loss=0.04361, over 2382623.72 frames. ], batch size: 26, lr: 4.90e-03, grad_scale: 8.0 2023-05-15 21:22:07,809 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.215e+02 3.079e+02 3.640e+02 4.177e+02 8.310e+02, threshold=7.280e+02, percent-clipped=2.0 2023-05-15 21:22:11,564 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.1019, 4.9718, 5.0481, 5.1135, 4.6945, 4.7658, 4.5429, 4.9888], device='cuda:0'), covar=tensor([0.0613, 0.0451, 0.0697, 0.0480, 0.1607, 0.1288, 0.0559, 0.0860], device='cuda:0'), in_proj_covar=tensor([0.0480, 0.0617, 0.0526, 0.0578, 0.0757, 0.0687, 0.0505, 0.0459], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0005, 0.0004, 0.0004, 0.0005, 0.0005, 0.0004, 0.0004], device='cuda:0') 2023-05-15 21:22:15,212 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.6198, 2.8515, 3.7127, 4.5908, 4.1345, 4.5109, 3.7717, 3.3847], device='cuda:0'), covar=tensor([0.0023, 0.0288, 0.0139, 0.0039, 0.0098, 0.0061, 0.0124, 0.0275], device='cuda:0'), in_proj_covar=tensor([0.0080, 0.0119, 0.0100, 0.0074, 0.0099, 0.0108, 0.0088, 0.0133], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-15 21:22:29,518 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=128951.0, num_to_drop=1, layers_to_drop={3} 2023-05-15 21:22:30,278 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.4310, 3.0908, 4.8716, 2.6339, 2.6032, 3.7867, 2.9520, 3.8728], device='cuda:0'), covar=tensor([0.0478, 0.1196, 0.0255, 0.1119, 0.1863, 0.1175, 0.1451, 0.0818], device='cuda:0'), in_proj_covar=tensor([0.0227, 0.0225, 0.0227, 0.0175, 0.0231, 0.0272, 0.0221, 0.0245], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:0') 2023-05-15 21:22:40,748 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.4287, 5.2265, 5.3586, 5.3907, 4.9846, 5.0683, 4.7978, 5.2445], device='cuda:0'), covar=tensor([0.0633, 0.0507, 0.0732, 0.0586, 0.1666, 0.1176, 0.0618, 0.0990], device='cuda:0'), in_proj_covar=tensor([0.0479, 0.0617, 0.0525, 0.0577, 0.0754, 0.0684, 0.0503, 0.0458], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0005, 0.0004, 0.0004, 0.0005, 0.0005, 0.0004, 0.0004], device='cuda:0') 2023-05-15 21:22:41,331 INFO [finetune.py:992] (0/2) Epoch 3, batch 4500, loss[loss=0.1745, simple_loss=0.2617, pruned_loss=0.04369, over 12254.00 frames. ], tot_loss[loss=0.1751, simple_loss=0.263, pruned_loss=0.0436, over 2374415.53 frames. ], batch size: 32, lr: 4.90e-03, grad_scale: 8.0 2023-05-15 21:22:46,933 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=128976.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 21:23:17,339 INFO [finetune.py:992] (0/2) Epoch 3, batch 4550, loss[loss=0.1567, simple_loss=0.2485, pruned_loss=0.03245, over 12293.00 frames. ], tot_loss[loss=0.1753, simple_loss=0.2632, pruned_loss=0.04367, over 2369843.62 frames. ], batch size: 33, lr: 4.90e-03, grad_scale: 8.0 2023-05-15 21:23:19,513 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.979e+02 3.068e+02 3.552e+02 4.377e+02 9.495e+02, threshold=7.104e+02, percent-clipped=2.0 2023-05-15 21:23:21,788 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=129024.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 21:23:48,941 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.84 vs. limit=2.0 2023-05-15 21:23:53,377 INFO [finetune.py:992] (0/2) Epoch 3, batch 4600, loss[loss=0.1855, simple_loss=0.2638, pruned_loss=0.05356, over 12134.00 frames. ], tot_loss[loss=0.1758, simple_loss=0.2639, pruned_loss=0.04386, over 2369861.89 frames. ], batch size: 30, lr: 4.90e-03, grad_scale: 8.0 2023-05-15 21:23:54,130 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=129069.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 21:24:28,068 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=129117.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 21:24:28,721 INFO [finetune.py:992] (0/2) Epoch 3, batch 4650, loss[loss=0.1638, simple_loss=0.2473, pruned_loss=0.04017, over 12178.00 frames. ], tot_loss[loss=0.1759, simple_loss=0.2639, pruned_loss=0.04396, over 2362614.09 frames. ], batch size: 31, lr: 4.90e-03, grad_scale: 8.0 2023-05-15 21:24:30,894 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.106e+02 2.760e+02 3.230e+02 3.906e+02 6.042e+02, threshold=6.460e+02, percent-clipped=0.0 2023-05-15 21:24:31,085 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.5101, 5.1271, 5.4677, 4.8863, 5.0519, 4.9020, 5.5433, 5.1425], device='cuda:0'), covar=tensor([0.0258, 0.0271, 0.0227, 0.0187, 0.0315, 0.0255, 0.0156, 0.0223], device='cuda:0'), in_proj_covar=tensor([0.0234, 0.0240, 0.0259, 0.0232, 0.0234, 0.0234, 0.0213, 0.0203], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-05-15 21:24:42,441 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=129137.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 21:24:47,440 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.5887, 2.7108, 4.0650, 4.3199, 2.8984, 2.6651, 2.7671, 2.1420], device='cuda:0'), covar=tensor([0.1355, 0.2503, 0.0519, 0.0434, 0.1060, 0.1870, 0.2405, 0.3507], device='cuda:0'), in_proj_covar=tensor([0.0270, 0.0370, 0.0262, 0.0284, 0.0251, 0.0277, 0.0347, 0.0353], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-15 21:25:04,895 INFO [finetune.py:992] (0/2) Epoch 3, batch 4700, loss[loss=0.1612, simple_loss=0.2567, pruned_loss=0.03286, over 12344.00 frames. ], tot_loss[loss=0.1762, simple_loss=0.2642, pruned_loss=0.0441, over 2354933.81 frames. ], batch size: 36, lr: 4.90e-03, grad_scale: 8.0 2023-05-15 21:25:10,095 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.4485, 4.7925, 4.1450, 5.1759, 4.5802, 3.1319, 4.3450, 3.2285], device='cuda:0'), covar=tensor([0.0664, 0.0659, 0.1175, 0.0244, 0.1014, 0.1311, 0.0850, 0.2602], device='cuda:0'), in_proj_covar=tensor([0.0291, 0.0362, 0.0340, 0.0255, 0.0349, 0.0257, 0.0326, 0.0350], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-15 21:25:24,882 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.8237, 3.5285, 5.1821, 2.7797, 2.8378, 3.8577, 3.2782, 3.9961], device='cuda:0'), covar=tensor([0.0427, 0.1025, 0.0255, 0.1095, 0.1867, 0.1348, 0.1341, 0.0943], device='cuda:0'), in_proj_covar=tensor([0.0230, 0.0228, 0.0229, 0.0177, 0.0233, 0.0276, 0.0224, 0.0248], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:0') 2023-05-15 21:25:41,426 INFO [finetune.py:992] (0/2) Epoch 3, batch 4750, loss[loss=0.146, simple_loss=0.2361, pruned_loss=0.02792, over 12185.00 frames. ], tot_loss[loss=0.1764, simple_loss=0.2645, pruned_loss=0.04412, over 2361176.18 frames. ], batch size: 31, lr: 4.90e-03, grad_scale: 8.0 2023-05-15 21:25:43,565 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.882e+02 2.721e+02 3.330e+02 3.956e+02 7.224e+02, threshold=6.660e+02, percent-clipped=1.0 2023-05-15 21:26:01,350 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=129246.0, num_to_drop=1, layers_to_drop={3} 2023-05-15 21:26:15,710 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.2949, 2.3972, 3.1892, 4.1810, 2.2329, 4.2896, 4.2424, 4.4437], device='cuda:0'), covar=tensor([0.0154, 0.0994, 0.0380, 0.0148, 0.1094, 0.0187, 0.0158, 0.0075], device='cuda:0'), in_proj_covar=tensor([0.0115, 0.0199, 0.0178, 0.0112, 0.0182, 0.0170, 0.0161, 0.0117], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-15 21:26:16,857 INFO [finetune.py:992] (0/2) Epoch 3, batch 4800, loss[loss=0.1936, simple_loss=0.287, pruned_loss=0.05015, over 12261.00 frames. ], tot_loss[loss=0.1757, simple_loss=0.2636, pruned_loss=0.04383, over 2370034.10 frames. ], batch size: 37, lr: 4.90e-03, grad_scale: 8.0 2023-05-15 21:26:24,833 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.0884, 2.2153, 3.6952, 3.0122, 3.3673, 3.0758, 2.3716, 3.5168], device='cuda:0'), covar=tensor([0.0092, 0.0319, 0.0100, 0.0198, 0.0109, 0.0158, 0.0322, 0.0087], device='cuda:0'), in_proj_covar=tensor([0.0158, 0.0192, 0.0168, 0.0174, 0.0191, 0.0148, 0.0181, 0.0162], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-15 21:26:52,926 INFO [finetune.py:992] (0/2) Epoch 3, batch 4850, loss[loss=0.176, simple_loss=0.2687, pruned_loss=0.04159, over 10807.00 frames. ], tot_loss[loss=0.1761, simple_loss=0.264, pruned_loss=0.04409, over 2375151.96 frames. ], batch size: 68, lr: 4.90e-03, grad_scale: 8.0 2023-05-15 21:26:55,088 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.959e+02 2.802e+02 3.179e+02 3.791e+02 1.091e+03, threshold=6.357e+02, percent-clipped=2.0 2023-05-15 21:27:29,165 INFO [finetune.py:992] (0/2) Epoch 3, batch 4900, loss[loss=0.1524, simple_loss=0.2387, pruned_loss=0.03307, over 12177.00 frames. ], tot_loss[loss=0.1768, simple_loss=0.2645, pruned_loss=0.0445, over 2364773.47 frames. ], batch size: 29, lr: 4.90e-03, grad_scale: 4.0 2023-05-15 21:28:01,410 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=129413.0, num_to_drop=1, layers_to_drop={0} 2023-05-15 21:28:04,576 INFO [finetune.py:992] (0/2) Epoch 3, batch 4950, loss[loss=0.2136, simple_loss=0.3017, pruned_loss=0.06276, over 12130.00 frames. ], tot_loss[loss=0.1773, simple_loss=0.2649, pruned_loss=0.04481, over 2369036.36 frames. ], batch size: 38, lr: 4.90e-03, grad_scale: 4.0 2023-05-15 21:28:07,400 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.019e+02 2.853e+02 3.450e+02 4.427e+02 1.611e+03, threshold=6.900e+02, percent-clipped=5.0 2023-05-15 21:28:18,282 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=129437.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 21:28:41,016 INFO [finetune.py:992] (0/2) Epoch 3, batch 5000, loss[loss=0.1466, simple_loss=0.2306, pruned_loss=0.03127, over 12303.00 frames. ], tot_loss[loss=0.1764, simple_loss=0.2642, pruned_loss=0.04434, over 2369412.26 frames. ], batch size: 28, lr: 4.90e-03, grad_scale: 4.0 2023-05-15 21:28:45,537 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=129474.0, num_to_drop=1, layers_to_drop={2} 2023-05-15 21:28:53,901 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=129485.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 21:29:17,365 INFO [finetune.py:992] (0/2) Epoch 3, batch 5050, loss[loss=0.148, simple_loss=0.2286, pruned_loss=0.03374, over 12113.00 frames. ], tot_loss[loss=0.1756, simple_loss=0.2635, pruned_loss=0.04383, over 2360897.69 frames. ], batch size: 30, lr: 4.90e-03, grad_scale: 4.0 2023-05-15 21:29:20,129 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.632e+02 2.734e+02 3.226e+02 3.747e+02 5.556e+02, threshold=6.452e+02, percent-clipped=0.0 2023-05-15 21:29:37,775 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=129546.0, num_to_drop=1, layers_to_drop={1} 2023-05-15 21:29:53,343 INFO [finetune.py:992] (0/2) Epoch 3, batch 5100, loss[loss=0.1456, simple_loss=0.2349, pruned_loss=0.02811, over 12178.00 frames. ], tot_loss[loss=0.1757, simple_loss=0.2637, pruned_loss=0.04389, over 2357811.63 frames. ], batch size: 29, lr: 4.90e-03, grad_scale: 4.0 2023-05-15 21:30:12,236 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=129594.0, num_to_drop=1, layers_to_drop={0} 2023-05-15 21:30:25,258 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=129612.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 21:30:27,681 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=3.34 vs. limit=5.0 2023-05-15 21:30:29,420 INFO [finetune.py:992] (0/2) Epoch 3, batch 5150, loss[loss=0.1735, simple_loss=0.2643, pruned_loss=0.04136, over 11177.00 frames. ], tot_loss[loss=0.1758, simple_loss=0.264, pruned_loss=0.04384, over 2362713.66 frames. ], batch size: 55, lr: 4.90e-03, grad_scale: 4.0 2023-05-15 21:30:32,844 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.912e+02 2.725e+02 3.094e+02 3.822e+02 6.071e+02, threshold=6.187e+02, percent-clipped=0.0 2023-05-15 21:31:05,827 INFO [finetune.py:992] (0/2) Epoch 3, batch 5200, loss[loss=0.1621, simple_loss=0.2543, pruned_loss=0.03492, over 12109.00 frames. ], tot_loss[loss=0.175, simple_loss=0.2631, pruned_loss=0.04348, over 2370753.92 frames. ], batch size: 30, lr: 4.90e-03, grad_scale: 8.0 2023-05-15 21:31:09,626 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=129673.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 21:31:40,631 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.6285, 2.3508, 3.3867, 4.4499, 2.4232, 4.5396, 4.5677, 4.7616], device='cuda:0'), covar=tensor([0.0103, 0.1190, 0.0387, 0.0157, 0.1138, 0.0178, 0.0140, 0.0069], device='cuda:0'), in_proj_covar=tensor([0.0114, 0.0199, 0.0178, 0.0111, 0.0181, 0.0169, 0.0161, 0.0116], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-15 21:31:41,156 INFO [finetune.py:992] (0/2) Epoch 3, batch 5250, loss[loss=0.1762, simple_loss=0.2588, pruned_loss=0.04678, over 12331.00 frames. ], tot_loss[loss=0.1746, simple_loss=0.2623, pruned_loss=0.04346, over 2375097.01 frames. ], batch size: 31, lr: 4.89e-03, grad_scale: 8.0 2023-05-15 21:31:43,947 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.946e+02 2.785e+02 3.303e+02 4.067e+02 7.536e+02, threshold=6.606e+02, percent-clipped=4.0 2023-05-15 21:32:17,405 INFO [finetune.py:992] (0/2) Epoch 3, batch 5300, loss[loss=0.1579, simple_loss=0.2379, pruned_loss=0.03895, over 12282.00 frames. ], tot_loss[loss=0.1746, simple_loss=0.2624, pruned_loss=0.04342, over 2377786.77 frames. ], batch size: 28, lr: 4.89e-03, grad_scale: 8.0 2023-05-15 21:32:18,203 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=129769.0, num_to_drop=1, layers_to_drop={3} 2023-05-15 21:32:23,275 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.4180, 4.0770, 4.2175, 4.5073, 3.0016, 4.0387, 2.6633, 4.0157], device='cuda:0'), covar=tensor([0.1508, 0.0639, 0.0869, 0.0515, 0.1103, 0.0546, 0.1813, 0.1477], device='cuda:0'), in_proj_covar=tensor([0.0224, 0.0262, 0.0295, 0.0351, 0.0241, 0.0238, 0.0259, 0.0389], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-15 21:32:34,759 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=129791.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 21:32:53,848 INFO [finetune.py:992] (0/2) Epoch 3, batch 5350, loss[loss=0.1681, simple_loss=0.2432, pruned_loss=0.04644, over 12170.00 frames. ], tot_loss[loss=0.1748, simple_loss=0.2622, pruned_loss=0.04366, over 2379030.50 frames. ], batch size: 29, lr: 4.89e-03, grad_scale: 8.0 2023-05-15 21:32:56,589 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.064e+02 2.918e+02 3.504e+02 4.192e+02 1.918e+03, threshold=7.007e+02, percent-clipped=5.0 2023-05-15 21:33:03,552 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.83 vs. limit=2.0 2023-05-15 21:33:03,794 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.5706, 4.5165, 4.4408, 4.5530, 4.1266, 4.6113, 4.6525, 4.8391], device='cuda:0'), covar=tensor([0.0199, 0.0153, 0.0179, 0.0262, 0.0671, 0.0246, 0.0142, 0.0137], device='cuda:0'), in_proj_covar=tensor([0.0180, 0.0180, 0.0180, 0.0225, 0.0228, 0.0198, 0.0166, 0.0212], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0005, 0.0005, 0.0005, 0.0004, 0.0005], device='cuda:0') 2023-05-15 21:33:18,120 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=129852.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 21:33:29,121 INFO [finetune.py:992] (0/2) Epoch 3, batch 5400, loss[loss=0.1491, simple_loss=0.231, pruned_loss=0.03359, over 12093.00 frames. ], tot_loss[loss=0.1745, simple_loss=0.262, pruned_loss=0.0435, over 2384024.65 frames. ], batch size: 32, lr: 4.89e-03, grad_scale: 8.0 2023-05-15 21:33:40,113 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.6546, 4.8537, 4.2883, 5.1753, 4.7516, 3.2541, 4.4596, 3.2432], device='cuda:0'), covar=tensor([0.0664, 0.0698, 0.1205, 0.0322, 0.0961, 0.1386, 0.0882, 0.2999], device='cuda:0'), in_proj_covar=tensor([0.0296, 0.0368, 0.0346, 0.0259, 0.0356, 0.0262, 0.0332, 0.0355], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-15 21:34:06,045 INFO [finetune.py:992] (0/2) Epoch 3, batch 5450, loss[loss=0.1944, simple_loss=0.2885, pruned_loss=0.05014, over 10353.00 frames. ], tot_loss[loss=0.1745, simple_loss=0.2622, pruned_loss=0.04342, over 2389011.36 frames. ], batch size: 68, lr: 4.89e-03, grad_scale: 8.0 2023-05-15 21:34:08,846 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.118e+02 2.953e+02 3.315e+02 4.012e+02 8.490e+02, threshold=6.630e+02, percent-clipped=3.0 2023-05-15 21:34:41,407 INFO [finetune.py:992] (0/2) Epoch 3, batch 5500, loss[loss=0.1894, simple_loss=0.2764, pruned_loss=0.05122, over 12162.00 frames. ], tot_loss[loss=0.1751, simple_loss=0.2626, pruned_loss=0.0438, over 2382372.42 frames. ], batch size: 34, lr: 4.89e-03, grad_scale: 8.0 2023-05-15 21:34:41,484 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=129968.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 21:34:41,783 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=3.14 vs. limit=5.0 2023-05-15 21:35:04,351 INFO [checkpoint.py:75] (0/2) Saving checkpoint to pruned_transducer_stateless7/exp_giga_finetune/checkpoint-30000.pt 2023-05-15 21:35:07,613 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=130000.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 21:35:09,133 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=130002.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 21:35:19,911 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=3.06 vs. limit=5.0 2023-05-15 21:35:20,235 INFO [finetune.py:992] (0/2) Epoch 3, batch 5550, loss[loss=0.175, simple_loss=0.2682, pruned_loss=0.04089, over 12109.00 frames. ], tot_loss[loss=0.1761, simple_loss=0.2637, pruned_loss=0.04421, over 2380026.85 frames. ], batch size: 33, lr: 4.89e-03, grad_scale: 8.0 2023-05-15 21:35:23,115 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.979e+02 2.826e+02 3.312e+02 3.995e+02 7.380e+02, threshold=6.624e+02, percent-clipped=1.0 2023-05-15 21:35:23,316 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.8719, 4.4753, 4.4238, 4.6970, 4.5244, 4.6715, 4.5828, 2.4812], device='cuda:0'), covar=tensor([0.0101, 0.0079, 0.0129, 0.0073, 0.0073, 0.0113, 0.0097, 0.0837], device='cuda:0'), in_proj_covar=tensor([0.0064, 0.0073, 0.0077, 0.0070, 0.0057, 0.0086, 0.0076, 0.0090], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-15 21:35:37,065 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.3989, 5.0971, 5.4427, 5.3532, 4.4490, 4.6841, 4.7721, 5.1299], device='cuda:0'), covar=tensor([0.0858, 0.0927, 0.0608, 0.0801, 0.2946, 0.1727, 0.0696, 0.1459], device='cuda:0'), in_proj_covar=tensor([0.0478, 0.0616, 0.0525, 0.0580, 0.0762, 0.0693, 0.0506, 0.0455], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0005, 0.0004, 0.0004, 0.0005, 0.0005, 0.0004, 0.0004], device='cuda:0') 2023-05-15 21:35:51,776 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=130061.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 21:35:53,205 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=130063.0, num_to_drop=1, layers_to_drop={3} 2023-05-15 21:35:56,622 INFO [finetune.py:992] (0/2) Epoch 3, batch 5600, loss[loss=0.146, simple_loss=0.237, pruned_loss=0.02752, over 12020.00 frames. ], tot_loss[loss=0.1758, simple_loss=0.2635, pruned_loss=0.04409, over 2383069.93 frames. ], batch size: 31, lr: 4.89e-03, grad_scale: 8.0 2023-05-15 21:35:57,488 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=130069.0, num_to_drop=1, layers_to_drop={1} 2023-05-15 21:36:31,263 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=130117.0, num_to_drop=1, layers_to_drop={0} 2023-05-15 21:36:31,841 INFO [finetune.py:992] (0/2) Epoch 3, batch 5650, loss[loss=0.2007, simple_loss=0.2853, pruned_loss=0.05806, over 12062.00 frames. ], tot_loss[loss=0.1768, simple_loss=0.2641, pruned_loss=0.04473, over 2381911.98 frames. ], batch size: 37, lr: 4.89e-03, grad_scale: 8.0 2023-05-15 21:36:34,703 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.967e+02 2.738e+02 3.427e+02 4.073e+02 1.202e+03, threshold=6.854e+02, percent-clipped=1.0 2023-05-15 21:36:36,987 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.4687, 2.2773, 3.3290, 4.3124, 2.3994, 4.3943, 4.3880, 4.6085], device='cuda:0'), covar=tensor([0.0103, 0.1170, 0.0401, 0.0125, 0.1146, 0.0194, 0.0121, 0.0061], device='cuda:0'), in_proj_covar=tensor([0.0117, 0.0203, 0.0183, 0.0114, 0.0186, 0.0174, 0.0165, 0.0119], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-15 21:36:52,384 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=130147.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 21:37:03,204 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.5124, 3.6399, 3.2516, 3.1925, 2.8433, 2.6378, 3.5579, 2.1715], device='cuda:0'), covar=tensor([0.0318, 0.0099, 0.0133, 0.0156, 0.0317, 0.0281, 0.0104, 0.0408], device='cuda:0'), in_proj_covar=tensor([0.0184, 0.0156, 0.0147, 0.0175, 0.0199, 0.0191, 0.0155, 0.0196], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-15 21:37:07,263 INFO [finetune.py:992] (0/2) Epoch 3, batch 5700, loss[loss=0.1444, simple_loss=0.2303, pruned_loss=0.0292, over 12180.00 frames. ], tot_loss[loss=0.1754, simple_loss=0.2625, pruned_loss=0.04414, over 2380773.64 frames. ], batch size: 31, lr: 4.89e-03, grad_scale: 8.0 2023-05-15 21:37:29,209 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-05-15 21:37:44,817 INFO [finetune.py:992] (0/2) Epoch 3, batch 5750, loss[loss=0.1681, simple_loss=0.2532, pruned_loss=0.04149, over 12346.00 frames. ], tot_loss[loss=0.176, simple_loss=0.2631, pruned_loss=0.0444, over 2378788.69 frames. ], batch size: 31, lr: 4.89e-03, grad_scale: 8.0 2023-05-15 21:37:47,432 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.153e+02 3.005e+02 3.316e+02 3.944e+02 6.494e+02, threshold=6.631e+02, percent-clipped=0.0 2023-05-15 21:38:02,142 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.38 vs. limit=2.0 2023-05-15 21:38:20,141 INFO [finetune.py:992] (0/2) Epoch 3, batch 5800, loss[loss=0.2087, simple_loss=0.2872, pruned_loss=0.06512, over 12040.00 frames. ], tot_loss[loss=0.1773, simple_loss=0.2644, pruned_loss=0.04512, over 2370154.55 frames. ], batch size: 37, lr: 4.89e-03, grad_scale: 8.0 2023-05-15 21:38:20,238 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=130268.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 21:38:36,139 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.92 vs. limit=2.0 2023-05-15 21:38:54,443 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=130316.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 21:38:55,825 INFO [finetune.py:992] (0/2) Epoch 3, batch 5850, loss[loss=0.1668, simple_loss=0.2608, pruned_loss=0.03644, over 11255.00 frames. ], tot_loss[loss=0.1753, simple_loss=0.2627, pruned_loss=0.04398, over 2378754.80 frames. ], batch size: 55, lr: 4.89e-03, grad_scale: 8.0 2023-05-15 21:38:58,542 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.062e+02 2.888e+02 3.485e+02 4.475e+02 7.584e+02, threshold=6.969e+02, percent-clipped=2.0 2023-05-15 21:39:23,964 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=130356.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 21:39:25,420 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=130358.0, num_to_drop=1, layers_to_drop={3} 2023-05-15 21:39:32,433 INFO [finetune.py:992] (0/2) Epoch 3, batch 5900, loss[loss=0.1692, simple_loss=0.2542, pruned_loss=0.04209, over 12083.00 frames. ], tot_loss[loss=0.1764, simple_loss=0.2637, pruned_loss=0.04449, over 2380005.43 frames. ], batch size: 32, lr: 4.89e-03, grad_scale: 8.0 2023-05-15 21:39:43,809 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=130384.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 21:39:44,649 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.2646, 3.8955, 3.9105, 4.3521, 2.8937, 3.9915, 2.5523, 3.9596], device='cuda:0'), covar=tensor([0.1618, 0.0719, 0.1015, 0.0621, 0.1172, 0.0552, 0.1812, 0.1328], device='cuda:0'), in_proj_covar=tensor([0.0223, 0.0261, 0.0294, 0.0350, 0.0237, 0.0237, 0.0255, 0.0385], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-15 21:40:08,137 INFO [finetune.py:992] (0/2) Epoch 3, batch 5950, loss[loss=0.1564, simple_loss=0.2364, pruned_loss=0.0382, over 12174.00 frames. ], tot_loss[loss=0.1762, simple_loss=0.2638, pruned_loss=0.04428, over 2386315.89 frames. ], batch size: 29, lr: 4.89e-03, grad_scale: 8.0 2023-05-15 21:40:10,939 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.014e+02 2.936e+02 3.497e+02 4.249e+02 9.499e+02, threshold=6.995e+02, percent-clipped=1.0 2023-05-15 21:40:18,373 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=130432.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 21:40:19,810 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.1298, 4.9376, 4.9634, 5.0455, 4.7189, 5.0229, 4.9329, 2.8409], device='cuda:0'), covar=tensor([0.0116, 0.0062, 0.0072, 0.0064, 0.0056, 0.0083, 0.0091, 0.0601], device='cuda:0'), in_proj_covar=tensor([0.0065, 0.0074, 0.0077, 0.0070, 0.0057, 0.0086, 0.0076, 0.0090], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-15 21:40:22,567 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=130438.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 21:40:27,606 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=130445.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 21:40:29,061 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=130447.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 21:40:43,771 INFO [finetune.py:992] (0/2) Epoch 3, batch 6000, loss[loss=0.1699, simple_loss=0.2676, pruned_loss=0.03604, over 12007.00 frames. ], tot_loss[loss=0.1773, simple_loss=0.265, pruned_loss=0.04477, over 2382678.57 frames. ], batch size: 42, lr: 4.89e-03, grad_scale: 8.0 2023-05-15 21:40:43,772 INFO [finetune.py:1017] (0/2) Computing validation loss 2023-05-15 21:41:01,521 INFO [finetune.py:1026] (0/2) Epoch 3, validation: loss=0.3355, simple_loss=0.4069, pruned_loss=0.1321, over 1020973.00 frames. 2023-05-15 21:41:01,522 INFO [finetune.py:1027] (0/2) Maximum memory allocated so far is 12745MB 2023-05-15 21:41:06,204 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=4.52 vs. limit=5.0 2023-05-15 21:41:07,720 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.40 vs. limit=2.0 2023-05-15 21:41:19,542 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=130493.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 21:41:20,829 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=130495.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 21:41:23,854 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=130499.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 21:41:37,112 INFO [finetune.py:992] (0/2) Epoch 3, batch 6050, loss[loss=0.1921, simple_loss=0.2832, pruned_loss=0.05043, over 12140.00 frames. ], tot_loss[loss=0.1779, simple_loss=0.2653, pruned_loss=0.04527, over 2379328.93 frames. ], batch size: 38, lr: 4.89e-03, grad_scale: 8.0 2023-05-15 21:41:39,968 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.197e+02 2.959e+02 3.443e+02 4.363e+02 1.639e+03, threshold=6.886e+02, percent-clipped=4.0 2023-05-15 21:41:58,619 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=130548.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 21:42:10,770 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=130565.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 21:42:12,584 INFO [finetune.py:992] (0/2) Epoch 3, batch 6100, loss[loss=0.1839, simple_loss=0.2815, pruned_loss=0.04309, over 11403.00 frames. ], tot_loss[loss=0.1786, simple_loss=0.2656, pruned_loss=0.04578, over 2368568.73 frames. ], batch size: 55, lr: 4.89e-03, grad_scale: 8.0 2023-05-15 21:42:43,453 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=130609.0, num_to_drop=1, layers_to_drop={2} 2023-05-15 21:42:47,194 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2023-05-15 21:42:49,482 INFO [finetune.py:992] (0/2) Epoch 3, batch 6150, loss[loss=0.2032, simple_loss=0.301, pruned_loss=0.05269, over 12016.00 frames. ], tot_loss[loss=0.1782, simple_loss=0.2655, pruned_loss=0.04546, over 2371741.29 frames. ], batch size: 40, lr: 4.89e-03, grad_scale: 8.0 2023-05-15 21:42:52,330 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.836e+02 2.993e+02 3.524e+02 4.331e+02 7.227e+02, threshold=7.047e+02, percent-clipped=1.0 2023-05-15 21:42:55,361 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=130626.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 21:43:16,728 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=130656.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 21:43:18,027 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=130658.0, num_to_drop=1, layers_to_drop={1} 2023-05-15 21:43:24,952 INFO [finetune.py:992] (0/2) Epoch 3, batch 6200, loss[loss=0.1734, simple_loss=0.2675, pruned_loss=0.03965, over 12150.00 frames. ], tot_loss[loss=0.1789, simple_loss=0.2665, pruned_loss=0.04563, over 2363851.70 frames. ], batch size: 36, lr: 4.89e-03, grad_scale: 8.0 2023-05-15 21:43:50,274 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=130704.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 21:43:51,744 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=130706.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 21:44:00,966 INFO [finetune.py:992] (0/2) Epoch 3, batch 6250, loss[loss=0.1799, simple_loss=0.2688, pruned_loss=0.04551, over 12000.00 frames. ], tot_loss[loss=0.1791, simple_loss=0.2669, pruned_loss=0.0457, over 2369280.03 frames. ], batch size: 40, lr: 4.89e-03, grad_scale: 8.0 2023-05-15 21:44:03,768 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.835e+02 2.911e+02 3.273e+02 3.794e+02 8.028e+02, threshold=6.546e+02, percent-clipped=1.0 2023-05-15 21:44:10,253 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.6626, 2.3937, 3.4534, 4.5457, 2.3361, 4.6158, 4.5901, 4.7420], device='cuda:0'), covar=tensor([0.0095, 0.1180, 0.0399, 0.0117, 0.1220, 0.0145, 0.0120, 0.0081], device='cuda:0'), in_proj_covar=tensor([0.0117, 0.0200, 0.0181, 0.0113, 0.0183, 0.0171, 0.0163, 0.0118], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-15 21:44:16,543 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=130740.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 21:44:36,998 INFO [finetune.py:992] (0/2) Epoch 3, batch 6300, loss[loss=0.1887, simple_loss=0.2758, pruned_loss=0.05077, over 12282.00 frames. ], tot_loss[loss=0.1789, simple_loss=0.2665, pruned_loss=0.0457, over 2362544.16 frames. ], batch size: 33, lr: 4.89e-03, grad_scale: 8.0 2023-05-15 21:44:51,364 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=130788.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 21:44:55,558 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=130794.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 21:45:12,703 INFO [finetune.py:992] (0/2) Epoch 3, batch 6350, loss[loss=0.1573, simple_loss=0.2389, pruned_loss=0.03786, over 12263.00 frames. ], tot_loss[loss=0.1796, simple_loss=0.2673, pruned_loss=0.04602, over 2361640.73 frames. ], batch size: 28, lr: 4.89e-03, grad_scale: 8.0 2023-05-15 21:45:14,219 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.1495, 5.9994, 5.5084, 5.4985, 6.1072, 5.3545, 5.5315, 5.5427], device='cuda:0'), covar=tensor([0.1506, 0.0837, 0.0866, 0.1683, 0.0844, 0.1994, 0.1648, 0.0920], device='cuda:0'), in_proj_covar=tensor([0.0335, 0.0457, 0.0360, 0.0412, 0.0438, 0.0412, 0.0369, 0.0360], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-15 21:45:15,393 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.159e+02 2.855e+02 3.388e+02 4.377e+02 6.922e+02, threshold=6.775e+02, percent-clipped=1.0 2023-05-15 21:45:21,433 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.0073, 4.8903, 4.7121, 4.8395, 4.4651, 4.9337, 4.9572, 5.1793], device='cuda:0'), covar=tensor([0.0230, 0.0126, 0.0189, 0.0237, 0.0680, 0.0230, 0.0135, 0.0139], device='cuda:0'), in_proj_covar=tensor([0.0179, 0.0178, 0.0178, 0.0223, 0.0225, 0.0196, 0.0164, 0.0210], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0005, 0.0005, 0.0005, 0.0004, 0.0005], device='cuda:0') 2023-05-15 21:45:28,019 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2023-05-15 21:45:48,721 INFO [finetune.py:992] (0/2) Epoch 3, batch 6400, loss[loss=0.1594, simple_loss=0.2524, pruned_loss=0.03315, over 12308.00 frames. ], tot_loss[loss=0.1791, simple_loss=0.2667, pruned_loss=0.04572, over 2370258.59 frames. ], batch size: 34, lr: 4.89e-03, grad_scale: 8.0 2023-05-15 21:46:15,221 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=130904.0, num_to_drop=1, layers_to_drop={2} 2023-05-15 21:46:25,320 INFO [finetune.py:992] (0/2) Epoch 3, batch 6450, loss[loss=0.178, simple_loss=0.2663, pruned_loss=0.04482, over 11260.00 frames. ], tot_loss[loss=0.1788, simple_loss=0.2666, pruned_loss=0.04553, over 2369664.30 frames. ], batch size: 55, lr: 4.89e-03, grad_scale: 8.0 2023-05-15 21:46:27,460 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=130921.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 21:46:28,035 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.906e+02 2.956e+02 3.435e+02 4.132e+02 8.465e+02, threshold=6.870e+02, percent-clipped=2.0 2023-05-15 21:47:00,294 INFO [finetune.py:992] (0/2) Epoch 3, batch 6500, loss[loss=0.2043, simple_loss=0.2888, pruned_loss=0.05996, over 12020.00 frames. ], tot_loss[loss=0.1795, simple_loss=0.2671, pruned_loss=0.04589, over 2366263.33 frames. ], batch size: 42, lr: 4.89e-03, grad_scale: 8.0 2023-05-15 21:47:15,315 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.0165, 4.5899, 4.7317, 4.8389, 4.5613, 4.8308, 4.7224, 2.5473], device='cuda:0'), covar=tensor([0.0093, 0.0066, 0.0095, 0.0065, 0.0054, 0.0099, 0.0082, 0.0701], device='cuda:0'), in_proj_covar=tensor([0.0064, 0.0072, 0.0076, 0.0070, 0.0057, 0.0086, 0.0075, 0.0089], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-15 21:47:36,999 INFO [finetune.py:992] (0/2) Epoch 3, batch 6550, loss[loss=0.2312, simple_loss=0.3156, pruned_loss=0.07344, over 11249.00 frames. ], tot_loss[loss=0.179, simple_loss=0.2668, pruned_loss=0.04562, over 2369194.97 frames. ], batch size: 55, lr: 4.89e-03, grad_scale: 8.0 2023-05-15 21:47:37,865 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.4321, 4.9039, 5.3403, 4.7509, 4.9720, 4.7103, 5.3829, 5.0276], device='cuda:0'), covar=tensor([0.0230, 0.0325, 0.0245, 0.0219, 0.0320, 0.0300, 0.0175, 0.0294], device='cuda:0'), in_proj_covar=tensor([0.0234, 0.0238, 0.0261, 0.0232, 0.0232, 0.0233, 0.0214, 0.0201], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-05-15 21:47:39,768 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.991e+02 2.885e+02 3.348e+02 4.078e+02 8.437e+02, threshold=6.696e+02, percent-clipped=3.0 2023-05-15 21:47:53,106 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=131040.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 21:47:57,747 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.92 vs. limit=2.0 2023-05-15 21:48:12,996 INFO [finetune.py:992] (0/2) Epoch 3, batch 6600, loss[loss=0.1849, simple_loss=0.282, pruned_loss=0.04393, over 12150.00 frames. ], tot_loss[loss=0.1789, simple_loss=0.2664, pruned_loss=0.04567, over 2374740.98 frames. ], batch size: 36, lr: 4.89e-03, grad_scale: 8.0 2023-05-15 21:48:14,585 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.9653, 5.9368, 5.7265, 5.3323, 5.0929, 5.8837, 5.3897, 5.2808], device='cuda:0'), covar=tensor([0.0728, 0.0840, 0.0576, 0.1301, 0.0600, 0.0633, 0.1538, 0.1070], device='cuda:0'), in_proj_covar=tensor([0.0570, 0.0507, 0.0464, 0.0579, 0.0380, 0.0656, 0.0715, 0.0527], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0004, 0.0004, 0.0003], device='cuda:0') 2023-05-15 21:48:20,283 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.8210, 2.4790, 3.6612, 3.7799, 3.0059, 2.7425, 2.6153, 2.4817], device='cuda:0'), covar=tensor([0.0971, 0.2402, 0.0533, 0.0415, 0.0734, 0.1521, 0.2146, 0.2739], device='cuda:0'), in_proj_covar=tensor([0.0267, 0.0367, 0.0263, 0.0284, 0.0249, 0.0276, 0.0343, 0.0348], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-15 21:48:27,140 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=131088.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 21:48:27,259 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=131088.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 21:48:31,391 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=131094.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 21:48:48,251 INFO [finetune.py:992] (0/2) Epoch 3, batch 6650, loss[loss=0.1705, simple_loss=0.2667, pruned_loss=0.03715, over 12353.00 frames. ], tot_loss[loss=0.1799, simple_loss=0.267, pruned_loss=0.04638, over 2363536.16 frames. ], batch size: 35, lr: 4.89e-03, grad_scale: 8.0 2023-05-15 21:48:51,059 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.820e+02 3.035e+02 3.644e+02 4.405e+02 1.056e+03, threshold=7.288e+02, percent-clipped=5.0 2023-05-15 21:48:55,502 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.5564, 2.6146, 3.8951, 4.4760, 4.1815, 4.4770, 3.8633, 3.2647], device='cuda:0'), covar=tensor([0.0022, 0.0327, 0.0087, 0.0039, 0.0074, 0.0064, 0.0108, 0.0291], device='cuda:0'), in_proj_covar=tensor([0.0080, 0.0120, 0.0100, 0.0074, 0.0099, 0.0108, 0.0089, 0.0133], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-15 21:49:01,266 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=131136.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 21:49:05,030 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.3004, 2.6264, 3.8658, 3.3084, 3.6420, 3.4088, 2.7971, 3.7589], device='cuda:0'), covar=tensor([0.0093, 0.0268, 0.0091, 0.0181, 0.0108, 0.0123, 0.0251, 0.0083], device='cuda:0'), in_proj_covar=tensor([0.0158, 0.0190, 0.0169, 0.0171, 0.0191, 0.0147, 0.0180, 0.0160], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-15 21:49:05,545 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=131142.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 21:49:24,684 INFO [finetune.py:992] (0/2) Epoch 3, batch 6700, loss[loss=0.1799, simple_loss=0.2754, pruned_loss=0.04218, over 12137.00 frames. ], tot_loss[loss=0.1796, simple_loss=0.2667, pruned_loss=0.04626, over 2359381.56 frames. ], batch size: 39, lr: 4.88e-03, grad_scale: 8.0 2023-05-15 21:49:29,145 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=131174.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 21:49:51,657 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=131204.0, num_to_drop=1, layers_to_drop={2} 2023-05-15 21:50:01,491 INFO [finetune.py:992] (0/2) Epoch 3, batch 6750, loss[loss=0.1828, simple_loss=0.2748, pruned_loss=0.04536, over 12104.00 frames. ], tot_loss[loss=0.1779, simple_loss=0.2652, pruned_loss=0.04527, over 2364483.20 frames. ], batch size: 33, lr: 4.88e-03, grad_scale: 8.0 2023-05-15 21:50:03,681 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=131221.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 21:50:04,170 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.886e+02 2.729e+02 3.205e+02 3.799e+02 1.174e+03, threshold=6.410e+02, percent-clipped=3.0 2023-05-15 21:50:13,655 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=131235.0, num_to_drop=1, layers_to_drop={2} 2023-05-15 21:50:25,469 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=131252.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 21:50:36,909 INFO [finetune.py:992] (0/2) Epoch 3, batch 6800, loss[loss=0.2195, simple_loss=0.3199, pruned_loss=0.05951, over 11479.00 frames. ], tot_loss[loss=0.178, simple_loss=0.2652, pruned_loss=0.04535, over 2367359.65 frames. ], batch size: 48, lr: 4.88e-03, grad_scale: 8.0 2023-05-15 21:50:37,595 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=131269.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 21:50:57,010 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.51 vs. limit=2.0 2023-05-15 21:51:04,368 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 2023-05-15 21:51:12,861 INFO [finetune.py:992] (0/2) Epoch 3, batch 6850, loss[loss=0.2061, simple_loss=0.2901, pruned_loss=0.06112, over 12039.00 frames. ], tot_loss[loss=0.1777, simple_loss=0.2648, pruned_loss=0.04529, over 2369129.44 frames. ], batch size: 40, lr: 4.88e-03, grad_scale: 8.0 2023-05-15 21:51:15,646 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.997e+02 3.070e+02 3.537e+02 4.360e+02 1.019e+03, threshold=7.074e+02, percent-clipped=8.0 2023-05-15 21:51:48,832 INFO [finetune.py:992] (0/2) Epoch 3, batch 6900, loss[loss=0.1781, simple_loss=0.2615, pruned_loss=0.04732, over 12034.00 frames. ], tot_loss[loss=0.1777, simple_loss=0.2648, pruned_loss=0.04526, over 2365642.52 frames. ], batch size: 31, lr: 4.88e-03, grad_scale: 8.0 2023-05-15 21:51:51,098 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([6.0837, 6.0075, 6.0910, 5.2309, 5.0587, 6.0847, 5.1856, 5.6520], device='cuda:0'), covar=tensor([0.0802, 0.1082, 0.0958, 0.2374, 0.1181, 0.1205, 0.3029, 0.1717], device='cuda:0'), in_proj_covar=tensor([0.0572, 0.0511, 0.0471, 0.0584, 0.0381, 0.0662, 0.0724, 0.0531], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0004, 0.0004, 0.0003], device='cuda:0') 2023-05-15 21:51:54,929 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2023-05-15 21:52:17,991 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=131409.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 21:52:18,879 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 2023-05-15 21:52:22,951 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=131416.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 21:52:24,192 INFO [finetune.py:992] (0/2) Epoch 3, batch 6950, loss[loss=0.1471, simple_loss=0.2321, pruned_loss=0.03108, over 12295.00 frames. ], tot_loss[loss=0.1789, simple_loss=0.2662, pruned_loss=0.04582, over 2362620.74 frames. ], batch size: 33, lr: 4.88e-03, grad_scale: 8.0 2023-05-15 21:52:27,692 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.982e+02 3.023e+02 3.604e+02 4.404e+02 1.310e+03, threshold=7.207e+02, percent-clipped=2.0 2023-05-15 21:52:45,727 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.93 vs. limit=2.0 2023-05-15 21:52:47,641 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.0791, 5.0121, 4.8451, 4.9483, 4.5096, 4.9925, 5.0309, 5.2736], device='cuda:0'), covar=tensor([0.0151, 0.0124, 0.0171, 0.0257, 0.0687, 0.0220, 0.0131, 0.0142], device='cuda:0'), in_proj_covar=tensor([0.0179, 0.0179, 0.0179, 0.0226, 0.0227, 0.0198, 0.0164, 0.0211], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0005, 0.0005, 0.0005, 0.0004, 0.0005], device='cuda:0') 2023-05-15 21:52:56,733 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.0346, 4.9678, 4.7773, 4.8514, 4.4813, 4.9392, 5.0109, 5.1800], device='cuda:0'), covar=tensor([0.0154, 0.0126, 0.0179, 0.0293, 0.0676, 0.0267, 0.0130, 0.0154], device='cuda:0'), in_proj_covar=tensor([0.0179, 0.0180, 0.0179, 0.0226, 0.0228, 0.0199, 0.0164, 0.0211], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0005, 0.0005, 0.0005, 0.0004, 0.0005], device='cuda:0') 2023-05-15 21:53:00,824 INFO [finetune.py:992] (0/2) Epoch 3, batch 7000, loss[loss=0.1871, simple_loss=0.2813, pruned_loss=0.04646, over 12300.00 frames. ], tot_loss[loss=0.1777, simple_loss=0.2652, pruned_loss=0.04512, over 2375831.39 frames. ], batch size: 34, lr: 4.88e-03, grad_scale: 8.0 2023-05-15 21:53:02,392 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=131470.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 21:53:07,400 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=131477.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 21:53:17,926 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=131492.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 21:53:34,203 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.7058, 3.2594, 5.0908, 2.5838, 2.7990, 3.8542, 3.0852, 3.9360], device='cuda:0'), covar=tensor([0.0392, 0.1127, 0.0244, 0.1137, 0.1766, 0.1146, 0.1384, 0.0953], device='cuda:0'), in_proj_covar=tensor([0.0229, 0.0224, 0.0230, 0.0174, 0.0231, 0.0276, 0.0221, 0.0249], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:0') 2023-05-15 21:53:35,920 INFO [finetune.py:992] (0/2) Epoch 3, batch 7050, loss[loss=0.1759, simple_loss=0.2719, pruned_loss=0.03999, over 11651.00 frames. ], tot_loss[loss=0.1781, simple_loss=0.2655, pruned_loss=0.04537, over 2365808.10 frames. ], batch size: 48, lr: 4.88e-03, grad_scale: 4.0 2023-05-15 21:53:40,221 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.138e+02 2.887e+02 3.480e+02 4.466e+02 8.621e+02, threshold=6.960e+02, percent-clipped=3.0 2023-05-15 21:53:44,738 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=131530.0, num_to_drop=1, layers_to_drop={3} 2023-05-15 21:54:01,190 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=131553.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 21:54:11,244 INFO [finetune.py:992] (0/2) Epoch 3, batch 7100, loss[loss=0.1781, simple_loss=0.2583, pruned_loss=0.04894, over 12290.00 frames. ], tot_loss[loss=0.1777, simple_loss=0.2654, pruned_loss=0.04497, over 2377796.43 frames. ], batch size: 33, lr: 4.88e-03, grad_scale: 4.0 2023-05-15 21:54:48,030 INFO [finetune.py:992] (0/2) Epoch 3, batch 7150, loss[loss=0.2106, simple_loss=0.2915, pruned_loss=0.06483, over 12152.00 frames. ], tot_loss[loss=0.1768, simple_loss=0.2645, pruned_loss=0.04458, over 2378742.20 frames. ], batch size: 34, lr: 4.88e-03, grad_scale: 4.0 2023-05-15 21:54:52,330 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.052e+02 2.829e+02 3.273e+02 4.121e+02 5.730e+02, threshold=6.545e+02, percent-clipped=0.0 2023-05-15 21:55:03,137 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2023-05-15 21:55:23,463 INFO [finetune.py:992] (0/2) Epoch 3, batch 7200, loss[loss=0.1496, simple_loss=0.2342, pruned_loss=0.03251, over 12266.00 frames. ], tot_loss[loss=0.1772, simple_loss=0.2645, pruned_loss=0.04496, over 2372408.17 frames. ], batch size: 28, lr: 4.88e-03, grad_scale: 8.0 2023-05-15 21:55:58,533 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.2974, 4.9393, 5.2648, 4.5883, 4.9802, 4.5268, 5.2256, 5.0428], device='cuda:0'), covar=tensor([0.0328, 0.0389, 0.0428, 0.0280, 0.0296, 0.0343, 0.0385, 0.0273], device='cuda:0'), in_proj_covar=tensor([0.0236, 0.0241, 0.0261, 0.0235, 0.0234, 0.0236, 0.0215, 0.0202], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-05-15 21:55:59,062 INFO [finetune.py:992] (0/2) Epoch 3, batch 7250, loss[loss=0.1862, simple_loss=0.2719, pruned_loss=0.05028, over 12105.00 frames. ], tot_loss[loss=0.1774, simple_loss=0.2647, pruned_loss=0.04504, over 2373834.91 frames. ], batch size: 33, lr: 4.88e-03, grad_scale: 8.0 2023-05-15 21:55:59,965 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.4821, 5.3298, 5.4731, 5.4940, 5.1046, 5.0693, 4.8685, 5.4370], device='cuda:0'), covar=tensor([0.0704, 0.0515, 0.0603, 0.0498, 0.1626, 0.1229, 0.0565, 0.0735], device='cuda:0'), in_proj_covar=tensor([0.0487, 0.0633, 0.0541, 0.0591, 0.0772, 0.0706, 0.0516, 0.0468], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0005, 0.0004, 0.0004, 0.0005, 0.0005, 0.0004, 0.0004], device='cuda:0') 2023-05-15 21:56:03,158 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.970e+02 2.906e+02 3.464e+02 4.213e+02 7.151e+02, threshold=6.928e+02, percent-clipped=2.0 2023-05-15 21:56:33,714 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=131765.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 21:56:35,726 INFO [finetune.py:992] (0/2) Epoch 3, batch 7300, loss[loss=0.2161, simple_loss=0.2963, pruned_loss=0.06794, over 11474.00 frames. ], tot_loss[loss=0.1787, simple_loss=0.2659, pruned_loss=0.04569, over 2367903.48 frames. ], batch size: 48, lr: 4.88e-03, grad_scale: 8.0 2023-05-15 21:56:35,904 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.9736, 4.9786, 4.8422, 4.8883, 4.4813, 5.0541, 5.0708, 5.3449], device='cuda:0'), covar=tensor([0.0231, 0.0158, 0.0214, 0.0315, 0.0806, 0.0284, 0.0164, 0.0156], device='cuda:0'), in_proj_covar=tensor([0.0182, 0.0183, 0.0182, 0.0231, 0.0232, 0.0201, 0.0167, 0.0215], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0005, 0.0005, 0.0005, 0.0004, 0.0005], device='cuda:0') 2023-05-15 21:56:38,622 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=131772.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 21:57:03,491 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-05-15 21:57:11,663 INFO [finetune.py:992] (0/2) Epoch 3, batch 7350, loss[loss=0.1884, simple_loss=0.2755, pruned_loss=0.0506, over 12136.00 frames. ], tot_loss[loss=0.1781, simple_loss=0.2652, pruned_loss=0.0455, over 2366685.63 frames. ], batch size: 38, lr: 4.88e-03, grad_scale: 8.0 2023-05-15 21:57:15,909 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.193e+02 2.957e+02 3.600e+02 4.199e+02 8.110e+02, threshold=7.200e+02, percent-clipped=1.0 2023-05-15 21:57:20,334 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=131830.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 21:57:33,417 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=131848.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 21:57:39,847 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=131857.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 21:57:42,701 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.3255, 4.2923, 4.2351, 4.2723, 3.9862, 4.3902, 4.3968, 4.5800], device='cuda:0'), covar=tensor([0.0230, 0.0172, 0.0214, 0.0352, 0.0714, 0.0310, 0.0169, 0.0175], device='cuda:0'), in_proj_covar=tensor([0.0182, 0.0182, 0.0181, 0.0229, 0.0230, 0.0200, 0.0165, 0.0214], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0005, 0.0005, 0.0005, 0.0004, 0.0005], device='cuda:0') 2023-05-15 21:57:47,475 INFO [finetune.py:992] (0/2) Epoch 3, batch 7400, loss[loss=0.1541, simple_loss=0.235, pruned_loss=0.03658, over 12121.00 frames. ], tot_loss[loss=0.1782, simple_loss=0.2656, pruned_loss=0.04538, over 2369682.73 frames. ], batch size: 30, lr: 4.88e-03, grad_scale: 8.0 2023-05-15 21:57:55,319 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=131878.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 21:58:24,324 INFO [finetune.py:992] (0/2) Epoch 3, batch 7450, loss[loss=0.147, simple_loss=0.2366, pruned_loss=0.02873, over 12263.00 frames. ], tot_loss[loss=0.178, simple_loss=0.2652, pruned_loss=0.04541, over 2362769.66 frames. ], batch size: 32, lr: 4.88e-03, grad_scale: 8.0 2023-05-15 21:58:24,530 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=131918.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 21:58:28,561 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.765e+02 2.924e+02 3.430e+02 4.215e+02 1.172e+03, threshold=6.861e+02, percent-clipped=1.0 2023-05-15 21:58:59,885 INFO [finetune.py:992] (0/2) Epoch 3, batch 7500, loss[loss=0.1523, simple_loss=0.2471, pruned_loss=0.02882, over 12358.00 frames. ], tot_loss[loss=0.1786, simple_loss=0.2656, pruned_loss=0.04578, over 2354092.27 frames. ], batch size: 36, lr: 4.88e-03, grad_scale: 8.0 2023-05-15 21:59:22,743 INFO [checkpoint.py:75] (0/2) Saving checkpoint to pruned_transducer_stateless7/exp_giga_finetune/checkpoint-32000.pt 2023-05-15 21:59:39,391 INFO [finetune.py:992] (0/2) Epoch 3, batch 7550, loss[loss=0.1795, simple_loss=0.257, pruned_loss=0.05105, over 12197.00 frames. ], tot_loss[loss=0.1794, simple_loss=0.2659, pruned_loss=0.04648, over 2348999.34 frames. ], batch size: 29, lr: 4.88e-03, grad_scale: 8.0 2023-05-15 21:59:43,563 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.982e+02 3.013e+02 3.558e+02 4.169e+02 8.465e+02, threshold=7.117e+02, percent-clipped=1.0 2023-05-15 21:59:48,709 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=132031.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 22:00:05,748 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=132054.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 22:00:13,552 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=132065.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 22:00:15,544 INFO [finetune.py:992] (0/2) Epoch 3, batch 7600, loss[loss=0.1699, simple_loss=0.2483, pruned_loss=0.04576, over 11341.00 frames. ], tot_loss[loss=0.1785, simple_loss=0.2648, pruned_loss=0.04609, over 2351454.28 frames. ], batch size: 25, lr: 4.88e-03, grad_scale: 8.0 2023-05-15 22:00:18,561 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=132072.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 22:00:32,887 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=132092.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 22:00:47,648 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=132113.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 22:00:49,244 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=132115.0, num_to_drop=1, layers_to_drop={3} 2023-05-15 22:00:51,024 INFO [finetune.py:992] (0/2) Epoch 3, batch 7650, loss[loss=0.2178, simple_loss=0.302, pruned_loss=0.0668, over 12030.00 frames. ], tot_loss[loss=0.1782, simple_loss=0.2647, pruned_loss=0.04586, over 2362451.00 frames. ], batch size: 40, lr: 4.88e-03, grad_scale: 8.0 2023-05-15 22:00:52,533 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=132120.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 22:00:55,400 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.740e+02 2.907e+02 3.519e+02 4.376e+02 6.204e+02, threshold=7.039e+02, percent-clipped=0.0 2023-05-15 22:01:02,921 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=4.61 vs. limit=5.0 2023-05-15 22:01:12,544 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=132148.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 22:01:14,189 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.49 vs. limit=2.0 2023-05-15 22:01:14,734 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.4372, 4.9531, 5.4133, 4.7836, 5.0490, 4.8188, 5.4262, 5.0303], device='cuda:0'), covar=tensor([0.0244, 0.0304, 0.0220, 0.0218, 0.0312, 0.0272, 0.0183, 0.0199], device='cuda:0'), in_proj_covar=tensor([0.0238, 0.0243, 0.0265, 0.0236, 0.0237, 0.0237, 0.0218, 0.0204], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-05-15 22:01:27,165 INFO [finetune.py:992] (0/2) Epoch 3, batch 7700, loss[loss=0.155, simple_loss=0.2424, pruned_loss=0.03384, over 12352.00 frames. ], tot_loss[loss=0.1777, simple_loss=0.2649, pruned_loss=0.04531, over 2364149.67 frames. ], batch size: 30, lr: 4.88e-03, grad_scale: 8.0 2023-05-15 22:01:46,954 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=132196.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 22:01:48,620 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 2023-05-15 22:01:59,750 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=132213.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 22:02:03,244 INFO [finetune.py:992] (0/2) Epoch 3, batch 7750, loss[loss=0.1762, simple_loss=0.2653, pruned_loss=0.04358, over 12038.00 frames. ], tot_loss[loss=0.1787, simple_loss=0.266, pruned_loss=0.04571, over 2368164.76 frames. ], batch size: 37, lr: 4.88e-03, grad_scale: 8.0 2023-05-15 22:02:07,534 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.760e+02 2.978e+02 3.505e+02 4.131e+02 6.274e+02, threshold=7.010e+02, percent-clipped=0.0 2023-05-15 22:02:38,771 INFO [finetune.py:992] (0/2) Epoch 3, batch 7800, loss[loss=0.2056, simple_loss=0.2928, pruned_loss=0.05921, over 11412.00 frames. ], tot_loss[loss=0.1782, simple_loss=0.2653, pruned_loss=0.04555, over 2366584.52 frames. ], batch size: 55, lr: 4.88e-03, grad_scale: 8.0 2023-05-15 22:03:07,735 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.1328, 2.1937, 3.5533, 2.9870, 3.4029, 3.2606, 2.4173, 3.4637], device='cuda:0'), covar=tensor([0.0094, 0.0368, 0.0133, 0.0212, 0.0135, 0.0134, 0.0329, 0.0120], device='cuda:0'), in_proj_covar=tensor([0.0158, 0.0192, 0.0171, 0.0172, 0.0194, 0.0148, 0.0182, 0.0162], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-15 22:03:14,541 INFO [finetune.py:992] (0/2) Epoch 3, batch 7850, loss[loss=0.1688, simple_loss=0.2536, pruned_loss=0.04201, over 12169.00 frames. ], tot_loss[loss=0.1792, simple_loss=0.2664, pruned_loss=0.04601, over 2361557.53 frames. ], batch size: 29, lr: 4.88e-03, grad_scale: 8.0 2023-05-15 22:03:16,105 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.3241, 5.0031, 5.1525, 5.2236, 4.7511, 4.8739, 4.6019, 5.1246], device='cuda:0'), covar=tensor([0.0563, 0.0642, 0.0832, 0.0602, 0.2304, 0.1331, 0.0635, 0.1055], device='cuda:0'), in_proj_covar=tensor([0.0489, 0.0638, 0.0551, 0.0596, 0.0783, 0.0715, 0.0519, 0.0474], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0005, 0.0004, 0.0005, 0.0005, 0.0005, 0.0004, 0.0004], device='cuda:0') 2023-05-15 22:03:18,729 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.021e+02 2.804e+02 3.373e+02 4.263e+02 8.998e+02, threshold=6.746e+02, percent-clipped=4.0 2023-05-15 22:03:49,913 INFO [finetune.py:992] (0/2) Epoch 3, batch 7900, loss[loss=0.1819, simple_loss=0.2801, pruned_loss=0.04182, over 12369.00 frames. ], tot_loss[loss=0.1789, simple_loss=0.2661, pruned_loss=0.04585, over 2368148.36 frames. ], batch size: 36, lr: 4.88e-03, grad_scale: 8.0 2023-05-15 22:04:03,508 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=132387.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 22:04:08,616 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.3480, 4.8120, 2.7443, 2.4524, 3.9170, 2.3833, 4.0038, 3.1848], device='cuda:0'), covar=tensor([0.0669, 0.0376, 0.1125, 0.1595, 0.0322, 0.1509, 0.0392, 0.0862], device='cuda:0'), in_proj_covar=tensor([0.0189, 0.0246, 0.0171, 0.0196, 0.0138, 0.0177, 0.0191, 0.0171], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-15 22:04:20,199 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=132410.0, num_to_drop=1, layers_to_drop={3} 2023-05-15 22:04:25,738 INFO [finetune.py:992] (0/2) Epoch 3, batch 7950, loss[loss=0.1432, simple_loss=0.2297, pruned_loss=0.02831, over 12033.00 frames. ], tot_loss[loss=0.1769, simple_loss=0.2639, pruned_loss=0.04498, over 2379216.11 frames. ], batch size: 31, lr: 4.88e-03, grad_scale: 8.0 2023-05-15 22:04:28,336 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=2.53 vs. limit=5.0 2023-05-15 22:04:30,098 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.797e+02 2.816e+02 3.359e+02 4.037e+02 1.124e+03, threshold=6.718e+02, percent-clipped=1.0 2023-05-15 22:04:33,098 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.4493, 4.7465, 2.6601, 2.0080, 4.0713, 2.2167, 4.0529, 2.9052], device='cuda:0'), covar=tensor([0.0503, 0.0382, 0.1034, 0.1898, 0.0241, 0.1562, 0.0363, 0.0992], device='cuda:0'), in_proj_covar=tensor([0.0189, 0.0246, 0.0172, 0.0197, 0.0138, 0.0177, 0.0192, 0.0171], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-15 22:04:44,249 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.1533, 4.8379, 4.9475, 5.0998, 4.9151, 5.0978, 4.9053, 2.8197], device='cuda:0'), covar=tensor([0.0090, 0.0057, 0.0066, 0.0049, 0.0040, 0.0068, 0.0058, 0.0591], device='cuda:0'), in_proj_covar=tensor([0.0064, 0.0072, 0.0075, 0.0068, 0.0056, 0.0084, 0.0075, 0.0088], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-15 22:05:01,811 INFO [finetune.py:992] (0/2) Epoch 3, batch 8000, loss[loss=0.2981, simple_loss=0.3459, pruned_loss=0.1252, over 8036.00 frames. ], tot_loss[loss=0.1769, simple_loss=0.2639, pruned_loss=0.04493, over 2380575.00 frames. ], batch size: 100, lr: 4.88e-03, grad_scale: 8.0 2023-05-15 22:05:34,415 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=132513.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 22:05:37,691 INFO [finetune.py:992] (0/2) Epoch 3, batch 8050, loss[loss=0.2111, simple_loss=0.2941, pruned_loss=0.06398, over 12115.00 frames. ], tot_loss[loss=0.1775, simple_loss=0.2645, pruned_loss=0.04523, over 2380960.05 frames. ], batch size: 38, lr: 4.88e-03, grad_scale: 8.0 2023-05-15 22:05:41,976 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.080e+02 3.034e+02 3.557e+02 4.293e+02 8.713e+02, threshold=7.114e+02, percent-clipped=4.0 2023-05-15 22:06:05,540 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.3458, 4.2793, 4.2095, 4.2706, 3.9215, 4.3731, 4.3773, 4.5860], device='cuda:0'), covar=tensor([0.0232, 0.0174, 0.0202, 0.0313, 0.0748, 0.0288, 0.0165, 0.0169], device='cuda:0'), in_proj_covar=tensor([0.0184, 0.0184, 0.0183, 0.0232, 0.0233, 0.0201, 0.0166, 0.0217], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0005, 0.0005, 0.0005, 0.0004, 0.0005], device='cuda:0') 2023-05-15 22:06:05,894 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=4.62 vs. limit=5.0 2023-05-15 22:06:08,229 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=132561.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 22:06:13,215 INFO [finetune.py:992] (0/2) Epoch 3, batch 8100, loss[loss=0.3307, simple_loss=0.3678, pruned_loss=0.1467, over 7630.00 frames. ], tot_loss[loss=0.178, simple_loss=0.2652, pruned_loss=0.04544, over 2384237.28 frames. ], batch size: 97, lr: 4.88e-03, grad_scale: 8.0 2023-05-15 22:06:22,802 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=2.49 vs. limit=5.0 2023-05-15 22:06:49,190 INFO [finetune.py:992] (0/2) Epoch 3, batch 8150, loss[loss=0.2153, simple_loss=0.2954, pruned_loss=0.06764, over 12285.00 frames. ], tot_loss[loss=0.1781, simple_loss=0.265, pruned_loss=0.04566, over 2380801.84 frames. ], batch size: 37, lr: 4.87e-03, grad_scale: 8.0 2023-05-15 22:06:53,458 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.801e+02 2.840e+02 3.244e+02 3.868e+02 7.076e+02, threshold=6.489e+02, percent-clipped=0.0 2023-05-15 22:07:19,863 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=132660.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 22:07:25,382 INFO [finetune.py:992] (0/2) Epoch 3, batch 8200, loss[loss=0.1912, simple_loss=0.2836, pruned_loss=0.04935, over 12020.00 frames. ], tot_loss[loss=0.1777, simple_loss=0.2647, pruned_loss=0.04531, over 2384429.10 frames. ], batch size: 40, lr: 4.87e-03, grad_scale: 8.0 2023-05-15 22:07:30,717 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.2757, 3.1466, 3.2615, 3.5703, 2.7554, 3.1516, 2.5604, 3.0868], device='cuda:0'), covar=tensor([0.1282, 0.0685, 0.0811, 0.0644, 0.0819, 0.0592, 0.1316, 0.0997], device='cuda:0'), in_proj_covar=tensor([0.0221, 0.0257, 0.0290, 0.0348, 0.0238, 0.0234, 0.0252, 0.0381], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-15 22:07:39,192 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=132687.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 22:07:55,483 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=132710.0, num_to_drop=1, layers_to_drop={0} 2023-05-15 22:08:01,008 INFO [finetune.py:992] (0/2) Epoch 3, batch 8250, loss[loss=0.1648, simple_loss=0.2398, pruned_loss=0.04493, over 12000.00 frames. ], tot_loss[loss=0.1787, simple_loss=0.2654, pruned_loss=0.04603, over 2374846.60 frames. ], batch size: 28, lr: 4.87e-03, grad_scale: 8.0 2023-05-15 22:08:03,358 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=132721.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 22:08:05,194 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.911e+02 2.844e+02 3.378e+02 4.059e+02 1.339e+03, threshold=6.755e+02, percent-clipped=5.0 2023-05-15 22:08:12,931 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=132735.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 22:08:20,898 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=132746.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 22:08:23,105 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.2395, 2.1916, 2.6375, 3.2182, 2.1583, 3.2518, 3.1821, 3.3597], device='cuda:0'), covar=tensor([0.0127, 0.0896, 0.0440, 0.0139, 0.0926, 0.0266, 0.0249, 0.0098], device='cuda:0'), in_proj_covar=tensor([0.0116, 0.0198, 0.0182, 0.0113, 0.0180, 0.0171, 0.0165, 0.0118], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-15 22:08:27,185 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.3630, 5.1578, 5.2565, 5.3190, 4.8880, 4.9593, 4.7624, 5.2071], device='cuda:0'), covar=tensor([0.0638, 0.0559, 0.0767, 0.0538, 0.1945, 0.1227, 0.0581, 0.1040], device='cuda:0'), in_proj_covar=tensor([0.0480, 0.0626, 0.0541, 0.0583, 0.0769, 0.0702, 0.0509, 0.0467], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0005, 0.0004, 0.0004, 0.0005, 0.0005, 0.0004, 0.0004], device='cuda:0') 2023-05-15 22:08:29,449 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.4944, 3.5616, 3.1461, 3.1857, 2.7845, 2.8123, 3.6255, 2.0805], device='cuda:0'), covar=tensor([0.0334, 0.0139, 0.0172, 0.0159, 0.0375, 0.0278, 0.0109, 0.0475], device='cuda:0'), in_proj_covar=tensor([0.0183, 0.0154, 0.0147, 0.0175, 0.0200, 0.0192, 0.0154, 0.0194], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], device='cuda:0') 2023-05-15 22:08:29,998 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=132758.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 22:08:36,928 INFO [finetune.py:992] (0/2) Epoch 3, batch 8300, loss[loss=0.1655, simple_loss=0.2435, pruned_loss=0.04377, over 11815.00 frames. ], tot_loss[loss=0.1789, simple_loss=0.2656, pruned_loss=0.0461, over 2367988.96 frames. ], batch size: 26, lr: 4.87e-03, grad_scale: 8.0 2023-05-15 22:08:37,880 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.1564, 5.0742, 4.9702, 5.0364, 4.6788, 5.1697, 5.1175, 5.3788], device='cuda:0'), covar=tensor([0.0205, 0.0129, 0.0163, 0.0278, 0.0710, 0.0214, 0.0141, 0.0135], device='cuda:0'), in_proj_covar=tensor([0.0183, 0.0183, 0.0182, 0.0231, 0.0232, 0.0200, 0.0166, 0.0216], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0005, 0.0005, 0.0005, 0.0004, 0.0005], device='cuda:0') 2023-05-15 22:08:53,669 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.15 vs. limit=2.0 2023-05-15 22:09:05,886 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=132807.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 22:09:13,338 INFO [finetune.py:992] (0/2) Epoch 3, batch 8350, loss[loss=0.1574, simple_loss=0.2472, pruned_loss=0.03379, over 12088.00 frames. ], tot_loss[loss=0.1787, simple_loss=0.2654, pruned_loss=0.04599, over 2359425.04 frames. ], batch size: 33, lr: 4.87e-03, grad_scale: 8.0 2023-05-15 22:09:17,511 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.059e+02 2.966e+02 3.421e+02 4.323e+02 8.228e+02, threshold=6.842e+02, percent-clipped=4.0 2023-05-15 22:09:48,613 INFO [finetune.py:992] (0/2) Epoch 3, batch 8400, loss[loss=0.2663, simple_loss=0.3333, pruned_loss=0.09964, over 7746.00 frames. ], tot_loss[loss=0.1782, simple_loss=0.2652, pruned_loss=0.04562, over 2366589.45 frames. ], batch size: 97, lr: 4.87e-03, grad_scale: 8.0 2023-05-15 22:10:24,913 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=4.74 vs. limit=5.0 2023-05-15 22:10:25,111 INFO [finetune.py:992] (0/2) Epoch 3, batch 8450, loss[loss=0.1981, simple_loss=0.2787, pruned_loss=0.05879, over 11653.00 frames. ], tot_loss[loss=0.1783, simple_loss=0.2652, pruned_loss=0.04571, over 2367867.77 frames. ], batch size: 48, lr: 4.87e-03, grad_scale: 8.0 2023-05-15 22:10:25,564 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.92 vs. limit=2.0 2023-05-15 22:10:29,295 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.816e+02 2.961e+02 3.332e+02 4.096e+02 8.306e+02, threshold=6.665e+02, percent-clipped=5.0 2023-05-15 22:11:00,957 INFO [finetune.py:992] (0/2) Epoch 3, batch 8500, loss[loss=0.2055, simple_loss=0.2886, pruned_loss=0.06122, over 12155.00 frames. ], tot_loss[loss=0.1793, simple_loss=0.2662, pruned_loss=0.04622, over 2368681.66 frames. ], batch size: 34, lr: 4.87e-03, grad_scale: 8.0 2023-05-15 22:11:07,475 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.0974, 5.9608, 5.5670, 5.4878, 6.0385, 5.4019, 5.5872, 5.5277], device='cuda:0'), covar=tensor([0.1521, 0.0901, 0.0954, 0.2016, 0.0954, 0.2176, 0.1612, 0.1005], device='cuda:0'), in_proj_covar=tensor([0.0333, 0.0464, 0.0361, 0.0415, 0.0439, 0.0421, 0.0369, 0.0363], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-15 22:11:35,592 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=133016.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 22:11:36,947 INFO [finetune.py:992] (0/2) Epoch 3, batch 8550, loss[loss=0.1747, simple_loss=0.2601, pruned_loss=0.04462, over 12110.00 frames. ], tot_loss[loss=0.179, simple_loss=0.2659, pruned_loss=0.04602, over 2371470.32 frames. ], batch size: 33, lr: 4.87e-03, grad_scale: 8.0 2023-05-15 22:11:41,142 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.874e+02 2.845e+02 3.482e+02 4.287e+02 2.922e+03, threshold=6.963e+02, percent-clipped=3.0 2023-05-15 22:12:06,445 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.4426, 3.5294, 3.2685, 3.7177, 3.5285, 2.6748, 3.2711, 2.8381], device='cuda:0'), covar=tensor([0.0717, 0.0923, 0.1240, 0.0584, 0.0944, 0.1301, 0.1012, 0.2371], device='cuda:0'), in_proj_covar=tensor([0.0293, 0.0366, 0.0346, 0.0256, 0.0357, 0.0259, 0.0329, 0.0352], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-15 22:12:13,342 INFO [finetune.py:992] (0/2) Epoch 3, batch 8600, loss[loss=0.1766, simple_loss=0.27, pruned_loss=0.04164, over 12180.00 frames. ], tot_loss[loss=0.1793, simple_loss=0.2659, pruned_loss=0.04634, over 2369116.54 frames. ], batch size: 35, lr: 4.87e-03, grad_scale: 8.0 2023-05-15 22:12:29,669 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.8870, 5.7324, 5.2140, 5.3162, 5.8625, 5.1216, 5.4416, 5.3012], device='cuda:0'), covar=tensor([0.1425, 0.1026, 0.0909, 0.1902, 0.0880, 0.2174, 0.1475, 0.1189], device='cuda:0'), in_proj_covar=tensor([0.0334, 0.0463, 0.0360, 0.0416, 0.0438, 0.0418, 0.0369, 0.0363], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-15 22:12:38,108 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=133102.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 22:12:40,883 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.0451, 4.8304, 4.9606, 5.0056, 4.5834, 4.6922, 4.4665, 4.9024], device='cuda:0'), covar=tensor([0.0566, 0.0568, 0.0695, 0.0525, 0.1845, 0.1175, 0.0561, 0.0889], device='cuda:0'), in_proj_covar=tensor([0.0480, 0.0624, 0.0539, 0.0583, 0.0764, 0.0700, 0.0509, 0.0464], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0005, 0.0004, 0.0004, 0.0005, 0.0005, 0.0004, 0.0004], device='cuda:0') 2023-05-15 22:12:48,619 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.1612, 4.2363, 4.2467, 4.5673, 3.3087, 3.9735, 2.6726, 4.2321], device='cuda:0'), covar=tensor([0.1543, 0.0619, 0.0809, 0.0530, 0.0906, 0.0537, 0.1623, 0.1306], device='cuda:0'), in_proj_covar=tensor([0.0221, 0.0256, 0.0291, 0.0349, 0.0237, 0.0235, 0.0252, 0.0379], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-15 22:12:49,048 INFO [finetune.py:992] (0/2) Epoch 3, batch 8650, loss[loss=0.1725, simple_loss=0.2517, pruned_loss=0.04664, over 12167.00 frames. ], tot_loss[loss=0.1798, simple_loss=0.2669, pruned_loss=0.04637, over 2363508.15 frames. ], batch size: 29, lr: 4.87e-03, grad_scale: 8.0 2023-05-15 22:12:53,338 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.967e+02 3.020e+02 3.510e+02 4.241e+02 9.799e+02, threshold=7.020e+02, percent-clipped=1.0 2023-05-15 22:13:24,468 INFO [finetune.py:992] (0/2) Epoch 3, batch 8700, loss[loss=0.1484, simple_loss=0.232, pruned_loss=0.03239, over 11882.00 frames. ], tot_loss[loss=0.1781, simple_loss=0.2654, pruned_loss=0.04537, over 2376436.26 frames. ], batch size: 26, lr: 4.87e-03, grad_scale: 8.0 2023-05-15 22:14:01,481 INFO [finetune.py:992] (0/2) Epoch 3, batch 8750, loss[loss=0.1514, simple_loss=0.2403, pruned_loss=0.03119, over 12248.00 frames. ], tot_loss[loss=0.1783, simple_loss=0.2654, pruned_loss=0.04559, over 2369604.32 frames. ], batch size: 32, lr: 4.87e-03, grad_scale: 8.0 2023-05-15 22:14:05,853 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.998e+02 2.947e+02 3.495e+02 4.217e+02 7.819e+02, threshold=6.990e+02, percent-clipped=0.0 2023-05-15 22:14:06,948 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 2023-05-15 22:14:37,129 INFO [finetune.py:992] (0/2) Epoch 3, batch 8800, loss[loss=0.174, simple_loss=0.2575, pruned_loss=0.04524, over 12334.00 frames. ], tot_loss[loss=0.1776, simple_loss=0.2646, pruned_loss=0.04531, over 2379478.67 frames. ], batch size: 35, lr: 4.87e-03, grad_scale: 8.0 2023-05-15 22:15:11,411 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=133316.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 22:15:12,696 INFO [finetune.py:992] (0/2) Epoch 3, batch 8850, loss[loss=0.1809, simple_loss=0.2703, pruned_loss=0.04573, over 12084.00 frames. ], tot_loss[loss=0.1777, simple_loss=0.2646, pruned_loss=0.04535, over 2371874.24 frames. ], batch size: 32, lr: 4.87e-03, grad_scale: 8.0 2023-05-15 22:15:16,896 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.926e+02 2.778e+02 3.188e+02 3.991e+02 7.101e+02, threshold=6.375e+02, percent-clipped=2.0 2023-05-15 22:15:33,569 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=2.24 vs. limit=5.0 2023-05-15 22:15:46,747 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=133364.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 22:15:48,902 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([6.0011, 5.9674, 5.7688, 5.4506, 5.1282, 5.9492, 5.5029, 5.3324], device='cuda:0'), covar=tensor([0.0740, 0.0811, 0.0572, 0.1268, 0.0692, 0.0633, 0.1424, 0.1000], device='cuda:0'), in_proj_covar=tensor([0.0566, 0.0509, 0.0472, 0.0581, 0.0383, 0.0662, 0.0715, 0.0523], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0004, 0.0004, 0.0003], device='cuda:0') 2023-05-15 22:15:49,503 INFO [finetune.py:992] (0/2) Epoch 3, batch 8900, loss[loss=0.1468, simple_loss=0.2415, pruned_loss=0.02605, over 12157.00 frames. ], tot_loss[loss=0.1766, simple_loss=0.2639, pruned_loss=0.04467, over 2382800.22 frames. ], batch size: 34, lr: 4.87e-03, grad_scale: 8.0 2023-05-15 22:15:53,983 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=133374.0, num_to_drop=1, layers_to_drop={0} 2023-05-15 22:15:55,355 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=133376.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 22:16:14,019 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=133402.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 22:16:17,752 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=4.76 vs. limit=5.0 2023-05-15 22:16:25,126 INFO [finetune.py:992] (0/2) Epoch 3, batch 8950, loss[loss=0.1669, simple_loss=0.2564, pruned_loss=0.03867, over 12295.00 frames. ], tot_loss[loss=0.1767, simple_loss=0.2641, pruned_loss=0.04465, over 2379659.80 frames. ], batch size: 33, lr: 4.87e-03, grad_scale: 8.0 2023-05-15 22:16:26,007 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.5568, 4.4812, 4.3530, 4.4833, 4.1023, 4.5956, 4.5972, 4.8042], device='cuda:0'), covar=tensor([0.0233, 0.0168, 0.0227, 0.0309, 0.0730, 0.0265, 0.0169, 0.0171], device='cuda:0'), in_proj_covar=tensor([0.0182, 0.0180, 0.0182, 0.0229, 0.0229, 0.0199, 0.0165, 0.0212], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0005, 0.0005, 0.0005, 0.0004, 0.0005], device='cuda:0') 2023-05-15 22:16:29,468 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.233e+02 2.921e+02 3.651e+02 4.216e+02 9.647e+02, threshold=7.301e+02, percent-clipped=4.0 2023-05-15 22:16:37,778 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=133435.0, num_to_drop=1, layers_to_drop={0} 2023-05-15 22:16:39,178 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=133437.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 22:16:39,974 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=2.79 vs. limit=5.0 2023-05-15 22:16:44,043 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.2435, 2.0023, 2.4284, 2.2288, 2.4024, 2.4079, 1.9132, 2.4148], device='cuda:0'), covar=tensor([0.0086, 0.0224, 0.0141, 0.0160, 0.0125, 0.0118, 0.0221, 0.0104], device='cuda:0'), in_proj_covar=tensor([0.0160, 0.0193, 0.0171, 0.0174, 0.0194, 0.0148, 0.0184, 0.0164], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-15 22:16:47,945 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=133450.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 22:17:00,526 INFO [finetune.py:992] (0/2) Epoch 3, batch 9000, loss[loss=0.1843, simple_loss=0.2775, pruned_loss=0.04558, over 12296.00 frames. ], tot_loss[loss=0.1781, simple_loss=0.2655, pruned_loss=0.04535, over 2378085.33 frames. ], batch size: 33, lr: 4.87e-03, grad_scale: 8.0 2023-05-15 22:17:00,527 INFO [finetune.py:1017] (0/2) Computing validation loss 2023-05-15 22:17:20,672 INFO [finetune.py:1026] (0/2) Epoch 3, validation: loss=0.3381, simple_loss=0.4084, pruned_loss=0.1339, over 1020973.00 frames. 2023-05-15 22:17:20,673 INFO [finetune.py:1027] (0/2) Maximum memory allocated so far is 12745MB 2023-05-15 22:17:25,212 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=133474.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 22:17:56,194 INFO [finetune.py:992] (0/2) Epoch 3, batch 9050, loss[loss=0.1854, simple_loss=0.2717, pruned_loss=0.04952, over 11131.00 frames. ], tot_loss[loss=0.1788, simple_loss=0.2659, pruned_loss=0.04584, over 2370246.80 frames. ], batch size: 55, lr: 4.87e-03, grad_scale: 16.0 2023-05-15 22:18:00,451 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.048e+02 2.817e+02 3.375e+02 4.169e+02 1.335e+03, threshold=6.751e+02, percent-clipped=5.0 2023-05-15 22:18:08,500 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=133535.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 22:18:24,935 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.39 vs. limit=2.0 2023-05-15 22:18:31,245 INFO [finetune.py:992] (0/2) Epoch 3, batch 9100, loss[loss=0.1875, simple_loss=0.2741, pruned_loss=0.0504, over 11735.00 frames. ], tot_loss[loss=0.1791, simple_loss=0.2662, pruned_loss=0.04599, over 2370682.68 frames. ], batch size: 44, lr: 4.87e-03, grad_scale: 16.0 2023-05-15 22:18:45,817 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.6730, 3.0180, 5.0426, 2.6637, 2.7518, 3.8428, 3.3625, 3.9526], device='cuda:0'), covar=tensor([0.0387, 0.1221, 0.0278, 0.1088, 0.1758, 0.1256, 0.1181, 0.1017], device='cuda:0'), in_proj_covar=tensor([0.0234, 0.0227, 0.0234, 0.0176, 0.0233, 0.0278, 0.0223, 0.0249], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:0') 2023-05-15 22:18:55,155 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.4112, 4.7215, 2.7988, 2.7420, 3.9295, 2.6378, 3.9981, 3.2788], device='cuda:0'), covar=tensor([0.0575, 0.0410, 0.1064, 0.1308, 0.0303, 0.1235, 0.0416, 0.0735], device='cuda:0'), in_proj_covar=tensor([0.0188, 0.0248, 0.0174, 0.0196, 0.0138, 0.0178, 0.0192, 0.0172], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-15 22:18:59,951 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.5755, 2.2935, 3.1315, 4.4272, 2.6252, 4.4152, 4.5496, 4.7082], device='cuda:0'), covar=tensor([0.0117, 0.1196, 0.0475, 0.0137, 0.1025, 0.0182, 0.0164, 0.0079], device='cuda:0'), in_proj_covar=tensor([0.0117, 0.0199, 0.0183, 0.0114, 0.0182, 0.0173, 0.0167, 0.0119], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-15 22:19:08,105 INFO [finetune.py:992] (0/2) Epoch 3, batch 9150, loss[loss=0.161, simple_loss=0.2537, pruned_loss=0.03419, over 12263.00 frames. ], tot_loss[loss=0.1784, simple_loss=0.2657, pruned_loss=0.0456, over 2377402.20 frames. ], batch size: 32, lr: 4.87e-03, grad_scale: 16.0 2023-05-15 22:19:12,163 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.012e+02 2.917e+02 3.668e+02 4.519e+02 9.899e+02, threshold=7.336e+02, percent-clipped=4.0 2023-05-15 22:19:32,049 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=133652.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 22:19:43,290 INFO [finetune.py:992] (0/2) Epoch 3, batch 9200, loss[loss=0.2027, simple_loss=0.2899, pruned_loss=0.05777, over 10702.00 frames. ], tot_loss[loss=0.1779, simple_loss=0.2652, pruned_loss=0.04526, over 2376526.41 frames. ], batch size: 68, lr: 4.87e-03, grad_scale: 16.0 2023-05-15 22:20:07,913 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.5859, 3.6780, 3.3059, 3.2775, 2.9238, 2.8687, 3.6874, 2.2975], device='cuda:0'), covar=tensor([0.0310, 0.0116, 0.0155, 0.0137, 0.0374, 0.0296, 0.0107, 0.0417], device='cuda:0'), in_proj_covar=tensor([0.0181, 0.0153, 0.0147, 0.0172, 0.0199, 0.0190, 0.0153, 0.0193], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], device='cuda:0') 2023-05-15 22:20:15,895 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=133713.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 22:20:19,356 INFO [finetune.py:992] (0/2) Epoch 3, batch 9250, loss[loss=0.2821, simple_loss=0.3363, pruned_loss=0.1139, over 7825.00 frames. ], tot_loss[loss=0.1765, simple_loss=0.2641, pruned_loss=0.04442, over 2378572.55 frames. ], batch size: 97, lr: 4.87e-03, grad_scale: 8.0 2023-05-15 22:20:25,100 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.930e+02 2.732e+02 3.186e+02 3.762e+02 2.059e+03, threshold=6.372e+02, percent-clipped=2.0 2023-05-15 22:20:28,702 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=133730.0, num_to_drop=1, layers_to_drop={3} 2023-05-15 22:20:30,151 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=133732.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 22:20:56,405 INFO [finetune.py:992] (0/2) Epoch 3, batch 9300, loss[loss=0.2591, simple_loss=0.3187, pruned_loss=0.0997, over 8369.00 frames. ], tot_loss[loss=0.1778, simple_loss=0.2651, pruned_loss=0.04528, over 2366201.25 frames. ], batch size: 97, lr: 4.87e-03, grad_scale: 8.0 2023-05-15 22:21:07,156 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.3458, 3.4692, 3.1809, 3.0832, 2.8045, 2.6066, 3.4371, 2.1464], device='cuda:0'), covar=tensor([0.0338, 0.0116, 0.0135, 0.0147, 0.0324, 0.0297, 0.0105, 0.0411], device='cuda:0'), in_proj_covar=tensor([0.0181, 0.0153, 0.0147, 0.0173, 0.0198, 0.0190, 0.0153, 0.0193], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], device='cuda:0') 2023-05-15 22:21:07,209 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.5081, 2.5468, 4.3851, 4.7559, 3.1399, 2.4652, 2.7596, 1.8834], device='cuda:0'), covar=tensor([0.1397, 0.3130, 0.0416, 0.0279, 0.0940, 0.2162, 0.2488, 0.4139], device='cuda:0'), in_proj_covar=tensor([0.0268, 0.0370, 0.0262, 0.0283, 0.0249, 0.0275, 0.0343, 0.0347], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-15 22:21:11,387 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.49 vs. limit=2.0 2023-05-15 22:21:31,717 INFO [finetune.py:992] (0/2) Epoch 3, batch 9350, loss[loss=0.1662, simple_loss=0.2619, pruned_loss=0.03525, over 12151.00 frames. ], tot_loss[loss=0.1783, simple_loss=0.2659, pruned_loss=0.04531, over 2366796.64 frames. ], batch size: 34, lr: 4.87e-03, grad_scale: 8.0 2023-05-15 22:21:36,743 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.088e+02 2.993e+02 3.452e+02 4.472e+02 7.220e+02, threshold=6.904e+02, percent-clipped=5.0 2023-05-15 22:21:40,255 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=133830.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 22:21:58,846 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2023-05-15 22:22:07,701 INFO [finetune.py:992] (0/2) Epoch 3, batch 9400, loss[loss=0.1693, simple_loss=0.247, pruned_loss=0.04581, over 12170.00 frames. ], tot_loss[loss=0.1785, simple_loss=0.2661, pruned_loss=0.04547, over 2356504.91 frames. ], batch size: 31, lr: 4.87e-03, grad_scale: 8.0 2023-05-15 22:22:43,384 INFO [finetune.py:992] (0/2) Epoch 3, batch 9450, loss[loss=0.1665, simple_loss=0.2622, pruned_loss=0.03538, over 12349.00 frames. ], tot_loss[loss=0.1786, simple_loss=0.2665, pruned_loss=0.04539, over 2362414.36 frames. ], batch size: 36, lr: 4.87e-03, grad_scale: 8.0 2023-05-15 22:22:48,354 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.031e+02 2.929e+02 3.427e+02 4.452e+02 8.212e+02, threshold=6.855e+02, percent-clipped=3.0 2023-05-15 22:23:06,120 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([6.1833, 6.1039, 5.9875, 5.4336, 5.2240, 6.0924, 5.6649, 5.5740], device='cuda:0'), covar=tensor([0.0574, 0.0914, 0.0556, 0.1458, 0.0645, 0.0648, 0.1378, 0.0817], device='cuda:0'), in_proj_covar=tensor([0.0575, 0.0517, 0.0479, 0.0588, 0.0388, 0.0668, 0.0723, 0.0533], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0004, 0.0004, 0.0003], device='cuda:0') 2023-05-15 22:23:18,711 INFO [finetune.py:992] (0/2) Epoch 3, batch 9500, loss[loss=0.1904, simple_loss=0.2686, pruned_loss=0.05604, over 11171.00 frames. ], tot_loss[loss=0.1775, simple_loss=0.2653, pruned_loss=0.04486, over 2368383.31 frames. ], batch size: 55, lr: 4.86e-03, grad_scale: 8.0 2023-05-15 22:23:39,412 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=133997.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 22:23:41,606 INFO [checkpoint.py:75] (0/2) Saving checkpoint to pruned_transducer_stateless7/exp_giga_finetune/checkpoint-34000.pt 2023-05-15 22:23:51,093 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=134008.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 22:23:54,437 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=4.65 vs. limit=5.0 2023-05-15 22:23:58,376 INFO [finetune.py:992] (0/2) Epoch 3, batch 9550, loss[loss=0.1635, simple_loss=0.2443, pruned_loss=0.04141, over 12014.00 frames. ], tot_loss[loss=0.177, simple_loss=0.2644, pruned_loss=0.04481, over 2364757.89 frames. ], batch size: 28, lr: 4.86e-03, grad_scale: 8.0 2023-05-15 22:24:03,374 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.739e+02 3.023e+02 3.557e+02 4.480e+02 3.253e+03, threshold=7.114e+02, percent-clipped=3.0 2023-05-15 22:24:07,115 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=134030.0, num_to_drop=1, layers_to_drop={1} 2023-05-15 22:24:08,600 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=134032.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 22:24:18,281 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.74 vs. limit=2.0 2023-05-15 22:24:27,677 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=134058.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 22:24:34,663 INFO [finetune.py:992] (0/2) Epoch 3, batch 9600, loss[loss=0.1612, simple_loss=0.2584, pruned_loss=0.03201, over 12181.00 frames. ], tot_loss[loss=0.1772, simple_loss=0.2645, pruned_loss=0.04493, over 2364829.62 frames. ], batch size: 35, lr: 4.86e-03, grad_scale: 8.0 2023-05-15 22:24:41,991 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=134078.0, num_to_drop=1, layers_to_drop={0} 2023-05-15 22:24:43,334 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=134080.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 22:24:52,677 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.4061, 5.1383, 5.4094, 5.3444, 4.5670, 4.6207, 4.8190, 5.1628], device='cuda:0'), covar=tensor([0.0817, 0.0976, 0.0733, 0.0907, 0.3319, 0.2104, 0.0724, 0.1548], device='cuda:0'), in_proj_covar=tensor([0.0482, 0.0632, 0.0540, 0.0594, 0.0771, 0.0705, 0.0514, 0.0465], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0005, 0.0004, 0.0004, 0.0005, 0.0005, 0.0004, 0.0004], device='cuda:0') 2023-05-15 22:25:10,211 INFO [finetune.py:992] (0/2) Epoch 3, batch 9650, loss[loss=0.1987, simple_loss=0.2835, pruned_loss=0.05696, over 12110.00 frames. ], tot_loss[loss=0.1769, simple_loss=0.2643, pruned_loss=0.04475, over 2367575.86 frames. ], batch size: 38, lr: 4.86e-03, grad_scale: 8.0 2023-05-15 22:25:14,035 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 2023-05-15 22:25:15,051 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.605e+02 2.666e+02 3.292e+02 4.068e+02 8.915e+02, threshold=6.584e+02, percent-clipped=1.0 2023-05-15 22:25:18,759 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=134130.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 22:25:46,047 INFO [finetune.py:992] (0/2) Epoch 3, batch 9700, loss[loss=0.1851, simple_loss=0.2732, pruned_loss=0.04852, over 12063.00 frames. ], tot_loss[loss=0.176, simple_loss=0.2633, pruned_loss=0.04433, over 2366622.34 frames. ], batch size: 42, lr: 4.86e-03, grad_scale: 8.0 2023-05-15 22:25:53,289 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=134178.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 22:26:11,999 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.9026, 4.5664, 4.5707, 4.7533, 4.6035, 4.8238, 4.6324, 2.5060], device='cuda:0'), covar=tensor([0.0102, 0.0064, 0.0085, 0.0065, 0.0049, 0.0084, 0.0079, 0.0717], device='cuda:0'), in_proj_covar=tensor([0.0065, 0.0073, 0.0077, 0.0070, 0.0057, 0.0087, 0.0076, 0.0090], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-15 22:26:19,064 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.9652, 5.9425, 5.7180, 5.2780, 5.1593, 5.8740, 5.4527, 5.2910], device='cuda:0'), covar=tensor([0.0631, 0.0733, 0.0560, 0.1455, 0.0622, 0.0666, 0.1423, 0.0917], device='cuda:0'), in_proj_covar=tensor([0.0575, 0.0516, 0.0481, 0.0591, 0.0387, 0.0667, 0.0724, 0.0536], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0004, 0.0004, 0.0003], device='cuda:0') 2023-05-15 22:26:22,407 INFO [finetune.py:992] (0/2) Epoch 3, batch 9750, loss[loss=0.2762, simple_loss=0.3267, pruned_loss=0.1128, over 8365.00 frames. ], tot_loss[loss=0.1777, simple_loss=0.2647, pruned_loss=0.04536, over 2359176.43 frames. ], batch size: 98, lr: 4.86e-03, grad_scale: 8.0 2023-05-15 22:26:27,473 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.987e+02 2.841e+02 3.593e+02 4.266e+02 1.145e+03, threshold=7.185e+02, percent-clipped=4.0 2023-05-15 22:26:57,989 INFO [finetune.py:992] (0/2) Epoch 3, batch 9800, loss[loss=0.1767, simple_loss=0.2751, pruned_loss=0.03912, over 12014.00 frames. ], tot_loss[loss=0.1771, simple_loss=0.2642, pruned_loss=0.04501, over 2365608.78 frames. ], batch size: 42, lr: 4.86e-03, grad_scale: 8.0 2023-05-15 22:27:24,362 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=134304.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 22:27:27,298 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=134308.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 22:27:27,963 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=134309.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 22:27:34,154 INFO [finetune.py:992] (0/2) Epoch 3, batch 9850, loss[loss=0.1587, simple_loss=0.2426, pruned_loss=0.03742, over 12356.00 frames. ], tot_loss[loss=0.1778, simple_loss=0.2652, pruned_loss=0.04522, over 2369153.17 frames. ], batch size: 30, lr: 4.86e-03, grad_scale: 8.0 2023-05-15 22:27:39,275 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.082e+02 3.023e+02 3.597e+02 4.284e+02 7.979e+02, threshold=7.194e+02, percent-clipped=2.0 2023-05-15 22:27:59,977 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=134353.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 22:28:02,154 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=134356.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 22:28:07,612 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.33 vs. limit=2.0 2023-05-15 22:28:08,730 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=134365.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 22:28:10,637 INFO [finetune.py:992] (0/2) Epoch 3, batch 9900, loss[loss=0.1864, simple_loss=0.2831, pruned_loss=0.04483, over 12041.00 frames. ], tot_loss[loss=0.1772, simple_loss=0.2647, pruned_loss=0.04485, over 2373797.38 frames. ], batch size: 40, lr: 4.86e-03, grad_scale: 8.0 2023-05-15 22:28:12,247 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=134370.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 22:28:46,538 INFO [finetune.py:992] (0/2) Epoch 3, batch 9950, loss[loss=0.2289, simple_loss=0.3105, pruned_loss=0.07368, over 11230.00 frames. ], tot_loss[loss=0.1788, simple_loss=0.2662, pruned_loss=0.04569, over 2363164.49 frames. ], batch size: 55, lr: 4.86e-03, grad_scale: 8.0 2023-05-15 22:28:50,453 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2023-05-15 22:28:51,550 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.944e+02 2.773e+02 3.327e+02 3.836e+02 9.753e+02, threshold=6.655e+02, percent-clipped=1.0 2023-05-15 22:29:22,951 INFO [finetune.py:992] (0/2) Epoch 3, batch 10000, loss[loss=0.2765, simple_loss=0.3395, pruned_loss=0.1068, over 8429.00 frames. ], tot_loss[loss=0.1782, simple_loss=0.2654, pruned_loss=0.04552, over 2370968.55 frames. ], batch size: 97, lr: 4.86e-03, grad_scale: 8.0 2023-05-15 22:29:59,099 INFO [finetune.py:992] (0/2) Epoch 3, batch 10050, loss[loss=0.1797, simple_loss=0.2739, pruned_loss=0.04277, over 11656.00 frames. ], tot_loss[loss=0.1772, simple_loss=0.2647, pruned_loss=0.04482, over 2379490.98 frames. ], batch size: 48, lr: 4.86e-03, grad_scale: 8.0 2023-05-15 22:30:04,120 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.757e+02 2.679e+02 3.237e+02 3.824e+02 1.026e+03, threshold=6.474e+02, percent-clipped=1.0 2023-05-15 22:30:29,271 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.44 vs. limit=2.0 2023-05-15 22:30:34,530 INFO [finetune.py:992] (0/2) Epoch 3, batch 10100, loss[loss=0.1424, simple_loss=0.2217, pruned_loss=0.03154, over 12181.00 frames. ], tot_loss[loss=0.1764, simple_loss=0.264, pruned_loss=0.04437, over 2387759.89 frames. ], batch size: 29, lr: 4.86e-03, grad_scale: 8.0 2023-05-15 22:31:11,410 INFO [finetune.py:992] (0/2) Epoch 3, batch 10150, loss[loss=0.166, simple_loss=0.2442, pruned_loss=0.0439, over 11823.00 frames. ], tot_loss[loss=0.1762, simple_loss=0.2636, pruned_loss=0.04438, over 2381323.16 frames. ], batch size: 26, lr: 4.86e-03, grad_scale: 8.0 2023-05-15 22:31:16,371 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.738e+02 2.878e+02 3.408e+02 4.078e+02 1.033e+03, threshold=6.816e+02, percent-clipped=2.0 2023-05-15 22:31:37,523 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=134653.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 22:31:42,473 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=134660.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 22:31:45,982 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=134665.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 22:31:47,997 INFO [finetune.py:992] (0/2) Epoch 3, batch 10200, loss[loss=0.1676, simple_loss=0.2585, pruned_loss=0.0383, over 12295.00 frames. ], tot_loss[loss=0.1771, simple_loss=0.2645, pruned_loss=0.04481, over 2383071.32 frames. ], batch size: 37, lr: 4.86e-03, grad_scale: 8.0 2023-05-15 22:32:07,199 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.3407, 6.2178, 5.6660, 5.7721, 6.1912, 5.7090, 5.8935, 5.7981], device='cuda:0'), covar=tensor([0.1402, 0.0752, 0.1083, 0.1946, 0.0848, 0.1968, 0.1408, 0.0876], device='cuda:0'), in_proj_covar=tensor([0.0334, 0.0470, 0.0363, 0.0417, 0.0442, 0.0419, 0.0371, 0.0365], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-15 22:32:11,336 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=134701.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 22:32:23,037 INFO [finetune.py:992] (0/2) Epoch 3, batch 10250, loss[loss=0.1512, simple_loss=0.2237, pruned_loss=0.03941, over 12007.00 frames. ], tot_loss[loss=0.1762, simple_loss=0.2635, pruned_loss=0.04438, over 2386186.95 frames. ], batch size: 28, lr: 4.86e-03, grad_scale: 8.0 2023-05-15 22:32:27,935 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.956e+02 2.962e+02 3.516e+02 4.268e+02 7.106e+02, threshold=7.031e+02, percent-clipped=2.0 2023-05-15 22:32:59,184 INFO [finetune.py:992] (0/2) Epoch 3, batch 10300, loss[loss=0.1918, simple_loss=0.2803, pruned_loss=0.05172, over 12345.00 frames. ], tot_loss[loss=0.1762, simple_loss=0.2636, pruned_loss=0.0444, over 2378699.82 frames. ], batch size: 35, lr: 4.86e-03, grad_scale: 8.0 2023-05-15 22:33:03,080 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.5205, 2.4880, 3.3028, 4.3417, 2.3637, 4.4419, 4.5145, 4.6068], device='cuda:0'), covar=tensor([0.0110, 0.1208, 0.0430, 0.0128, 0.1250, 0.0168, 0.0114, 0.0069], device='cuda:0'), in_proj_covar=tensor([0.0119, 0.0204, 0.0185, 0.0117, 0.0187, 0.0175, 0.0171, 0.0121], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-15 22:33:35,688 INFO [finetune.py:992] (0/2) Epoch 3, batch 10350, loss[loss=0.1633, simple_loss=0.2489, pruned_loss=0.03886, over 12268.00 frames. ], tot_loss[loss=0.1755, simple_loss=0.2632, pruned_loss=0.04393, over 2378210.71 frames. ], batch size: 28, lr: 4.86e-03, grad_scale: 8.0 2023-05-15 22:33:40,539 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.955e+02 2.825e+02 3.250e+02 4.015e+02 7.345e+02, threshold=6.499e+02, percent-clipped=1.0 2023-05-15 22:34:02,792 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.45 vs. limit=2.0 2023-05-15 22:34:07,244 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=134862.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 22:34:11,316 INFO [finetune.py:992] (0/2) Epoch 3, batch 10400, loss[loss=0.1742, simple_loss=0.2665, pruned_loss=0.04098, over 12291.00 frames. ], tot_loss[loss=0.1757, simple_loss=0.2635, pruned_loss=0.04393, over 2379995.89 frames. ], batch size: 33, lr: 4.86e-03, grad_scale: 8.0 2023-05-15 22:34:26,403 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.6074, 4.4539, 4.5705, 4.5944, 4.2266, 4.3127, 4.1841, 4.4832], device='cuda:0'), covar=tensor([0.0543, 0.0554, 0.0692, 0.0552, 0.1692, 0.1114, 0.0509, 0.0921], device='cuda:0'), in_proj_covar=tensor([0.0487, 0.0640, 0.0542, 0.0597, 0.0776, 0.0707, 0.0519, 0.0470], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0005, 0.0004, 0.0004, 0.0005, 0.0005, 0.0004, 0.0004], device='cuda:0') 2023-05-15 22:34:47,498 INFO [finetune.py:992] (0/2) Epoch 3, batch 10450, loss[loss=0.1612, simple_loss=0.2588, pruned_loss=0.03186, over 12277.00 frames. ], tot_loss[loss=0.1768, simple_loss=0.2646, pruned_loss=0.04453, over 2365667.99 frames. ], batch size: 37, lr: 4.86e-03, grad_scale: 8.0 2023-05-15 22:34:51,218 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=134923.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 22:34:52,406 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.107e+02 2.965e+02 3.504e+02 4.131e+02 9.829e+02, threshold=7.008e+02, percent-clipped=3.0 2023-05-15 22:35:06,129 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.1834, 4.4724, 3.9902, 4.8488, 4.4764, 2.8930, 4.2091, 2.9399], device='cuda:0'), covar=tensor([0.0767, 0.0736, 0.1281, 0.0347, 0.0961, 0.1520, 0.0864, 0.3092], device='cuda:0'), in_proj_covar=tensor([0.0293, 0.0364, 0.0346, 0.0258, 0.0356, 0.0260, 0.0330, 0.0351], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-15 22:35:18,031 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=134960.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 22:35:21,478 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=134965.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 22:35:23,489 INFO [finetune.py:992] (0/2) Epoch 3, batch 10500, loss[loss=0.1644, simple_loss=0.2455, pruned_loss=0.04165, over 11984.00 frames. ], tot_loss[loss=0.1766, simple_loss=0.2642, pruned_loss=0.04443, over 2370033.79 frames. ], batch size: 28, lr: 4.86e-03, grad_scale: 8.0 2023-05-15 22:35:52,439 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=135008.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 22:35:56,064 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=135013.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 22:35:56,367 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2023-05-15 22:35:59,387 INFO [finetune.py:992] (0/2) Epoch 3, batch 10550, loss[loss=0.1536, simple_loss=0.2437, pruned_loss=0.03178, over 12094.00 frames. ], tot_loss[loss=0.1778, simple_loss=0.2654, pruned_loss=0.04512, over 2353260.19 frames. ], batch size: 32, lr: 4.86e-03, grad_scale: 8.0 2023-05-15 22:36:04,296 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.006e+02 2.851e+02 3.385e+02 4.106e+02 8.437e+02, threshold=6.769e+02, percent-clipped=3.0 2023-05-15 22:36:35,664 INFO [finetune.py:992] (0/2) Epoch 3, batch 10600, loss[loss=0.1858, simple_loss=0.2764, pruned_loss=0.04763, over 12135.00 frames. ], tot_loss[loss=0.1767, simple_loss=0.2643, pruned_loss=0.0446, over 2355684.63 frames. ], batch size: 39, lr: 4.86e-03, grad_scale: 8.0 2023-05-15 22:36:55,099 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.1868, 2.0950, 2.6388, 3.1587, 2.1711, 3.2508, 3.1707, 3.3164], device='cuda:0'), covar=tensor([0.0119, 0.0949, 0.0418, 0.0158, 0.0931, 0.0255, 0.0232, 0.0110], device='cuda:0'), in_proj_covar=tensor([0.0118, 0.0202, 0.0183, 0.0115, 0.0184, 0.0173, 0.0168, 0.0120], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-15 22:37:02,254 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.94 vs. limit=2.0 2023-05-15 22:37:11,783 INFO [finetune.py:992] (0/2) Epoch 3, batch 10650, loss[loss=0.1627, simple_loss=0.2607, pruned_loss=0.03231, over 12158.00 frames. ], tot_loss[loss=0.1769, simple_loss=0.2644, pruned_loss=0.04472, over 2355168.27 frames. ], batch size: 36, lr: 4.86e-03, grad_scale: 8.0 2023-05-15 22:37:16,768 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.734e+02 2.600e+02 3.182e+02 4.069e+02 7.657e+02, threshold=6.365e+02, percent-clipped=1.0 2023-05-15 22:37:26,305 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.2555, 2.0060, 3.6412, 4.2856, 3.8654, 4.0675, 3.8524, 2.8377], device='cuda:0'), covar=tensor([0.0037, 0.0486, 0.0112, 0.0035, 0.0116, 0.0074, 0.0081, 0.0357], device='cuda:0'), in_proj_covar=tensor([0.0080, 0.0118, 0.0101, 0.0074, 0.0099, 0.0109, 0.0087, 0.0132], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-15 22:37:47,388 INFO [finetune.py:992] (0/2) Epoch 3, batch 10700, loss[loss=0.1754, simple_loss=0.2637, pruned_loss=0.04356, over 11645.00 frames. ], tot_loss[loss=0.1775, simple_loss=0.2649, pruned_loss=0.04507, over 2348823.47 frames. ], batch size: 48, lr: 4.86e-03, grad_scale: 8.0 2023-05-15 22:37:49,703 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=135171.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 22:38:24,697 INFO [finetune.py:992] (0/2) Epoch 3, batch 10750, loss[loss=0.1921, simple_loss=0.2823, pruned_loss=0.05101, over 10478.00 frames. ], tot_loss[loss=0.178, simple_loss=0.2657, pruned_loss=0.04515, over 2353740.56 frames. ], batch size: 68, lr: 4.86e-03, grad_scale: 8.0 2023-05-15 22:38:24,785 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=135218.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 22:38:29,766 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.207e+02 3.060e+02 3.393e+02 4.113e+02 1.124e+03, threshold=6.785e+02, percent-clipped=1.0 2023-05-15 22:38:34,928 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=135232.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 22:39:00,493 INFO [finetune.py:992] (0/2) Epoch 3, batch 10800, loss[loss=0.1467, simple_loss=0.2284, pruned_loss=0.03252, over 11978.00 frames. ], tot_loss[loss=0.177, simple_loss=0.2647, pruned_loss=0.04466, over 2360461.22 frames. ], batch size: 28, lr: 4.86e-03, grad_scale: 8.0 2023-05-15 22:39:36,144 INFO [finetune.py:992] (0/2) Epoch 3, batch 10850, loss[loss=0.1565, simple_loss=0.2398, pruned_loss=0.03658, over 11809.00 frames. ], tot_loss[loss=0.1766, simple_loss=0.2643, pruned_loss=0.04443, over 2366372.01 frames. ], batch size: 26, lr: 4.85e-03, grad_scale: 8.0 2023-05-15 22:39:41,720 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.242e+02 2.957e+02 3.427e+02 4.068e+02 7.800e+02, threshold=6.853e+02, percent-clipped=3.0 2023-05-15 22:39:42,780 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.4786, 2.3428, 3.6348, 4.4057, 3.8737, 4.3339, 3.7877, 3.1424], device='cuda:0'), covar=tensor([0.0025, 0.0373, 0.0129, 0.0032, 0.0123, 0.0065, 0.0093, 0.0284], device='cuda:0'), in_proj_covar=tensor([0.0080, 0.0118, 0.0100, 0.0074, 0.0098, 0.0108, 0.0087, 0.0131], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-15 22:40:13,861 INFO [finetune.py:992] (0/2) Epoch 3, batch 10900, loss[loss=0.2896, simple_loss=0.3585, pruned_loss=0.1103, over 7770.00 frames. ], tot_loss[loss=0.1762, simple_loss=0.2638, pruned_loss=0.04426, over 2372814.41 frames. ], batch size: 101, lr: 4.85e-03, grad_scale: 8.0 2023-05-15 22:40:41,915 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.1862, 2.2320, 3.0751, 4.0696, 1.9032, 4.1663, 4.1805, 4.3127], device='cuda:0'), covar=tensor([0.0136, 0.1183, 0.0409, 0.0111, 0.1252, 0.0146, 0.0127, 0.0073], device='cuda:0'), in_proj_covar=tensor([0.0119, 0.0202, 0.0184, 0.0116, 0.0186, 0.0174, 0.0169, 0.0121], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-15 22:40:49,337 INFO [finetune.py:992] (0/2) Epoch 3, batch 10950, loss[loss=0.1645, simple_loss=0.2428, pruned_loss=0.04309, over 12332.00 frames. ], tot_loss[loss=0.1776, simple_loss=0.2652, pruned_loss=0.04495, over 2375637.86 frames. ], batch size: 31, lr: 4.85e-03, grad_scale: 8.0 2023-05-15 22:40:54,097 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.012e+02 2.889e+02 3.575e+02 4.378e+02 6.782e+02, threshold=7.149e+02, percent-clipped=0.0 2023-05-15 22:41:02,016 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([6.1020, 6.0834, 5.8587, 5.4065, 5.1798, 6.0307, 5.6293, 5.4826], device='cuda:0'), covar=tensor([0.0690, 0.0780, 0.0593, 0.1405, 0.0573, 0.0564, 0.1340, 0.0849], device='cuda:0'), in_proj_covar=tensor([0.0572, 0.0512, 0.0481, 0.0587, 0.0385, 0.0659, 0.0720, 0.0527], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0004, 0.0004, 0.0003], device='cuda:0') 2023-05-15 22:41:14,562 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.0497, 2.2382, 2.2737, 2.2621, 1.9812, 1.8877, 2.1564, 1.7037], device='cuda:0'), covar=tensor([0.0291, 0.0173, 0.0167, 0.0145, 0.0309, 0.0225, 0.0148, 0.0332], device='cuda:0'), in_proj_covar=tensor([0.0182, 0.0157, 0.0148, 0.0177, 0.0199, 0.0191, 0.0155, 0.0194], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-15 22:41:23,685 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.4047, 4.8904, 5.3607, 4.6743, 4.9585, 4.7992, 5.3859, 5.0549], device='cuda:0'), covar=tensor([0.0208, 0.0366, 0.0233, 0.0239, 0.0281, 0.0284, 0.0179, 0.0234], device='cuda:0'), in_proj_covar=tensor([0.0234, 0.0240, 0.0260, 0.0235, 0.0233, 0.0235, 0.0212, 0.0198], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-05-15 22:41:24,964 INFO [finetune.py:992] (0/2) Epoch 3, batch 11000, loss[loss=0.2417, simple_loss=0.3245, pruned_loss=0.07942, over 11102.00 frames. ], tot_loss[loss=0.1812, simple_loss=0.2686, pruned_loss=0.04694, over 2358066.37 frames. ], batch size: 55, lr: 4.85e-03, grad_scale: 8.0 2023-05-15 22:41:36,519 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.75 vs. limit=2.0 2023-05-15 22:42:01,227 INFO [finetune.py:992] (0/2) Epoch 3, batch 11050, loss[loss=0.2774, simple_loss=0.3533, pruned_loss=0.1007, over 11028.00 frames. ], tot_loss[loss=0.1854, simple_loss=0.2723, pruned_loss=0.04924, over 2316598.23 frames. ], batch size: 55, lr: 4.85e-03, grad_scale: 8.0 2023-05-15 22:42:01,337 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=135518.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 22:42:03,775 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=2.00 vs. limit=2.0 2023-05-15 22:42:06,057 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.024e+02 3.260e+02 3.848e+02 4.589e+02 1.215e+03, threshold=7.697e+02, percent-clipped=4.0 2023-05-15 22:42:07,366 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=135527.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 22:42:24,916 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.1051, 5.9056, 5.4385, 5.4136, 5.9710, 5.4014, 5.5803, 5.5435], device='cuda:0'), covar=tensor([0.1507, 0.0907, 0.1059, 0.2053, 0.0878, 0.2167, 0.1735, 0.1016], device='cuda:0'), in_proj_covar=tensor([0.0334, 0.0463, 0.0361, 0.0415, 0.0442, 0.0420, 0.0368, 0.0365], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-15 22:42:35,348 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=135566.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 22:42:36,710 INFO [finetune.py:992] (0/2) Epoch 3, batch 11100, loss[loss=0.2354, simple_loss=0.3249, pruned_loss=0.07294, over 11049.00 frames. ], tot_loss[loss=0.1896, simple_loss=0.2758, pruned_loss=0.05166, over 2274059.69 frames. ], batch size: 55, lr: 4.85e-03, grad_scale: 8.0 2023-05-15 22:43:12,093 INFO [finetune.py:992] (0/2) Epoch 3, batch 11150, loss[loss=0.2007, simple_loss=0.2875, pruned_loss=0.05697, over 12263.00 frames. ], tot_loss[loss=0.1968, simple_loss=0.2815, pruned_loss=0.05606, over 2218948.61 frames. ], batch size: 37, lr: 4.85e-03, grad_scale: 8.0 2023-05-15 22:43:16,873 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.554e+02 3.344e+02 4.365e+02 5.107e+02 8.408e+02, threshold=8.730e+02, percent-clipped=1.0 2023-05-15 22:43:20,657 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=135630.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 22:43:31,493 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.1435, 4.1912, 4.1747, 4.5532, 3.2314, 4.0020, 2.7875, 4.1676], device='cuda:0'), covar=tensor([0.1761, 0.0718, 0.1016, 0.0579, 0.1094, 0.0634, 0.1758, 0.1514], device='cuda:0'), in_proj_covar=tensor([0.0218, 0.0252, 0.0285, 0.0341, 0.0232, 0.0230, 0.0248, 0.0377], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-15 22:43:43,883 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.5552, 3.2812, 3.4846, 3.5808, 3.5141, 3.6112, 3.4693, 2.7718], device='cuda:0'), covar=tensor([0.0081, 0.0085, 0.0120, 0.0065, 0.0062, 0.0096, 0.0076, 0.0531], device='cuda:0'), in_proj_covar=tensor([0.0062, 0.0070, 0.0074, 0.0067, 0.0055, 0.0083, 0.0073, 0.0086], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-15 22:43:47,807 INFO [finetune.py:992] (0/2) Epoch 3, batch 11200, loss[loss=0.2, simple_loss=0.2774, pruned_loss=0.0613, over 12254.00 frames. ], tot_loss[loss=0.2056, simple_loss=0.2888, pruned_loss=0.06117, over 2151898.74 frames. ], batch size: 37, lr: 4.85e-03, grad_scale: 8.0 2023-05-15 22:44:04,021 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=135691.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 22:44:19,853 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.0021, 4.7597, 4.7935, 4.8721, 4.6829, 4.9243, 4.7801, 2.6363], device='cuda:0'), covar=tensor([0.0110, 0.0068, 0.0091, 0.0065, 0.0058, 0.0093, 0.0161, 0.0722], device='cuda:0'), in_proj_covar=tensor([0.0063, 0.0071, 0.0075, 0.0068, 0.0055, 0.0084, 0.0073, 0.0087], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-15 22:44:22,508 INFO [finetune.py:992] (0/2) Epoch 3, batch 11250, loss[loss=0.3308, simple_loss=0.3762, pruned_loss=0.1426, over 7126.00 frames. ], tot_loss[loss=0.2142, simple_loss=0.296, pruned_loss=0.06618, over 2090077.44 frames. ], batch size: 98, lr: 4.85e-03, grad_scale: 16.0 2023-05-15 22:44:22,985 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=4.69 vs. limit=5.0 2023-05-15 22:44:27,925 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.526e+02 3.507e+02 4.335e+02 5.323e+02 9.145e+02, threshold=8.671e+02, percent-clipped=1.0 2023-05-15 22:44:32,014 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.4618, 4.4033, 4.3213, 3.9174, 4.1339, 4.4314, 4.1332, 3.9190], device='cuda:0'), covar=tensor([0.0727, 0.0898, 0.0690, 0.1317, 0.1995, 0.0797, 0.1382, 0.1149], device='cuda:0'), in_proj_covar=tensor([0.0553, 0.0495, 0.0464, 0.0564, 0.0370, 0.0636, 0.0690, 0.0509], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0004, 0.0004, 0.0003], device='cuda:0') 2023-05-15 22:44:53,647 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.7999, 4.1857, 3.8846, 4.6009, 4.0534, 2.6183, 3.9871, 3.0373], device='cuda:0'), covar=tensor([0.0851, 0.0826, 0.1127, 0.0303, 0.1157, 0.1633, 0.1027, 0.2723], device='cuda:0'), in_proj_covar=tensor([0.0284, 0.0354, 0.0334, 0.0247, 0.0344, 0.0253, 0.0320, 0.0343], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-15 22:44:57,266 INFO [finetune.py:992] (0/2) Epoch 3, batch 11300, loss[loss=0.2637, simple_loss=0.3429, pruned_loss=0.09229, over 10190.00 frames. ], tot_loss[loss=0.2225, simple_loss=0.3029, pruned_loss=0.07103, over 2016269.64 frames. ], batch size: 68, lr: 4.85e-03, grad_scale: 16.0 2023-05-15 22:45:04,482 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.3659, 4.6295, 4.2720, 5.0230, 4.6112, 2.9035, 4.4274, 3.1511], device='cuda:0'), covar=tensor([0.0726, 0.0846, 0.1021, 0.0310, 0.0866, 0.1461, 0.0783, 0.2944], device='cuda:0'), in_proj_covar=tensor([0.0285, 0.0355, 0.0335, 0.0248, 0.0345, 0.0253, 0.0321, 0.0344], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-15 22:45:20,540 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 2023-05-15 22:45:33,282 INFO [finetune.py:992] (0/2) Epoch 3, batch 11350, loss[loss=0.2645, simple_loss=0.3499, pruned_loss=0.08956, over 11430.00 frames. ], tot_loss[loss=0.2266, simple_loss=0.3067, pruned_loss=0.07327, over 1993365.55 frames. ], batch size: 48, lr: 4.85e-03, grad_scale: 16.0 2023-05-15 22:45:37,861 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.640e+02 3.541e+02 4.178e+02 4.948e+02 7.953e+02, threshold=8.356e+02, percent-clipped=0.0 2023-05-15 22:45:39,384 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=135827.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 22:45:40,140 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.5956, 4.4713, 4.6021, 4.6156, 4.2829, 4.3232, 4.2509, 4.5442], device='cuda:0'), covar=tensor([0.0720, 0.0536, 0.0697, 0.0551, 0.1661, 0.1262, 0.0495, 0.0822], device='cuda:0'), in_proj_covar=tensor([0.0456, 0.0594, 0.0507, 0.0556, 0.0716, 0.0661, 0.0485, 0.0441], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0005, 0.0005, 0.0003, 0.0003], device='cuda:0') 2023-05-15 22:46:07,770 INFO [finetune.py:992] (0/2) Epoch 3, batch 11400, loss[loss=0.3118, simple_loss=0.3694, pruned_loss=0.1271, over 6940.00 frames. ], tot_loss[loss=0.2333, simple_loss=0.3116, pruned_loss=0.07745, over 1918166.40 frames. ], batch size: 98, lr: 4.85e-03, grad_scale: 16.0 2023-05-15 22:46:13,231 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=135875.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 22:46:42,736 INFO [finetune.py:992] (0/2) Epoch 3, batch 11450, loss[loss=0.3177, simple_loss=0.365, pruned_loss=0.1352, over 7055.00 frames. ], tot_loss[loss=0.2391, simple_loss=0.3159, pruned_loss=0.08109, over 1875266.34 frames. ], batch size: 104, lr: 4.85e-03, grad_scale: 16.0 2023-05-15 22:46:47,389 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.608e+02 3.568e+02 4.271e+02 4.917e+02 1.252e+03, threshold=8.543e+02, percent-clipped=1.0 2023-05-15 22:47:12,833 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.4662, 4.4078, 4.3179, 3.9994, 4.1177, 4.4314, 4.1526, 4.0387], device='cuda:0'), covar=tensor([0.0677, 0.0875, 0.0628, 0.1206, 0.1573, 0.0763, 0.1272, 0.0992], device='cuda:0'), in_proj_covar=tensor([0.0539, 0.0484, 0.0455, 0.0553, 0.0360, 0.0619, 0.0671, 0.0498], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-05-15 22:47:16,841 INFO [finetune.py:992] (0/2) Epoch 3, batch 11500, loss[loss=0.2488, simple_loss=0.3224, pruned_loss=0.08764, over 6965.00 frames. ], tot_loss[loss=0.2437, simple_loss=0.3191, pruned_loss=0.08416, over 1829780.02 frames. ], batch size: 99, lr: 4.85e-03, grad_scale: 16.0 2023-05-15 22:47:23,064 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.8214, 3.7150, 3.8085, 3.5731, 3.6864, 3.5807, 3.7432, 3.4347], device='cuda:0'), covar=tensor([0.0353, 0.0294, 0.0349, 0.0233, 0.0299, 0.0285, 0.0355, 0.1232], device='cuda:0'), in_proj_covar=tensor([0.0217, 0.0221, 0.0240, 0.0218, 0.0216, 0.0217, 0.0197, 0.0181], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-15 22:47:26,918 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=135982.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 22:47:29,589 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=135986.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 22:47:35,173 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=135994.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 22:47:39,545 INFO [checkpoint.py:75] (0/2) Saving checkpoint to pruned_transducer_stateless7/exp_giga_finetune/checkpoint-36000.pt 2023-05-15 22:47:55,319 INFO [finetune.py:992] (0/2) Epoch 3, batch 11550, loss[loss=0.2936, simple_loss=0.3528, pruned_loss=0.1172, over 7070.00 frames. ], tot_loss[loss=0.2475, simple_loss=0.3215, pruned_loss=0.08677, over 1819258.48 frames. ], batch size: 99, lr: 4.85e-03, grad_scale: 16.0 2023-05-15 22:48:00,055 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.813e+02 3.551e+02 4.092e+02 4.768e+02 1.093e+03, threshold=8.184e+02, percent-clipped=1.0 2023-05-15 22:48:13,040 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=136043.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 22:48:21,183 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=136055.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 22:48:30,377 INFO [finetune.py:992] (0/2) Epoch 3, batch 11600, loss[loss=0.2062, simple_loss=0.2908, pruned_loss=0.06076, over 11093.00 frames. ], tot_loss[loss=0.2488, simple_loss=0.3223, pruned_loss=0.08761, over 1793043.37 frames. ], batch size: 55, lr: 4.85e-03, grad_scale: 16.0 2023-05-15 22:49:05,616 INFO [finetune.py:992] (0/2) Epoch 3, batch 11650, loss[loss=0.2393, simple_loss=0.3265, pruned_loss=0.07602, over 10520.00 frames. ], tot_loss[loss=0.2495, simple_loss=0.3222, pruned_loss=0.08837, over 1777449.23 frames. ], batch size: 68, lr: 4.85e-03, grad_scale: 8.0 2023-05-15 22:49:11,210 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.566e+02 3.696e+02 4.157e+02 5.000e+02 3.157e+03, threshold=8.314e+02, percent-clipped=5.0 2023-05-15 22:49:22,748 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=136142.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 22:49:29,631 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.94 vs. limit=2.0 2023-05-15 22:49:34,063 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.6915, 3.5708, 3.6433, 3.6949, 3.4226, 3.7385, 3.6617, 3.8080], device='cuda:0'), covar=tensor([0.0266, 0.0156, 0.0172, 0.0257, 0.0580, 0.0274, 0.0222, 0.0225], device='cuda:0'), in_proj_covar=tensor([0.0160, 0.0158, 0.0158, 0.0202, 0.0203, 0.0175, 0.0145, 0.0191], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:0') 2023-05-15 22:49:40,489 INFO [finetune.py:992] (0/2) Epoch 3, batch 11700, loss[loss=0.3058, simple_loss=0.3514, pruned_loss=0.1301, over 6603.00 frames. ], tot_loss[loss=0.25, simple_loss=0.322, pruned_loss=0.08898, over 1759999.49 frames. ], batch size: 105, lr: 4.85e-03, grad_scale: 8.0 2023-05-15 22:50:02,176 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([1.9965, 2.1762, 2.1695, 2.1381, 1.9295, 1.9368, 2.0701, 1.5610], device='cuda:0'), covar=tensor([0.0288, 0.0156, 0.0170, 0.0187, 0.0275, 0.0201, 0.0141, 0.0357], device='cuda:0'), in_proj_covar=tensor([0.0177, 0.0151, 0.0143, 0.0173, 0.0193, 0.0186, 0.0150, 0.0192], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], device='cuda:0') 2023-05-15 22:50:05,175 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=136203.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 22:50:15,108 INFO [finetune.py:992] (0/2) Epoch 3, batch 11750, loss[loss=0.2421, simple_loss=0.3252, pruned_loss=0.07954, over 11029.00 frames. ], tot_loss[loss=0.2498, simple_loss=0.3217, pruned_loss=0.0889, over 1749996.10 frames. ], batch size: 55, lr: 4.85e-03, grad_scale: 8.0 2023-05-15 22:50:20,310 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.982e+02 3.625e+02 4.141e+02 4.770e+02 9.665e+02, threshold=8.282e+02, percent-clipped=2.0 2023-05-15 22:50:21,509 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=4.79 vs. limit=5.0 2023-05-15 22:50:40,985 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.6788, 3.9291, 3.5744, 4.2957, 3.8682, 2.5978, 3.7319, 2.8153], device='cuda:0'), covar=tensor([0.0916, 0.0940, 0.1349, 0.0396, 0.1131, 0.1750, 0.1117, 0.3178], device='cuda:0'), in_proj_covar=tensor([0.0277, 0.0343, 0.0323, 0.0236, 0.0333, 0.0247, 0.0311, 0.0335], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-15 22:50:50,290 INFO [finetune.py:992] (0/2) Epoch 3, batch 11800, loss[loss=0.231, simple_loss=0.3152, pruned_loss=0.07342, over 12116.00 frames. ], tot_loss[loss=0.2516, simple_loss=0.3232, pruned_loss=0.09001, over 1722422.62 frames. ], batch size: 39, lr: 4.85e-03, grad_scale: 8.0 2023-05-15 22:51:02,689 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=136286.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 22:51:09,418 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.2674, 4.1023, 4.1405, 4.2040, 4.1740, 4.3182, 4.1541, 2.2251], device='cuda:0'), covar=tensor([0.0117, 0.0083, 0.0130, 0.0090, 0.0059, 0.0115, 0.0102, 0.1011], device='cuda:0'), in_proj_covar=tensor([0.0059, 0.0066, 0.0070, 0.0063, 0.0052, 0.0079, 0.0068, 0.0084], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-15 22:51:24,571 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.99 vs. limit=2.0 2023-05-15 22:51:25,552 INFO [finetune.py:992] (0/2) Epoch 3, batch 11850, loss[loss=0.2695, simple_loss=0.3368, pruned_loss=0.1011, over 6466.00 frames. ], tot_loss[loss=0.2521, simple_loss=0.3244, pruned_loss=0.0899, over 1705334.38 frames. ], batch size: 98, lr: 4.85e-03, grad_scale: 8.0 2023-05-15 22:51:30,896 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.669e+02 3.750e+02 4.377e+02 5.279e+02 9.259e+02, threshold=8.753e+02, percent-clipped=2.0 2023-05-15 22:51:36,270 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=136334.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 22:51:37,033 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=136335.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 22:51:38,914 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=136338.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 22:51:47,158 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=136350.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 22:51:52,227 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=2.01 vs. limit=2.0 2023-05-15 22:52:00,483 INFO [finetune.py:992] (0/2) Epoch 3, batch 11900, loss[loss=0.2465, simple_loss=0.3207, pruned_loss=0.08611, over 7088.00 frames. ], tot_loss[loss=0.2528, simple_loss=0.3252, pruned_loss=0.09025, over 1671388.43 frames. ], batch size: 98, lr: 4.85e-03, grad_scale: 8.0 2023-05-15 22:52:19,635 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=136396.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 22:52:34,986 INFO [finetune.py:992] (0/2) Epoch 3, batch 11950, loss[loss=0.186, simple_loss=0.2759, pruned_loss=0.04808, over 10925.00 frames. ], tot_loss[loss=0.2474, simple_loss=0.3213, pruned_loss=0.08682, over 1678642.76 frames. ], batch size: 55, lr: 4.85e-03, grad_scale: 8.0 2023-05-15 22:52:41,191 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.345e+02 3.513e+02 4.159e+02 5.094e+02 8.737e+02, threshold=8.317e+02, percent-clipped=0.0 2023-05-15 22:53:09,676 INFO [finetune.py:992] (0/2) Epoch 3, batch 12000, loss[loss=0.2193, simple_loss=0.2926, pruned_loss=0.07298, over 6945.00 frames. ], tot_loss[loss=0.2409, simple_loss=0.3162, pruned_loss=0.08283, over 1671653.17 frames. ], batch size: 98, lr: 4.85e-03, grad_scale: 8.0 2023-05-15 22:53:09,676 INFO [finetune.py:1017] (0/2) Computing validation loss 2023-05-15 22:53:28,729 INFO [finetune.py:1026] (0/2) Epoch 3, validation: loss=0.2937, simple_loss=0.369, pruned_loss=0.1092, over 1020973.00 frames. 2023-05-15 22:53:28,730 INFO [finetune.py:1027] (0/2) Maximum memory allocated so far is 12745MB 2023-05-15 22:53:48,926 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=136498.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 22:54:02,717 INFO [finetune.py:992] (0/2) Epoch 3, batch 12050, loss[loss=0.2286, simple_loss=0.3085, pruned_loss=0.07433, over 7286.00 frames. ], tot_loss[loss=0.235, simple_loss=0.3114, pruned_loss=0.07934, over 1653159.63 frames. ], batch size: 103, lr: 4.85e-03, grad_scale: 8.0 2023-05-15 22:54:08,051 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.114e+02 2.988e+02 3.634e+02 4.408e+02 6.442e+02, threshold=7.268e+02, percent-clipped=0.0 2023-05-15 22:54:31,244 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.6026, 5.5173, 5.3612, 4.9495, 4.8312, 5.5245, 5.1738, 5.0463], device='cuda:0'), covar=tensor([0.0581, 0.0836, 0.0648, 0.1328, 0.0960, 0.0650, 0.1351, 0.0798], device='cuda:0'), in_proj_covar=tensor([0.0529, 0.0476, 0.0445, 0.0538, 0.0356, 0.0608, 0.0653, 0.0490], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-15 22:54:35,214 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.8942, 3.7681, 3.8782, 3.6248, 3.7372, 3.6346, 3.8509, 3.5716], device='cuda:0'), covar=tensor([0.0372, 0.0359, 0.0374, 0.0253, 0.0328, 0.0286, 0.0304, 0.0979], device='cuda:0'), in_proj_covar=tensor([0.0209, 0.0211, 0.0229, 0.0211, 0.0206, 0.0208, 0.0189, 0.0173], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-15 22:54:35,719 INFO [finetune.py:992] (0/2) Epoch 3, batch 12100, loss[loss=0.2238, simple_loss=0.2938, pruned_loss=0.07691, over 6978.00 frames. ], tot_loss[loss=0.2339, simple_loss=0.3107, pruned_loss=0.07855, over 1645997.50 frames. ], batch size: 98, lr: 4.85e-03, grad_scale: 8.0 2023-05-15 22:54:41,759 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.9920, 2.1547, 2.6210, 2.9943, 2.2155, 3.1034, 3.0476, 3.1508], device='cuda:0'), covar=tensor([0.0133, 0.1061, 0.0435, 0.0174, 0.1011, 0.0220, 0.0250, 0.0123], device='cuda:0'), in_proj_covar=tensor([0.0109, 0.0191, 0.0172, 0.0107, 0.0173, 0.0158, 0.0153, 0.0111], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-15 22:55:07,960 INFO [finetune.py:992] (0/2) Epoch 3, batch 12150, loss[loss=0.3126, simple_loss=0.3619, pruned_loss=0.1317, over 6622.00 frames. ], tot_loss[loss=0.2345, simple_loss=0.3117, pruned_loss=0.07869, over 1663593.47 frames. ], batch size: 98, lr: 4.84e-03, grad_scale: 8.0 2023-05-15 22:55:12,829 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.198e+02 3.271e+02 3.722e+02 4.794e+02 9.063e+02, threshold=7.443e+02, percent-clipped=4.0 2023-05-15 22:55:20,436 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=136638.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 22:55:24,297 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=136643.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 22:55:28,774 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=136650.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 22:55:39,638 INFO [finetune.py:992] (0/2) Epoch 3, batch 12200, loss[loss=0.2801, simple_loss=0.3373, pruned_loss=0.1115, over 6845.00 frames. ], tot_loss[loss=0.2362, simple_loss=0.3127, pruned_loss=0.07982, over 1652739.15 frames. ], batch size: 98, lr: 4.84e-03, grad_scale: 8.0 2023-05-15 22:55:50,575 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=136686.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 22:55:53,535 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=136691.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 22:55:57,900 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=136698.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 22:56:01,177 INFO [checkpoint.py:75] (0/2) Saving checkpoint to pruned_transducer_stateless7/exp_giga_finetune/epoch-3.pt 2023-05-15 22:56:26,417 INFO [finetune.py:992] (0/2) Epoch 4, batch 0, loss[loss=0.2253, simple_loss=0.3085, pruned_loss=0.07102, over 12113.00 frames. ], tot_loss[loss=0.2253, simple_loss=0.3085, pruned_loss=0.07102, over 12113.00 frames. ], batch size: 39, lr: 4.84e-03, grad_scale: 8.0 2023-05-15 22:56:26,418 INFO [finetune.py:1017] (0/2) Computing validation loss 2023-05-15 22:56:39,686 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.6624, 4.5423, 4.5329, 4.4682, 4.1538, 4.5813, 4.5561, 4.7022], device='cuda:0'), covar=tensor([0.0213, 0.0139, 0.0166, 0.0254, 0.0670, 0.0278, 0.0153, 0.0176], device='cuda:0'), in_proj_covar=tensor([0.0150, 0.0148, 0.0149, 0.0190, 0.0189, 0.0166, 0.0137, 0.0182], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:0') 2023-05-15 22:56:44,024 INFO [finetune.py:1026] (0/2) Epoch 4, validation: loss=0.2991, simple_loss=0.3703, pruned_loss=0.1139, over 1020973.00 frames. 2023-05-15 22:56:44,025 INFO [finetune.py:1027] (0/2) Maximum memory allocated so far is 12745MB 2023-05-15 22:56:45,695 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=136704.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 22:57:00,962 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.149e+02 3.364e+02 3.979e+02 4.865e+02 1.021e+03, threshold=7.957e+02, percent-clipped=3.0 2023-05-15 22:57:04,344 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=4.55 vs. limit=5.0 2023-05-15 22:57:20,547 INFO [finetune.py:992] (0/2) Epoch 4, batch 50, loss[loss=0.1602, simple_loss=0.246, pruned_loss=0.03721, over 12021.00 frames. ], tot_loss[loss=0.1858, simple_loss=0.2728, pruned_loss=0.04936, over 534535.33 frames. ], batch size: 28, lr: 4.84e-03, grad_scale: 8.0 2023-05-15 22:57:26,947 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.8864, 5.8726, 5.5984, 5.1939, 5.0027, 5.8040, 5.4231, 5.2749], device='cuda:0'), covar=tensor([0.0844, 0.0935, 0.0751, 0.1582, 0.0719, 0.0725, 0.1579, 0.0901], device='cuda:0'), in_proj_covar=tensor([0.0533, 0.0476, 0.0448, 0.0541, 0.0357, 0.0609, 0.0656, 0.0492], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-15 22:57:53,328 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=136798.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 22:57:56,263 INFO [finetune.py:992] (0/2) Epoch 4, batch 100, loss[loss=0.189, simple_loss=0.2868, pruned_loss=0.04566, over 12098.00 frames. ], tot_loss[loss=0.1844, simple_loss=0.2724, pruned_loss=0.04818, over 947726.35 frames. ], batch size: 39, lr: 4.84e-03, grad_scale: 8.0 2023-05-15 22:57:59,315 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=136806.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 22:58:03,489 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.0770, 5.0636, 4.8756, 4.9793, 4.5847, 5.0944, 4.9219, 5.3321], device='cuda:0'), covar=tensor([0.0198, 0.0120, 0.0197, 0.0286, 0.0680, 0.0294, 0.0170, 0.0147], device='cuda:0'), in_proj_covar=tensor([0.0157, 0.0155, 0.0156, 0.0199, 0.0197, 0.0173, 0.0143, 0.0189], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:0') 2023-05-15 22:58:13,296 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.855e+02 2.875e+02 3.507e+02 4.053e+02 1.106e+03, threshold=7.015e+02, percent-clipped=2.0 2023-05-15 22:58:27,405 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=136846.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 22:58:31,619 INFO [finetune.py:992] (0/2) Epoch 4, batch 150, loss[loss=0.2038, simple_loss=0.2868, pruned_loss=0.06041, over 12143.00 frames. ], tot_loss[loss=0.1858, simple_loss=0.273, pruned_loss=0.04931, over 1260198.66 frames. ], batch size: 38, lr: 4.84e-03, grad_scale: 8.0 2023-05-15 22:58:42,581 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.9878, 5.9648, 5.6910, 5.1860, 5.0113, 5.9113, 5.5168, 5.2965], device='cuda:0'), covar=tensor([0.0815, 0.0909, 0.0732, 0.1726, 0.0706, 0.0675, 0.1494, 0.1021], device='cuda:0'), in_proj_covar=tensor([0.0546, 0.0488, 0.0458, 0.0557, 0.0366, 0.0624, 0.0675, 0.0505], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-05-15 22:58:42,662 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=136867.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 22:59:07,974 INFO [finetune.py:992] (0/2) Epoch 4, batch 200, loss[loss=0.1644, simple_loss=0.2598, pruned_loss=0.03451, over 12300.00 frames. ], tot_loss[loss=0.1826, simple_loss=0.2693, pruned_loss=0.04792, over 1508903.67 frames. ], batch size: 34, lr: 4.84e-03, grad_scale: 8.0 2023-05-15 22:59:13,830 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.4371, 2.5982, 3.2876, 4.2140, 2.2092, 4.3184, 4.3534, 4.4970], device='cuda:0'), covar=tensor([0.0092, 0.1127, 0.0435, 0.0123, 0.1250, 0.0177, 0.0119, 0.0070], device='cuda:0'), in_proj_covar=tensor([0.0111, 0.0196, 0.0175, 0.0108, 0.0178, 0.0162, 0.0157, 0.0113], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-15 22:59:25,718 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.852e+02 3.078e+02 3.576e+02 4.250e+02 9.794e+02, threshold=7.152e+02, percent-clipped=4.0 2023-05-15 22:59:33,032 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.9783, 4.6234, 4.8461, 4.8778, 4.5516, 4.9438, 4.7159, 2.3631], device='cuda:0'), covar=tensor([0.0092, 0.0068, 0.0081, 0.0062, 0.0058, 0.0089, 0.0073, 0.0836], device='cuda:0'), in_proj_covar=tensor([0.0058, 0.0066, 0.0070, 0.0063, 0.0052, 0.0079, 0.0068, 0.0084], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-15 22:59:38,682 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.1041, 4.3136, 3.8691, 4.6906, 4.2775, 2.7741, 4.0473, 2.9285], device='cuda:0'), covar=tensor([0.0768, 0.0855, 0.1377, 0.0393, 0.1055, 0.1669, 0.0983, 0.3205], device='cuda:0'), in_proj_covar=tensor([0.0283, 0.0353, 0.0334, 0.0240, 0.0344, 0.0254, 0.0319, 0.0346], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-15 22:59:44,102 INFO [finetune.py:992] (0/2) Epoch 4, batch 250, loss[loss=0.1549, simple_loss=0.2368, pruned_loss=0.03656, over 12265.00 frames. ], tot_loss[loss=0.1828, simple_loss=0.2697, pruned_loss=0.04793, over 1698300.45 frames. ], batch size: 32, lr: 4.84e-03, grad_scale: 8.0 2023-05-15 23:00:12,022 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=136991.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 23:00:17,635 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=136999.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 23:00:20,040 INFO [finetune.py:992] (0/2) Epoch 4, batch 300, loss[loss=0.2017, simple_loss=0.2864, pruned_loss=0.05852, over 10543.00 frames. ], tot_loss[loss=0.1812, simple_loss=0.2684, pruned_loss=0.04705, over 1848269.53 frames. ], batch size: 68, lr: 4.84e-03, grad_scale: 8.0 2023-05-15 23:00:36,828 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.656e+02 2.970e+02 3.404e+02 3.901e+02 9.506e+02, threshold=6.809e+02, percent-clipped=2.0 2023-05-15 23:00:46,929 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=137039.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 23:00:56,328 INFO [finetune.py:992] (0/2) Epoch 4, batch 350, loss[loss=0.167, simple_loss=0.2636, pruned_loss=0.03524, over 12308.00 frames. ], tot_loss[loss=0.1803, simple_loss=0.2675, pruned_loss=0.04661, over 1971240.00 frames. ], batch size: 34, lr: 4.84e-03, grad_scale: 8.0 2023-05-15 23:01:32,267 INFO [finetune.py:992] (0/2) Epoch 4, batch 400, loss[loss=0.1973, simple_loss=0.2871, pruned_loss=0.0537, over 11376.00 frames. ], tot_loss[loss=0.1802, simple_loss=0.2678, pruned_loss=0.04625, over 2060494.26 frames. ], batch size: 55, lr: 4.84e-03, grad_scale: 8.0 2023-05-15 23:01:49,110 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.920e+02 2.863e+02 3.332e+02 3.985e+02 7.745e+02, threshold=6.664e+02, percent-clipped=2.0 2023-05-15 23:02:06,216 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.2709, 4.1043, 4.0024, 4.3777, 2.7638, 3.9189, 2.5624, 4.1102], device='cuda:0'), covar=tensor([0.1647, 0.0671, 0.0978, 0.0651, 0.1192, 0.0592, 0.1847, 0.1200], device='cuda:0'), in_proj_covar=tensor([0.0223, 0.0255, 0.0284, 0.0339, 0.0233, 0.0232, 0.0252, 0.0375], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-15 23:02:07,354 INFO [finetune.py:992] (0/2) Epoch 4, batch 450, loss[loss=0.2, simple_loss=0.2735, pruned_loss=0.06326, over 12346.00 frames. ], tot_loss[loss=0.1789, simple_loss=0.2669, pruned_loss=0.04542, over 2139588.88 frames. ], batch size: 31, lr: 4.84e-03, grad_scale: 8.0 2023-05-15 23:02:14,499 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=137162.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 23:02:20,896 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.7942, 2.8222, 3.8599, 4.8013, 4.2468, 4.7652, 4.1917, 3.4374], device='cuda:0'), covar=tensor([0.0023, 0.0339, 0.0104, 0.0025, 0.0087, 0.0052, 0.0071, 0.0309], device='cuda:0'), in_proj_covar=tensor([0.0077, 0.0117, 0.0096, 0.0071, 0.0094, 0.0107, 0.0083, 0.0130], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-15 23:02:43,489 INFO [finetune.py:992] (0/2) Epoch 4, batch 500, loss[loss=0.1917, simple_loss=0.2846, pruned_loss=0.04945, over 11694.00 frames. ], tot_loss[loss=0.1783, simple_loss=0.266, pruned_loss=0.04532, over 2197599.08 frames. ], batch size: 48, lr: 4.84e-03, grad_scale: 8.0 2023-05-15 23:03:00,587 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.913e+02 2.872e+02 3.437e+02 3.927e+02 1.123e+03, threshold=6.874e+02, percent-clipped=1.0 2023-05-15 23:03:19,546 INFO [finetune.py:992] (0/2) Epoch 4, batch 550, loss[loss=0.1713, simple_loss=0.2661, pruned_loss=0.03821, over 12106.00 frames. ], tot_loss[loss=0.1783, simple_loss=0.2664, pruned_loss=0.04516, over 2237617.25 frames. ], batch size: 33, lr: 4.84e-03, grad_scale: 8.0 2023-05-15 23:03:37,966 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.3251, 4.8541, 2.9689, 2.5926, 4.1345, 2.6743, 4.1351, 3.3760], device='cuda:0'), covar=tensor([0.0661, 0.0351, 0.1092, 0.1531, 0.0214, 0.1260, 0.0421, 0.0788], device='cuda:0'), in_proj_covar=tensor([0.0187, 0.0239, 0.0173, 0.0198, 0.0133, 0.0180, 0.0187, 0.0172], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-15 23:03:52,654 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=137299.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 23:03:54,749 INFO [finetune.py:992] (0/2) Epoch 4, batch 600, loss[loss=0.1853, simple_loss=0.2766, pruned_loss=0.04699, over 12331.00 frames. ], tot_loss[loss=0.179, simple_loss=0.2671, pruned_loss=0.04547, over 2258976.07 frames. ], batch size: 36, lr: 4.84e-03, grad_scale: 8.0 2023-05-15 23:04:08,275 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.6741, 2.3803, 3.7715, 4.7329, 4.0858, 4.5666, 4.0749, 3.0798], device='cuda:0'), covar=tensor([0.0022, 0.0380, 0.0113, 0.0024, 0.0099, 0.0053, 0.0065, 0.0313], device='cuda:0'), in_proj_covar=tensor([0.0079, 0.0120, 0.0099, 0.0073, 0.0096, 0.0109, 0.0086, 0.0133], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-15 23:04:11,613 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.075e+02 2.902e+02 3.445e+02 4.026e+02 7.738e+02, threshold=6.890e+02, percent-clipped=2.0 2023-05-15 23:04:22,625 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.96 vs. limit=2.0 2023-05-15 23:04:27,065 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=137347.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 23:04:30,492 INFO [finetune.py:992] (0/2) Epoch 4, batch 650, loss[loss=0.1366, simple_loss=0.2225, pruned_loss=0.02529, over 12133.00 frames. ], tot_loss[loss=0.1782, simple_loss=0.2661, pruned_loss=0.04517, over 2286335.67 frames. ], batch size: 30, lr: 4.84e-03, grad_scale: 8.0 2023-05-15 23:04:38,404 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.7092, 2.9988, 3.7436, 4.8346, 4.1781, 4.8004, 4.1017, 3.3852], device='cuda:0'), covar=tensor([0.0028, 0.0294, 0.0133, 0.0021, 0.0096, 0.0056, 0.0076, 0.0296], device='cuda:0'), in_proj_covar=tensor([0.0080, 0.0121, 0.0100, 0.0074, 0.0097, 0.0110, 0.0087, 0.0134], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-15 23:04:47,442 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.0441, 2.4426, 3.5840, 3.0820, 3.4717, 3.1820, 2.3926, 3.5043], device='cuda:0'), covar=tensor([0.0132, 0.0325, 0.0115, 0.0221, 0.0157, 0.0153, 0.0353, 0.0123], device='cuda:0'), in_proj_covar=tensor([0.0154, 0.0188, 0.0162, 0.0168, 0.0188, 0.0144, 0.0180, 0.0157], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-15 23:05:06,776 INFO [finetune.py:992] (0/2) Epoch 4, batch 700, loss[loss=0.1855, simple_loss=0.2795, pruned_loss=0.04581, over 11345.00 frames. ], tot_loss[loss=0.1776, simple_loss=0.2655, pruned_loss=0.04483, over 2310458.40 frames. ], batch size: 55, lr: 4.84e-03, grad_scale: 8.0 2023-05-15 23:05:12,260 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.85 vs. limit=2.0 2023-05-15 23:05:23,829 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.847e+02 2.978e+02 3.456e+02 4.081e+02 5.922e+02, threshold=6.912e+02, percent-clipped=0.0 2023-05-15 23:05:25,642 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.49 vs. limit=2.0 2023-05-15 23:05:34,621 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.1419, 5.9720, 5.4660, 5.5798, 6.0388, 5.2897, 5.7185, 5.4904], device='cuda:0'), covar=tensor([0.1506, 0.0898, 0.0928, 0.1609, 0.0893, 0.2265, 0.1323, 0.1223], device='cuda:0'), in_proj_covar=tensor([0.0324, 0.0446, 0.0348, 0.0396, 0.0421, 0.0404, 0.0353, 0.0347], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-15 23:05:42,331 INFO [finetune.py:992] (0/2) Epoch 4, batch 750, loss[loss=0.1892, simple_loss=0.2789, pruned_loss=0.04975, over 11808.00 frames. ], tot_loss[loss=0.1782, simple_loss=0.266, pruned_loss=0.04521, over 2321582.25 frames. ], batch size: 44, lr: 4.84e-03, grad_scale: 8.0 2023-05-15 23:05:49,374 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=137462.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 23:06:18,128 INFO [finetune.py:992] (0/2) Epoch 4, batch 800, loss[loss=0.1867, simple_loss=0.2826, pruned_loss=0.04544, over 12350.00 frames. ], tot_loss[loss=0.1782, simple_loss=0.266, pruned_loss=0.04522, over 2331949.85 frames. ], batch size: 35, lr: 4.84e-03, grad_scale: 8.0 2023-05-15 23:06:24,144 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=137510.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 23:06:36,142 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.126e+02 2.963e+02 3.442e+02 4.311e+02 7.987e+02, threshold=6.885e+02, percent-clipped=2.0 2023-05-15 23:06:54,376 INFO [finetune.py:992] (0/2) Epoch 4, batch 850, loss[loss=0.1703, simple_loss=0.2667, pruned_loss=0.03689, over 10452.00 frames. ], tot_loss[loss=0.1778, simple_loss=0.2657, pruned_loss=0.04499, over 2341590.87 frames. ], batch size: 69, lr: 4.84e-03, grad_scale: 8.0 2023-05-15 23:07:29,857 INFO [finetune.py:992] (0/2) Epoch 4, batch 900, loss[loss=0.1582, simple_loss=0.2451, pruned_loss=0.03564, over 12276.00 frames. ], tot_loss[loss=0.1793, simple_loss=0.2668, pruned_loss=0.04589, over 2347908.80 frames. ], batch size: 28, lr: 4.84e-03, grad_scale: 8.0 2023-05-15 23:07:46,882 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.964e+02 2.861e+02 3.473e+02 4.370e+02 1.849e+03, threshold=6.947e+02, percent-clipped=5.0 2023-05-15 23:08:06,281 INFO [finetune.py:992] (0/2) Epoch 4, batch 950, loss[loss=0.1808, simple_loss=0.2683, pruned_loss=0.04667, over 12022.00 frames. ], tot_loss[loss=0.1777, simple_loss=0.2654, pruned_loss=0.04501, over 2360758.27 frames. ], batch size: 40, lr: 4.84e-03, grad_scale: 8.0 2023-05-15 23:08:19,731 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.8869, 4.9329, 4.6609, 4.7510, 4.3877, 4.9392, 4.8639, 5.1411], device='cuda:0'), covar=tensor([0.0194, 0.0121, 0.0187, 0.0331, 0.0766, 0.0251, 0.0140, 0.0170], device='cuda:0'), in_proj_covar=tensor([0.0172, 0.0171, 0.0172, 0.0222, 0.0219, 0.0191, 0.0157, 0.0207], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0005, 0.0004, 0.0004, 0.0003, 0.0005], device='cuda:0') 2023-05-15 23:08:42,204 INFO [finetune.py:992] (0/2) Epoch 4, batch 1000, loss[loss=0.2177, simple_loss=0.3015, pruned_loss=0.06693, over 11299.00 frames. ], tot_loss[loss=0.1783, simple_loss=0.266, pruned_loss=0.0453, over 2360934.37 frames. ], batch size: 55, lr: 4.84e-03, grad_scale: 8.0 2023-05-15 23:08:53,048 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.5157, 2.6392, 3.2842, 4.3544, 2.5244, 4.5226, 4.5806, 4.6802], device='cuda:0'), covar=tensor([0.0152, 0.1182, 0.0452, 0.0132, 0.1136, 0.0151, 0.0127, 0.0075], device='cuda:0'), in_proj_covar=tensor([0.0115, 0.0199, 0.0181, 0.0111, 0.0182, 0.0168, 0.0163, 0.0116], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-15 23:08:59,199 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.962e+02 2.918e+02 3.501e+02 4.066e+02 1.012e+03, threshold=7.003e+02, percent-clipped=3.0 2023-05-15 23:09:00,112 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.8319, 4.9252, 4.6292, 4.7530, 4.3118, 4.8248, 4.7854, 5.1119], device='cuda:0'), covar=tensor([0.0273, 0.0136, 0.0206, 0.0326, 0.0888, 0.0415, 0.0178, 0.0192], device='cuda:0'), in_proj_covar=tensor([0.0173, 0.0173, 0.0173, 0.0224, 0.0221, 0.0192, 0.0158, 0.0208], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0005, 0.0004, 0.0004, 0.0003, 0.0005], device='cuda:0') 2023-05-15 23:09:17,644 INFO [finetune.py:992] (0/2) Epoch 4, batch 1050, loss[loss=0.1694, simple_loss=0.2554, pruned_loss=0.04169, over 12131.00 frames. ], tot_loss[loss=0.178, simple_loss=0.2658, pruned_loss=0.04514, over 2366005.26 frames. ], batch size: 30, lr: 4.84e-03, grad_scale: 8.0 2023-05-15 23:09:40,186 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.6923, 3.1334, 5.0975, 2.5177, 2.8241, 3.7471, 3.1483, 3.8929], device='cuda:0'), covar=tensor([0.0391, 0.1239, 0.0230, 0.1240, 0.1951, 0.1359, 0.1443, 0.0989], device='cuda:0'), in_proj_covar=tensor([0.0230, 0.0226, 0.0224, 0.0177, 0.0232, 0.0274, 0.0221, 0.0246], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:0') 2023-05-15 23:09:53,712 INFO [finetune.py:992] (0/2) Epoch 4, batch 1100, loss[loss=0.1972, simple_loss=0.279, pruned_loss=0.0577, over 11291.00 frames. ], tot_loss[loss=0.1778, simple_loss=0.2654, pruned_loss=0.04506, over 2374913.88 frames. ], batch size: 55, lr: 4.83e-03, grad_scale: 8.0 2023-05-15 23:10:11,420 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.106e+02 2.964e+02 3.480e+02 4.053e+02 1.106e+03, threshold=6.959e+02, percent-clipped=4.0 2023-05-15 23:10:29,690 INFO [finetune.py:992] (0/2) Epoch 4, batch 1150, loss[loss=0.171, simple_loss=0.2658, pruned_loss=0.03807, over 12350.00 frames. ], tot_loss[loss=0.1771, simple_loss=0.2649, pruned_loss=0.04462, over 2385031.04 frames. ], batch size: 35, lr: 4.83e-03, grad_scale: 8.0 2023-05-15 23:10:32,669 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.4156, 5.2617, 5.3277, 5.3879, 5.0124, 4.9534, 4.8758, 5.3048], device='cuda:0'), covar=tensor([0.0620, 0.0488, 0.0686, 0.0575, 0.1716, 0.1507, 0.0494, 0.1000], device='cuda:0'), in_proj_covar=tensor([0.0483, 0.0620, 0.0536, 0.0579, 0.0760, 0.0694, 0.0504, 0.0463], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0005, 0.0004, 0.0004, 0.0005, 0.0005, 0.0004, 0.0004], device='cuda:0') 2023-05-15 23:10:58,419 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=137892.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 23:11:00,312 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=137895.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 23:11:04,983 INFO [finetune.py:992] (0/2) Epoch 4, batch 1200, loss[loss=0.2044, simple_loss=0.2922, pruned_loss=0.05831, over 12092.00 frames. ], tot_loss[loss=0.1772, simple_loss=0.2649, pruned_loss=0.04473, over 2380389.69 frames. ], batch size: 42, lr: 4.83e-03, grad_scale: 8.0 2023-05-15 23:11:20,554 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.46 vs. limit=2.0 2023-05-15 23:11:22,110 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.760e+02 2.876e+02 3.340e+02 4.046e+02 2.321e+03, threshold=6.680e+02, percent-clipped=1.0 2023-05-15 23:11:22,313 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.9834, 4.9920, 4.7018, 4.7925, 4.4558, 4.9370, 4.9444, 5.1646], device='cuda:0'), covar=tensor([0.0220, 0.0139, 0.0187, 0.0365, 0.0740, 0.0333, 0.0151, 0.0174], device='cuda:0'), in_proj_covar=tensor([0.0172, 0.0173, 0.0173, 0.0223, 0.0220, 0.0193, 0.0158, 0.0207], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0005, 0.0004, 0.0004, 0.0003, 0.0005], device='cuda:0') 2023-05-15 23:11:24,373 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.0253, 3.8752, 4.0308, 3.6819, 3.8592, 3.6911, 4.0282, 3.6161], device='cuda:0'), covar=tensor([0.0309, 0.0291, 0.0294, 0.0233, 0.0276, 0.0283, 0.0247, 0.1184], device='cuda:0'), in_proj_covar=tensor([0.0232, 0.0235, 0.0254, 0.0231, 0.0228, 0.0233, 0.0210, 0.0193], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-15 23:11:41,361 INFO [finetune.py:992] (0/2) Epoch 4, batch 1250, loss[loss=0.181, simple_loss=0.2676, pruned_loss=0.04724, over 11319.00 frames. ], tot_loss[loss=0.1775, simple_loss=0.2652, pruned_loss=0.04493, over 2374435.01 frames. ], batch size: 55, lr: 4.83e-03, grad_scale: 8.0 2023-05-15 23:11:42,266 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=137953.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 23:11:44,172 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=137956.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 23:11:57,564 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-05-15 23:12:07,378 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.8217, 3.6763, 3.6396, 3.7517, 3.5225, 3.8749, 3.8689, 4.0245], device='cuda:0'), covar=tensor([0.0217, 0.0190, 0.0225, 0.0424, 0.0625, 0.0364, 0.0183, 0.0215], device='cuda:0'), in_proj_covar=tensor([0.0173, 0.0174, 0.0174, 0.0224, 0.0220, 0.0193, 0.0159, 0.0207], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0005, 0.0004, 0.0004, 0.0003, 0.0005], device='cuda:0') 2023-05-15 23:12:16,211 INFO [checkpoint.py:75] (0/2) Saving checkpoint to pruned_transducer_stateless7/exp_giga_finetune/checkpoint-38000.pt 2023-05-15 23:12:21,033 INFO [finetune.py:992] (0/2) Epoch 4, batch 1300, loss[loss=0.1709, simple_loss=0.2532, pruned_loss=0.04427, over 12183.00 frames. ], tot_loss[loss=0.1773, simple_loss=0.2649, pruned_loss=0.04483, over 2368791.56 frames. ], batch size: 31, lr: 4.83e-03, grad_scale: 8.0 2023-05-15 23:12:38,028 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.233e+02 2.960e+02 3.353e+02 3.979e+02 8.906e+02, threshold=6.705e+02, percent-clipped=5.0 2023-05-15 23:12:56,434 INFO [finetune.py:992] (0/2) Epoch 4, batch 1350, loss[loss=0.1542, simple_loss=0.2479, pruned_loss=0.03022, over 12031.00 frames. ], tot_loss[loss=0.1773, simple_loss=0.2651, pruned_loss=0.0447, over 2373675.80 frames. ], batch size: 31, lr: 4.83e-03, grad_scale: 8.0 2023-05-15 23:13:32,559 INFO [finetune.py:992] (0/2) Epoch 4, batch 1400, loss[loss=0.1671, simple_loss=0.2478, pruned_loss=0.04326, over 12192.00 frames. ], tot_loss[loss=0.1778, simple_loss=0.2657, pruned_loss=0.04494, over 2374575.68 frames. ], batch size: 31, lr: 4.83e-03, grad_scale: 8.0 2023-05-15 23:13:33,533 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.6119, 2.5512, 4.5580, 4.9248, 3.2014, 2.6626, 2.8483, 2.0674], device='cuda:0'), covar=tensor([0.1353, 0.3196, 0.0421, 0.0295, 0.0913, 0.1905, 0.2590, 0.3745], device='cuda:0'), in_proj_covar=tensor([0.0272, 0.0368, 0.0261, 0.0282, 0.0248, 0.0278, 0.0347, 0.0352], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-15 23:13:50,455 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.116e+02 2.931e+02 3.271e+02 3.889e+02 6.456e+02, threshold=6.542e+02, percent-clipped=0.0 2023-05-15 23:13:57,958 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2023-05-15 23:14:08,823 INFO [finetune.py:992] (0/2) Epoch 4, batch 1450, loss[loss=0.1563, simple_loss=0.2408, pruned_loss=0.03592, over 12345.00 frames. ], tot_loss[loss=0.1782, simple_loss=0.266, pruned_loss=0.04526, over 2362083.24 frames. ], batch size: 30, lr: 4.83e-03, grad_scale: 16.0 2023-05-15 23:14:22,898 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=3.29 vs. limit=5.0 2023-05-15 23:14:44,742 INFO [finetune.py:992] (0/2) Epoch 4, batch 1500, loss[loss=0.1835, simple_loss=0.2682, pruned_loss=0.0494, over 12062.00 frames. ], tot_loss[loss=0.1771, simple_loss=0.2646, pruned_loss=0.04482, over 2372712.35 frames. ], batch size: 40, lr: 4.83e-03, grad_scale: 16.0 2023-05-15 23:14:47,143 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.3064, 2.7411, 3.8633, 3.2537, 3.6288, 3.3817, 2.7602, 3.7253], device='cuda:0'), covar=tensor([0.0096, 0.0278, 0.0104, 0.0227, 0.0124, 0.0142, 0.0294, 0.0111], device='cuda:0'), in_proj_covar=tensor([0.0159, 0.0191, 0.0167, 0.0172, 0.0193, 0.0146, 0.0185, 0.0162], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-15 23:15:02,645 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.007e+02 2.957e+02 3.712e+02 4.461e+02 1.458e+03, threshold=7.425e+02, percent-clipped=4.0 2023-05-15 23:15:07,833 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.7968, 2.1392, 3.2063, 3.7762, 3.3294, 3.7389, 3.2629, 2.5699], device='cuda:0'), covar=tensor([0.0040, 0.0385, 0.0144, 0.0043, 0.0166, 0.0077, 0.0127, 0.0355], device='cuda:0'), in_proj_covar=tensor([0.0078, 0.0119, 0.0098, 0.0073, 0.0097, 0.0108, 0.0086, 0.0131], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-15 23:15:11,272 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.9976, 5.9873, 5.7662, 5.2736, 5.0909, 5.9152, 5.4817, 5.3771], device='cuda:0'), covar=tensor([0.0748, 0.0927, 0.0642, 0.1467, 0.0806, 0.0646, 0.1595, 0.0867], device='cuda:0'), in_proj_covar=tensor([0.0569, 0.0512, 0.0483, 0.0586, 0.0386, 0.0657, 0.0714, 0.0535], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0004, 0.0004, 0.0003], device='cuda:0') 2023-05-15 23:15:18,379 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=138248.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 23:15:20,273 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.38 vs. limit=2.0 2023-05-15 23:15:20,513 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=138251.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 23:15:21,189 INFO [finetune.py:992] (0/2) Epoch 4, batch 1550, loss[loss=0.1491, simple_loss=0.2295, pruned_loss=0.03433, over 11810.00 frames. ], tot_loss[loss=0.1759, simple_loss=0.2636, pruned_loss=0.0441, over 2374903.65 frames. ], batch size: 26, lr: 4.83e-03, grad_scale: 16.0 2023-05-15 23:15:34,768 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=138270.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 23:15:48,154 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.3996, 4.9435, 5.3312, 4.6144, 4.9819, 4.6736, 5.4334, 5.0322], device='cuda:0'), covar=tensor([0.0274, 0.0319, 0.0306, 0.0263, 0.0340, 0.0355, 0.0198, 0.0278], device='cuda:0'), in_proj_covar=tensor([0.0234, 0.0238, 0.0258, 0.0233, 0.0231, 0.0236, 0.0212, 0.0195], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-05-15 23:15:56,170 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.6366, 2.8557, 4.5654, 4.8231, 2.8684, 2.6289, 2.9665, 2.1016], device='cuda:0'), covar=tensor([0.1354, 0.2692, 0.0451, 0.0333, 0.1116, 0.1922, 0.2421, 0.3702], device='cuda:0'), in_proj_covar=tensor([0.0270, 0.0364, 0.0258, 0.0279, 0.0247, 0.0276, 0.0344, 0.0350], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-15 23:15:57,252 INFO [finetune.py:992] (0/2) Epoch 4, batch 1600, loss[loss=0.1856, simple_loss=0.2653, pruned_loss=0.05292, over 12355.00 frames. ], tot_loss[loss=0.176, simple_loss=0.2636, pruned_loss=0.04419, over 2377171.95 frames. ], batch size: 31, lr: 4.83e-03, grad_scale: 16.0 2023-05-15 23:16:05,996 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.5574, 2.8079, 4.6445, 4.8840, 2.9306, 2.6749, 2.9028, 2.0868], device='cuda:0'), covar=tensor([0.1428, 0.2938, 0.0394, 0.0337, 0.1101, 0.1874, 0.2434, 0.3672], device='cuda:0'), in_proj_covar=tensor([0.0270, 0.0365, 0.0258, 0.0279, 0.0247, 0.0276, 0.0344, 0.0350], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-15 23:16:11,509 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=138322.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 23:16:14,174 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.128e+02 2.944e+02 3.290e+02 4.014e+02 5.311e+02, threshold=6.580e+02, percent-clipped=0.0 2023-05-15 23:16:18,000 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=138331.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 23:16:22,969 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.4831, 4.2568, 4.1938, 4.5177, 3.1942, 4.0928, 2.9884, 4.2318], device='cuda:0'), covar=tensor([0.1521, 0.0595, 0.0971, 0.0726, 0.0985, 0.0544, 0.1518, 0.1579], device='cuda:0'), in_proj_covar=tensor([0.0227, 0.0259, 0.0293, 0.0349, 0.0237, 0.0236, 0.0256, 0.0385], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-15 23:16:32,550 INFO [finetune.py:992] (0/2) Epoch 4, batch 1650, loss[loss=0.1898, simple_loss=0.2793, pruned_loss=0.05017, over 11694.00 frames. ], tot_loss[loss=0.176, simple_loss=0.2635, pruned_loss=0.04425, over 2385079.60 frames. ], batch size: 48, lr: 4.83e-03, grad_scale: 16.0 2023-05-15 23:16:55,444 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=138383.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 23:17:08,931 INFO [finetune.py:992] (0/2) Epoch 4, batch 1700, loss[loss=0.1952, simple_loss=0.2812, pruned_loss=0.05457, over 12125.00 frames. ], tot_loss[loss=0.1756, simple_loss=0.263, pruned_loss=0.04406, over 2385862.62 frames. ], batch size: 39, lr: 4.83e-03, grad_scale: 16.0 2023-05-15 23:17:26,533 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.236e+02 2.850e+02 3.347e+02 4.033e+02 6.121e+02, threshold=6.694e+02, percent-clipped=0.0 2023-05-15 23:17:44,968 INFO [finetune.py:992] (0/2) Epoch 4, batch 1750, loss[loss=0.1994, simple_loss=0.2902, pruned_loss=0.05426, over 12111.00 frames. ], tot_loss[loss=0.1756, simple_loss=0.263, pruned_loss=0.04406, over 2382013.33 frames. ], batch size: 39, lr: 4.83e-03, grad_scale: 16.0 2023-05-15 23:18:20,748 INFO [finetune.py:992] (0/2) Epoch 4, batch 1800, loss[loss=0.1605, simple_loss=0.2334, pruned_loss=0.04379, over 12167.00 frames. ], tot_loss[loss=0.1754, simple_loss=0.2627, pruned_loss=0.04405, over 2381613.58 frames. ], batch size: 27, lr: 4.83e-03, grad_scale: 16.0 2023-05-15 23:18:20,922 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.0193, 5.0438, 4.8359, 4.8820, 4.4702, 5.0710, 5.0011, 5.1867], device='cuda:0'), covar=tensor([0.0205, 0.0115, 0.0176, 0.0274, 0.0744, 0.0304, 0.0127, 0.0156], device='cuda:0'), in_proj_covar=tensor([0.0176, 0.0177, 0.0177, 0.0228, 0.0225, 0.0197, 0.0162, 0.0211], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0005, 0.0004, 0.0005, 0.0004, 0.0005], device='cuda:0') 2023-05-15 23:18:37,975 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.698e+02 2.878e+02 3.516e+02 4.441e+02 6.847e+02, threshold=7.032e+02, percent-clipped=2.0 2023-05-15 23:18:54,021 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=138548.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 23:18:56,055 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=138551.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 23:18:56,581 INFO [finetune.py:992] (0/2) Epoch 4, batch 1850, loss[loss=0.1725, simple_loss=0.2738, pruned_loss=0.03555, over 12343.00 frames. ], tot_loss[loss=0.1761, simple_loss=0.2635, pruned_loss=0.04439, over 2377593.40 frames. ], batch size: 36, lr: 4.83e-03, grad_scale: 16.0 2023-05-15 23:19:28,500 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=138596.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 23:19:30,579 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=138599.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 23:19:32,995 INFO [finetune.py:992] (0/2) Epoch 4, batch 1900, loss[loss=0.185, simple_loss=0.2697, pruned_loss=0.05017, over 12015.00 frames. ], tot_loss[loss=0.1758, simple_loss=0.2631, pruned_loss=0.04421, over 2382373.89 frames. ], batch size: 42, lr: 4.83e-03, grad_scale: 16.0 2023-05-15 23:19:39,804 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=4.77 vs. limit=5.0 2023-05-15 23:19:46,629 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.8174, 3.1251, 4.7162, 4.9710, 3.3402, 2.7668, 2.9595, 2.2413], device='cuda:0'), covar=tensor([0.1291, 0.2563, 0.0447, 0.0384, 0.0960, 0.1909, 0.2427, 0.3658], device='cuda:0'), in_proj_covar=tensor([0.0274, 0.0372, 0.0264, 0.0285, 0.0251, 0.0281, 0.0350, 0.0356], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-15 23:19:49,914 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.190e+02 2.936e+02 3.314e+02 4.022e+02 9.144e+02, threshold=6.629e+02, percent-clipped=5.0 2023-05-15 23:19:50,029 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=138626.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 23:20:08,463 INFO [finetune.py:992] (0/2) Epoch 4, batch 1950, loss[loss=0.1744, simple_loss=0.2735, pruned_loss=0.03768, over 12104.00 frames. ], tot_loss[loss=0.1771, simple_loss=0.264, pruned_loss=0.04504, over 2379232.47 frames. ], batch size: 32, lr: 4.83e-03, grad_scale: 16.0 2023-05-15 23:20:27,589 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=138678.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 23:20:29,726 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=138681.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 23:20:33,270 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=138686.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 23:20:40,993 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.3749, 4.9601, 5.1765, 5.2953, 5.0800, 5.2579, 5.1336, 3.0743], device='cuda:0'), covar=tensor([0.0055, 0.0058, 0.0069, 0.0042, 0.0043, 0.0078, 0.0075, 0.0581], device='cuda:0'), in_proj_covar=tensor([0.0064, 0.0072, 0.0076, 0.0068, 0.0056, 0.0086, 0.0074, 0.0091], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-15 23:20:41,011 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=138697.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 23:20:43,866 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.9061, 2.0525, 3.3364, 2.9102, 3.2054, 2.9875, 2.2246, 3.2349], device='cuda:0'), covar=tensor([0.0137, 0.0410, 0.0156, 0.0230, 0.0170, 0.0165, 0.0367, 0.0154], device='cuda:0'), in_proj_covar=tensor([0.0158, 0.0189, 0.0168, 0.0170, 0.0193, 0.0146, 0.0184, 0.0160], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-15 23:20:44,389 INFO [finetune.py:992] (0/2) Epoch 4, batch 2000, loss[loss=0.185, simple_loss=0.282, pruned_loss=0.04402, over 11825.00 frames. ], tot_loss[loss=0.1771, simple_loss=0.2641, pruned_loss=0.04509, over 2374549.65 frames. ], batch size: 44, lr: 4.83e-03, grad_scale: 16.0 2023-05-15 23:20:46,754 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.7308, 2.8004, 4.6868, 4.8624, 2.8067, 2.7412, 2.8793, 2.1703], device='cuda:0'), covar=tensor([0.1375, 0.3058, 0.0404, 0.0333, 0.1198, 0.1921, 0.2537, 0.3653], device='cuda:0'), in_proj_covar=tensor([0.0272, 0.0370, 0.0263, 0.0283, 0.0250, 0.0280, 0.0348, 0.0353], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-15 23:21:02,093 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.866e+02 2.852e+02 3.308e+02 4.055e+02 7.312e+02, threshold=6.616e+02, percent-clipped=3.0 2023-05-15 23:21:13,594 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=138742.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 23:21:14,272 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([6.0256, 6.0128, 5.8170, 5.4220, 5.1673, 5.9358, 5.5480, 5.4036], device='cuda:0'), covar=tensor([0.0722, 0.0844, 0.0549, 0.1389, 0.0632, 0.0643, 0.1416, 0.0885], device='cuda:0'), in_proj_covar=tensor([0.0565, 0.0512, 0.0483, 0.0579, 0.0383, 0.0655, 0.0716, 0.0535], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0004, 0.0004, 0.0003], device='cuda:0') 2023-05-15 23:21:17,173 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=138747.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 23:21:20,480 INFO [finetune.py:992] (0/2) Epoch 4, batch 2050, loss[loss=0.1994, simple_loss=0.2839, pruned_loss=0.05746, over 10558.00 frames. ], tot_loss[loss=0.1765, simple_loss=0.2639, pruned_loss=0.0445, over 2381258.92 frames. ], batch size: 68, lr: 4.83e-03, grad_scale: 16.0 2023-05-15 23:21:24,890 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=138758.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 23:21:42,872 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.59 vs. limit=2.0 2023-05-15 23:21:56,180 INFO [finetune.py:992] (0/2) Epoch 4, batch 2100, loss[loss=0.1552, simple_loss=0.2316, pruned_loss=0.03938, over 12165.00 frames. ], tot_loss[loss=0.1761, simple_loss=0.2636, pruned_loss=0.04433, over 2386576.86 frames. ], batch size: 29, lr: 4.83e-03, grad_scale: 16.0 2023-05-15 23:22:13,177 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.939e+02 2.809e+02 3.283e+02 4.095e+02 9.466e+02, threshold=6.566e+02, percent-clipped=3.0 2023-05-15 23:22:13,601 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=3.35 vs. limit=5.0 2023-05-15 23:22:32,001 INFO [finetune.py:992] (0/2) Epoch 4, batch 2150, loss[loss=0.1845, simple_loss=0.2739, pruned_loss=0.04758, over 12074.00 frames. ], tot_loss[loss=0.176, simple_loss=0.2634, pruned_loss=0.04431, over 2384418.64 frames. ], batch size: 42, lr: 4.83e-03, grad_scale: 16.0 2023-05-15 23:22:34,275 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.8721, 4.5440, 4.0328, 4.1180, 4.6344, 4.0689, 4.2479, 3.9809], device='cuda:0'), covar=tensor([0.1522, 0.1157, 0.1461, 0.2001, 0.1122, 0.1972, 0.1644, 0.1535], device='cuda:0'), in_proj_covar=tensor([0.0332, 0.0456, 0.0355, 0.0405, 0.0428, 0.0411, 0.0362, 0.0356], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-15 23:23:00,369 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([6.1211, 6.0866, 5.8466, 5.3365, 5.2134, 5.9764, 5.5285, 5.3803], device='cuda:0'), covar=tensor([0.0597, 0.0828, 0.0615, 0.1532, 0.0635, 0.0696, 0.1474, 0.0889], device='cuda:0'), in_proj_covar=tensor([0.0565, 0.0514, 0.0484, 0.0581, 0.0383, 0.0657, 0.0719, 0.0535], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0004, 0.0004, 0.0003], device='cuda:0') 2023-05-15 23:23:03,199 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.0205, 5.9845, 5.4064, 5.4953, 5.9909, 5.4254, 5.6456, 5.5095], device='cuda:0'), covar=tensor([0.1685, 0.0959, 0.1116, 0.1897, 0.1079, 0.2176, 0.1676, 0.1206], device='cuda:0'), in_proj_covar=tensor([0.0331, 0.0456, 0.0355, 0.0405, 0.0429, 0.0410, 0.0362, 0.0356], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-15 23:23:08,027 INFO [finetune.py:992] (0/2) Epoch 4, batch 2200, loss[loss=0.1602, simple_loss=0.2448, pruned_loss=0.0378, over 12187.00 frames. ], tot_loss[loss=0.1757, simple_loss=0.2631, pruned_loss=0.0441, over 2384581.34 frames. ], batch size: 31, lr: 4.83e-03, grad_scale: 16.0 2023-05-15 23:23:13,068 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=138909.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 23:23:24,913 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.112e+02 2.850e+02 3.488e+02 4.120e+02 8.663e+02, threshold=6.977e+02, percent-clipped=4.0 2023-05-15 23:23:25,063 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=138926.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 23:23:25,868 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.2270, 5.1815, 4.9931, 5.0148, 4.6771, 5.1933, 5.1022, 5.4662], device='cuda:0'), covar=tensor([0.0165, 0.0131, 0.0154, 0.0318, 0.0675, 0.0196, 0.0160, 0.0149], device='cuda:0'), in_proj_covar=tensor([0.0177, 0.0178, 0.0180, 0.0230, 0.0228, 0.0199, 0.0163, 0.0214], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0005, 0.0005, 0.0005, 0.0004, 0.0005], device='cuda:0') 2023-05-15 23:23:29,404 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.4721, 4.7783, 2.9783, 2.6920, 4.1218, 2.6430, 4.0960, 3.4094], device='cuda:0'), covar=tensor([0.0516, 0.0383, 0.0990, 0.1347, 0.0223, 0.1176, 0.0383, 0.0690], device='cuda:0'), in_proj_covar=tensor([0.0185, 0.0241, 0.0171, 0.0197, 0.0134, 0.0177, 0.0188, 0.0170], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-15 23:23:33,117 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2023-05-15 23:23:42,215 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.1516, 2.5078, 3.6379, 3.0773, 3.4301, 3.2179, 2.4878, 3.5916], device='cuda:0'), covar=tensor([0.0122, 0.0305, 0.0139, 0.0190, 0.0127, 0.0153, 0.0331, 0.0096], device='cuda:0'), in_proj_covar=tensor([0.0160, 0.0190, 0.0168, 0.0170, 0.0194, 0.0146, 0.0184, 0.0162], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-15 23:23:43,428 INFO [finetune.py:992] (0/2) Epoch 4, batch 2250, loss[loss=0.1644, simple_loss=0.2583, pruned_loss=0.03526, over 12255.00 frames. ], tot_loss[loss=0.1753, simple_loss=0.2627, pruned_loss=0.04393, over 2385156.04 frames. ], batch size: 37, lr: 4.83e-03, grad_scale: 16.0 2023-05-15 23:23:52,909 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.8766, 4.5123, 4.8437, 4.2660, 4.5292, 4.3269, 4.8630, 4.4661], device='cuda:0'), covar=tensor([0.0247, 0.0316, 0.0246, 0.0256, 0.0327, 0.0280, 0.0198, 0.0474], device='cuda:0'), in_proj_covar=tensor([0.0235, 0.0241, 0.0259, 0.0236, 0.0233, 0.0237, 0.0215, 0.0200], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-05-15 23:23:56,422 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=138970.0, num_to_drop=1, layers_to_drop={2} 2023-05-15 23:23:59,153 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=138974.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 23:24:02,123 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=138978.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 23:24:16,160 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.9910, 5.9677, 5.7164, 5.2134, 5.1386, 5.8581, 5.4252, 5.2649], device='cuda:0'), covar=tensor([0.0737, 0.0888, 0.0601, 0.1498, 0.0563, 0.0714, 0.1596, 0.0864], device='cuda:0'), in_proj_covar=tensor([0.0569, 0.0518, 0.0488, 0.0585, 0.0385, 0.0663, 0.0724, 0.0538], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0004, 0.0004, 0.0003], device='cuda:0') 2023-05-15 23:24:19,921 INFO [finetune.py:992] (0/2) Epoch 4, batch 2300, loss[loss=0.1471, simple_loss=0.2326, pruned_loss=0.03079, over 11800.00 frames. ], tot_loss[loss=0.1755, simple_loss=0.2631, pruned_loss=0.04391, over 2383626.28 frames. ], batch size: 26, lr: 4.83e-03, grad_scale: 16.0 2023-05-15 23:24:37,530 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.146e+02 2.924e+02 3.589e+02 4.110e+02 6.308e+02, threshold=7.178e+02, percent-clipped=0.0 2023-05-15 23:24:37,606 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=139026.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 23:24:39,749 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.4367, 4.9785, 5.4064, 4.7122, 5.0321, 4.7816, 5.4517, 5.1284], device='cuda:0'), covar=tensor([0.0210, 0.0252, 0.0203, 0.0210, 0.0271, 0.0244, 0.0169, 0.0253], device='cuda:0'), in_proj_covar=tensor([0.0235, 0.0241, 0.0258, 0.0236, 0.0233, 0.0237, 0.0214, 0.0199], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-05-15 23:24:40,099 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.35 vs. limit=2.0 2023-05-15 23:24:45,416 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=139037.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 23:24:48,897 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=139042.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 23:24:56,068 INFO [finetune.py:992] (0/2) Epoch 4, batch 2350, loss[loss=0.1897, simple_loss=0.2771, pruned_loss=0.05113, over 12032.00 frames. ], tot_loss[loss=0.1757, simple_loss=0.2636, pruned_loss=0.04388, over 2383938.59 frames. ], batch size: 42, lr: 4.82e-03, grad_scale: 16.0 2023-05-15 23:24:56,829 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=139053.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 23:25:06,912 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.6760, 3.6954, 3.3674, 3.3558, 3.0965, 3.0236, 3.9087, 2.4482], device='cuda:0'), covar=tensor([0.0327, 0.0138, 0.0186, 0.0156, 0.0336, 0.0286, 0.0101, 0.0419], device='cuda:0'), in_proj_covar=tensor([0.0180, 0.0155, 0.0148, 0.0176, 0.0199, 0.0191, 0.0153, 0.0194], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], device='cuda:0') 2023-05-15 23:25:26,289 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.46 vs. limit=2.0 2023-05-15 23:25:31,482 INFO [finetune.py:992] (0/2) Epoch 4, batch 2400, loss[loss=0.1897, simple_loss=0.2815, pruned_loss=0.04895, over 11852.00 frames. ], tot_loss[loss=0.1758, simple_loss=0.2636, pruned_loss=0.04396, over 2383243.57 frames. ], batch size: 44, lr: 4.82e-03, grad_scale: 16.0 2023-05-15 23:25:48,264 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.104e+02 2.841e+02 3.377e+02 3.980e+02 6.478e+02, threshold=6.753e+02, percent-clipped=0.0 2023-05-15 23:26:08,264 INFO [finetune.py:992] (0/2) Epoch 4, batch 2450, loss[loss=0.1739, simple_loss=0.2647, pruned_loss=0.0416, over 11236.00 frames. ], tot_loss[loss=0.1757, simple_loss=0.2635, pruned_loss=0.04398, over 2385695.88 frames. ], batch size: 55, lr: 4.82e-03, grad_scale: 16.0 2023-05-15 23:26:37,558 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.45 vs. limit=2.0 2023-05-15 23:26:43,800 INFO [finetune.py:992] (0/2) Epoch 4, batch 2500, loss[loss=0.1793, simple_loss=0.2757, pruned_loss=0.04146, over 12314.00 frames. ], tot_loss[loss=0.1761, simple_loss=0.264, pruned_loss=0.0441, over 2380070.60 frames. ], batch size: 34, lr: 4.82e-03, grad_scale: 16.0 2023-05-15 23:27:00,669 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.089e+02 2.899e+02 3.236e+02 3.898e+02 7.132e+02, threshold=6.472e+02, percent-clipped=1.0 2023-05-15 23:27:03,220 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.6943, 2.9484, 4.4260, 4.7263, 3.0963, 2.8192, 3.0359, 1.9735], device='cuda:0'), covar=tensor([0.1308, 0.2561, 0.0460, 0.0343, 0.0999, 0.1872, 0.2244, 0.3859], device='cuda:0'), in_proj_covar=tensor([0.0274, 0.0371, 0.0264, 0.0285, 0.0251, 0.0281, 0.0350, 0.0355], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-15 23:27:09,441 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=139238.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 23:27:19,333 INFO [finetune.py:992] (0/2) Epoch 4, batch 2550, loss[loss=0.164, simple_loss=0.2609, pruned_loss=0.03359, over 12295.00 frames. ], tot_loss[loss=0.176, simple_loss=0.2639, pruned_loss=0.04405, over 2371762.51 frames. ], batch size: 37, lr: 4.82e-03, grad_scale: 16.0 2023-05-15 23:27:28,645 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=139265.0, num_to_drop=1, layers_to_drop={0} 2023-05-15 23:27:36,514 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=139276.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 23:27:38,530 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.9588, 5.8803, 5.5936, 5.0719, 5.0315, 5.7641, 5.3832, 5.1656], device='cuda:0'), covar=tensor([0.0581, 0.0876, 0.0689, 0.1489, 0.0626, 0.0667, 0.1390, 0.0945], device='cuda:0'), in_proj_covar=tensor([0.0568, 0.0519, 0.0485, 0.0583, 0.0385, 0.0661, 0.0718, 0.0534], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0004, 0.0004, 0.0003], device='cuda:0') 2023-05-15 23:27:47,064 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=139289.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 23:27:49,902 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.2656, 4.4556, 2.6006, 2.4815, 3.8442, 2.4426, 3.9475, 3.0361], device='cuda:0'), covar=tensor([0.0603, 0.0459, 0.1148, 0.1454, 0.0280, 0.1262, 0.0360, 0.0803], device='cuda:0'), in_proj_covar=tensor([0.0185, 0.0241, 0.0171, 0.0197, 0.0135, 0.0175, 0.0186, 0.0169], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-15 23:27:54,195 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=139299.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 23:27:56,010 INFO [finetune.py:992] (0/2) Epoch 4, batch 2600, loss[loss=0.1728, simple_loss=0.2626, pruned_loss=0.04146, over 12117.00 frames. ], tot_loss[loss=0.1749, simple_loss=0.2631, pruned_loss=0.04336, over 2378978.93 frames. ], batch size: 33, lr: 4.82e-03, grad_scale: 16.0 2023-05-15 23:28:09,204 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=4.62 vs. limit=5.0 2023-05-15 23:28:13,101 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.024e+02 2.986e+02 3.375e+02 3.883e+02 7.968e+02, threshold=6.751e+02, percent-clipped=3.0 2023-05-15 23:28:17,031 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.6120, 2.8042, 4.5480, 4.7825, 2.9484, 2.6383, 2.8712, 2.0393], device='cuda:0'), covar=tensor([0.1347, 0.2725, 0.0409, 0.0343, 0.1058, 0.1873, 0.2604, 0.3694], device='cuda:0'), in_proj_covar=tensor([0.0274, 0.0371, 0.0265, 0.0285, 0.0252, 0.0281, 0.0350, 0.0355], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-15 23:28:21,149 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=139337.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 23:28:21,183 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=139337.0, num_to_drop=1, layers_to_drop={0} 2023-05-15 23:28:24,486 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=139342.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 23:28:30,055 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=139350.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 23:28:31,312 INFO [finetune.py:992] (0/2) Epoch 4, batch 2650, loss[loss=0.1571, simple_loss=0.2415, pruned_loss=0.03637, over 12335.00 frames. ], tot_loss[loss=0.1753, simple_loss=0.2633, pruned_loss=0.04369, over 2371439.35 frames. ], batch size: 31, lr: 4.82e-03, grad_scale: 16.0 2023-05-15 23:28:32,201 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=139353.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 23:28:54,714 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=139385.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 23:28:58,207 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=139390.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 23:29:06,364 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=139401.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 23:29:07,050 INFO [finetune.py:992] (0/2) Epoch 4, batch 2700, loss[loss=0.1662, simple_loss=0.2541, pruned_loss=0.03915, over 12293.00 frames. ], tot_loss[loss=0.1752, simple_loss=0.2632, pruned_loss=0.04358, over 2362982.53 frames. ], batch size: 33, lr: 4.82e-03, grad_scale: 16.0 2023-05-15 23:29:24,185 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.053e+02 2.943e+02 3.380e+02 4.126e+02 9.261e+02, threshold=6.761e+02, percent-clipped=2.0 2023-05-15 23:29:43,193 INFO [finetune.py:992] (0/2) Epoch 4, batch 2750, loss[loss=0.1541, simple_loss=0.2423, pruned_loss=0.03299, over 12340.00 frames. ], tot_loss[loss=0.1756, simple_loss=0.2639, pruned_loss=0.04371, over 2363419.03 frames. ], batch size: 31, lr: 4.82e-03, grad_scale: 16.0 2023-05-15 23:30:18,348 INFO [finetune.py:992] (0/2) Epoch 4, batch 2800, loss[loss=0.176, simple_loss=0.2616, pruned_loss=0.04513, over 12364.00 frames. ], tot_loss[loss=0.1756, simple_loss=0.2636, pruned_loss=0.04381, over 2361095.73 frames. ], batch size: 38, lr: 4.82e-03, grad_scale: 16.0 2023-05-15 23:30:35,532 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.134e+02 2.967e+02 3.317e+02 4.034e+02 8.886e+02, threshold=6.634e+02, percent-clipped=2.0 2023-05-15 23:30:53,847 INFO [finetune.py:992] (0/2) Epoch 4, batch 2850, loss[loss=0.1663, simple_loss=0.249, pruned_loss=0.04183, over 12036.00 frames. ], tot_loss[loss=0.1751, simple_loss=0.263, pruned_loss=0.04367, over 2365066.47 frames. ], batch size: 31, lr: 4.82e-03, grad_scale: 16.0 2023-05-15 23:31:03,337 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=139565.0, num_to_drop=1, layers_to_drop={1} 2023-05-15 23:31:08,373 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=139571.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 23:31:17,022 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=139582.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 23:31:18,468 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([1.9530, 3.9319, 3.9659, 4.2300, 2.7460, 3.6753, 2.4153, 3.8503], device='cuda:0'), covar=tensor([0.2092, 0.0797, 0.1124, 0.0773, 0.1342, 0.0807, 0.2190, 0.1340], device='cuda:0'), in_proj_covar=tensor([0.0228, 0.0263, 0.0296, 0.0355, 0.0239, 0.0240, 0.0259, 0.0387], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-15 23:31:25,471 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=139594.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 23:31:31,358 INFO [finetune.py:992] (0/2) Epoch 4, batch 2900, loss[loss=0.2126, simple_loss=0.2893, pruned_loss=0.06796, over 7981.00 frames. ], tot_loss[loss=0.1753, simple_loss=0.263, pruned_loss=0.04379, over 2356316.01 frames. ], batch size: 101, lr: 4.82e-03, grad_scale: 16.0 2023-05-15 23:31:32,299 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.7649, 2.8376, 4.5628, 4.7966, 3.0561, 2.7774, 3.1394, 2.1305], device='cuda:0'), covar=tensor([0.1374, 0.3098, 0.0459, 0.0362, 0.1071, 0.1998, 0.2260, 0.3833], device='cuda:0'), in_proj_covar=tensor([0.0275, 0.0372, 0.0265, 0.0286, 0.0253, 0.0282, 0.0351, 0.0356], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-15 23:31:39,011 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=139613.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 23:31:48,248 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.018e+02 2.682e+02 3.469e+02 4.078e+02 1.175e+03, threshold=6.939e+02, percent-clipped=1.0 2023-05-15 23:31:52,700 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=139632.0, num_to_drop=1, layers_to_drop={3} 2023-05-15 23:31:52,783 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=139632.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 23:31:54,240 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.2879, 4.0033, 4.0838, 4.3034, 3.0743, 3.7761, 2.6480, 3.9991], device='cuda:0'), covar=tensor([0.1523, 0.0658, 0.0859, 0.0630, 0.0968, 0.0619, 0.1695, 0.1148], device='cuda:0'), in_proj_covar=tensor([0.0226, 0.0261, 0.0292, 0.0351, 0.0237, 0.0237, 0.0256, 0.0382], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-15 23:32:00,629 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=139643.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 23:32:01,920 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=139645.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 23:32:06,698 INFO [finetune.py:992] (0/2) Epoch 4, batch 2950, loss[loss=0.1645, simple_loss=0.2561, pruned_loss=0.0365, over 12154.00 frames. ], tot_loss[loss=0.1759, simple_loss=0.2636, pruned_loss=0.04411, over 2361308.45 frames. ], batch size: 34, lr: 4.82e-03, grad_scale: 16.0 2023-05-15 23:32:42,132 INFO [finetune.py:992] (0/2) Epoch 4, batch 3000, loss[loss=0.2109, simple_loss=0.2869, pruned_loss=0.06749, over 7769.00 frames. ], tot_loss[loss=0.1762, simple_loss=0.2638, pruned_loss=0.04431, over 2345103.68 frames. ], batch size: 98, lr: 4.82e-03, grad_scale: 16.0 2023-05-15 23:32:42,133 INFO [finetune.py:1017] (0/2) Computing validation loss 2023-05-15 23:32:56,979 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.5312, 4.0866, 3.6525, 3.8085, 4.1269, 3.4444, 3.7318, 3.4326], device='cuda:0'), covar=tensor([0.1700, 0.0966, 0.1272, 0.1563, 0.0989, 0.1997, 0.1320, 0.1682], device='cuda:0'), in_proj_covar=tensor([0.0333, 0.0464, 0.0358, 0.0412, 0.0436, 0.0417, 0.0369, 0.0359], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-15 23:33:00,431 INFO [finetune.py:1026] (0/2) Epoch 4, validation: loss=0.3291, simple_loss=0.4045, pruned_loss=0.1269, over 1020973.00 frames. 2023-05-15 23:33:00,431 INFO [finetune.py:1027] (0/2) Maximum memory allocated so far is 12745MB 2023-05-15 23:33:06,330 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.7502, 2.5319, 4.8133, 5.1226, 3.2896, 2.5948, 2.8919, 1.9697], device='cuda:0'), covar=tensor([0.1544, 0.3719, 0.0405, 0.0276, 0.0961, 0.2181, 0.2833, 0.4843], device='cuda:0'), in_proj_covar=tensor([0.0272, 0.0367, 0.0263, 0.0283, 0.0250, 0.0278, 0.0347, 0.0351], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-15 23:33:17,218 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.814e+02 2.801e+02 3.363e+02 4.100e+02 7.836e+02, threshold=6.726e+02, percent-clipped=4.0 2023-05-15 23:33:18,911 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.6984, 2.8934, 3.6582, 4.7425, 3.9835, 4.6843, 3.9378, 3.3069], device='cuda:0'), covar=tensor([0.0023, 0.0299, 0.0157, 0.0025, 0.0129, 0.0057, 0.0121, 0.0279], device='cuda:0'), in_proj_covar=tensor([0.0078, 0.0117, 0.0097, 0.0072, 0.0096, 0.0108, 0.0085, 0.0131], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-15 23:33:35,833 INFO [finetune.py:992] (0/2) Epoch 4, batch 3050, loss[loss=0.1881, simple_loss=0.2791, pruned_loss=0.04856, over 12118.00 frames. ], tot_loss[loss=0.1766, simple_loss=0.2641, pruned_loss=0.04454, over 2345803.00 frames. ], batch size: 38, lr: 4.82e-03, grad_scale: 16.0 2023-05-15 23:33:47,951 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=139769.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 23:34:02,241 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=139789.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 23:34:03,718 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.0161, 2.2522, 3.4692, 2.9712, 3.2934, 3.1330, 2.4057, 3.4220], device='cuda:0'), covar=tensor([0.0099, 0.0303, 0.0118, 0.0201, 0.0147, 0.0124, 0.0291, 0.0107], device='cuda:0'), in_proj_covar=tensor([0.0160, 0.0192, 0.0169, 0.0171, 0.0195, 0.0148, 0.0184, 0.0164], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-15 23:34:06,434 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([6.1001, 6.0223, 5.8258, 5.3092, 5.2130, 5.9848, 5.5950, 5.3600], device='cuda:0'), covar=tensor([0.0650, 0.0862, 0.0629, 0.1575, 0.0617, 0.0716, 0.1475, 0.1055], device='cuda:0'), in_proj_covar=tensor([0.0570, 0.0515, 0.0486, 0.0586, 0.0382, 0.0660, 0.0716, 0.0534], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0004, 0.0004, 0.0003], device='cuda:0') 2023-05-15 23:34:11,590 INFO [finetune.py:992] (0/2) Epoch 4, batch 3100, loss[loss=0.2001, simple_loss=0.2979, pruned_loss=0.0511, over 12129.00 frames. ], tot_loss[loss=0.1763, simple_loss=0.2638, pruned_loss=0.04438, over 2341610.63 frames. ], batch size: 39, lr: 4.82e-03, grad_scale: 16.0 2023-05-15 23:34:13,146 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=139804.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 23:34:14,614 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=139806.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 23:34:29,391 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.635e+02 2.867e+02 3.242e+02 3.780e+02 8.522e+02, threshold=6.484e+02, percent-clipped=1.0 2023-05-15 23:34:32,412 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=139830.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 23:34:47,161 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=139850.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 23:34:47,231 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=139850.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 23:34:47,933 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.5595, 3.0319, 3.8258, 2.2487, 2.6711, 3.1189, 2.9050, 3.3213], device='cuda:0'), covar=tensor([0.0480, 0.0875, 0.0383, 0.1104, 0.1466, 0.1243, 0.1053, 0.0837], device='cuda:0'), in_proj_covar=tensor([0.0228, 0.0223, 0.0228, 0.0175, 0.0231, 0.0274, 0.0221, 0.0248], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:0') 2023-05-15 23:34:48,374 INFO [finetune.py:992] (0/2) Epoch 4, batch 3150, loss[loss=0.237, simple_loss=0.3178, pruned_loss=0.07812, over 7795.00 frames. ], tot_loss[loss=0.1759, simple_loss=0.2635, pruned_loss=0.04415, over 2346435.26 frames. ], batch size: 98, lr: 4.82e-03, grad_scale: 16.0 2023-05-15 23:34:54,693 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.1039, 5.9803, 5.4470, 5.5303, 6.0163, 5.3231, 5.5422, 5.5423], device='cuda:0'), covar=tensor([0.1303, 0.0845, 0.1042, 0.1687, 0.0956, 0.1948, 0.1619, 0.1136], device='cuda:0'), in_proj_covar=tensor([0.0331, 0.0461, 0.0355, 0.0408, 0.0432, 0.0412, 0.0368, 0.0356], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-15 23:34:56,153 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.6704, 2.5430, 4.4740, 4.8005, 3.2157, 2.6090, 2.7619, 1.8911], device='cuda:0'), covar=tensor([0.1330, 0.3325, 0.0423, 0.0312, 0.0937, 0.2057, 0.2673, 0.4133], device='cuda:0'), in_proj_covar=tensor([0.0273, 0.0369, 0.0263, 0.0284, 0.0252, 0.0280, 0.0348, 0.0352], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-15 23:34:57,562 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=139865.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 23:34:59,004 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=139867.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 23:35:06,339 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=4.41 vs. limit=5.0 2023-05-15 23:35:17,822 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=139894.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 23:35:23,437 INFO [finetune.py:992] (0/2) Epoch 4, batch 3200, loss[loss=0.1728, simple_loss=0.2698, pruned_loss=0.03792, over 12126.00 frames. ], tot_loss[loss=0.1749, simple_loss=0.2626, pruned_loss=0.04359, over 2356898.85 frames. ], batch size: 38, lr: 4.82e-03, grad_scale: 16.0 2023-05-15 23:35:26,123 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.43 vs. limit=2.0 2023-05-15 23:35:30,060 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=139911.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 23:35:40,347 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.887e+02 2.833e+02 3.398e+02 3.955e+02 6.334e+02, threshold=6.796e+02, percent-clipped=0.0 2023-05-15 23:35:41,132 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=139927.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 23:35:44,800 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=139932.0, num_to_drop=1, layers_to_drop={0} 2023-05-15 23:35:48,904 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=139938.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 23:35:51,733 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=139942.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 23:35:53,876 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=139945.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 23:35:58,568 INFO [finetune.py:992] (0/2) Epoch 4, batch 3250, loss[loss=0.1575, simple_loss=0.2528, pruned_loss=0.03108, over 12287.00 frames. ], tot_loss[loss=0.1751, simple_loss=0.2626, pruned_loss=0.04383, over 2363068.48 frames. ], batch size: 34, lr: 4.82e-03, grad_scale: 16.0 2023-05-15 23:36:18,961 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=139980.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 23:36:28,684 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=139993.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 23:36:33,880 INFO [checkpoint.py:75] (0/2) Saving checkpoint to pruned_transducer_stateless7/exp_giga_finetune/checkpoint-40000.pt 2023-05-15 23:36:38,555 INFO [finetune.py:992] (0/2) Epoch 4, batch 3300, loss[loss=0.1761, simple_loss=0.2665, pruned_loss=0.04289, over 12066.00 frames. ], tot_loss[loss=0.175, simple_loss=0.2626, pruned_loss=0.0437, over 2368777.73 frames. ], batch size: 37, lr: 4.82e-03, grad_scale: 16.0 2023-05-15 23:36:55,089 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.937e+02 2.787e+02 3.285e+02 4.062e+02 8.144e+02, threshold=6.569e+02, percent-clipped=3.0 2023-05-15 23:37:13,355 INFO [finetune.py:992] (0/2) Epoch 4, batch 3350, loss[loss=0.1826, simple_loss=0.277, pruned_loss=0.0441, over 12105.00 frames. ], tot_loss[loss=0.1767, simple_loss=0.2646, pruned_loss=0.04447, over 2368627.33 frames. ], batch size: 33, lr: 4.82e-03, grad_scale: 16.0 2023-05-15 23:37:17,850 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.7884, 5.7489, 5.5326, 5.0928, 5.0411, 5.6848, 5.2932, 5.1100], device='cuda:0'), covar=tensor([0.0647, 0.0838, 0.0610, 0.1464, 0.0716, 0.0727, 0.1582, 0.0970], device='cuda:0'), in_proj_covar=tensor([0.0574, 0.0519, 0.0489, 0.0591, 0.0383, 0.0667, 0.0726, 0.0539], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0004, 0.0004, 0.0003], device='cuda:0') 2023-05-15 23:37:21,115 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.88 vs. limit=2.0 2023-05-15 23:37:49,418 INFO [finetune.py:992] (0/2) Epoch 4, batch 3400, loss[loss=0.1836, simple_loss=0.2735, pruned_loss=0.04679, over 12297.00 frames. ], tot_loss[loss=0.1762, simple_loss=0.2638, pruned_loss=0.04431, over 2373169.71 frames. ], batch size: 33, lr: 4.82e-03, grad_scale: 16.0 2023-05-15 23:37:50,342 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=140103.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 23:37:55,284 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=140110.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 23:38:05,622 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=140125.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 23:38:05,825 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.1728, 4.5388, 4.0192, 4.9590, 4.4018, 2.8546, 4.1436, 3.0610], device='cuda:0'), covar=tensor([0.0800, 0.0820, 0.1274, 0.0372, 0.1083, 0.1470, 0.0942, 0.2986], device='cuda:0'), in_proj_covar=tensor([0.0293, 0.0364, 0.0344, 0.0254, 0.0352, 0.0256, 0.0330, 0.0351], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-15 23:38:06,247 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.070e+02 2.929e+02 3.583e+02 4.170e+02 6.929e+02, threshold=7.166e+02, percent-clipped=2.0 2023-05-15 23:38:12,214 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-05-15 23:38:20,397 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=140145.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 23:38:25,302 INFO [finetune.py:992] (0/2) Epoch 4, batch 3450, loss[loss=0.179, simple_loss=0.2743, pruned_loss=0.04182, over 12273.00 frames. ], tot_loss[loss=0.1768, simple_loss=0.2643, pruned_loss=0.04462, over 2381726.63 frames. ], batch size: 33, lr: 4.82e-03, grad_scale: 32.0 2023-05-15 23:38:31,206 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=140160.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 23:38:32,688 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=140162.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 23:38:34,156 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=140164.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 23:38:39,139 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=140171.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 23:39:01,152 INFO [finetune.py:992] (0/2) Epoch 4, batch 3500, loss[loss=0.2029, simple_loss=0.2862, pruned_loss=0.0598, over 10468.00 frames. ], tot_loss[loss=0.1764, simple_loss=0.2639, pruned_loss=0.04445, over 2375170.58 frames. ], batch size: 68, lr: 4.82e-03, grad_scale: 32.0 2023-05-15 23:39:04,142 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=140206.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 23:39:08,644 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.6716, 2.8462, 3.7940, 4.7594, 3.9920, 4.6343, 4.0461, 3.2502], device='cuda:0'), covar=tensor([0.0024, 0.0316, 0.0110, 0.0027, 0.0098, 0.0061, 0.0084, 0.0299], device='cuda:0'), in_proj_covar=tensor([0.0079, 0.0118, 0.0099, 0.0074, 0.0098, 0.0110, 0.0086, 0.0134], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-15 23:39:18,480 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.921e+02 2.882e+02 3.287e+02 3.945e+02 9.868e+02, threshold=6.574e+02, percent-clipped=1.0 2023-05-15 23:39:19,364 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=140227.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 23:39:27,207 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=140238.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 23:39:37,604 INFO [finetune.py:992] (0/2) Epoch 4, batch 3550, loss[loss=0.176, simple_loss=0.2673, pruned_loss=0.04233, over 12151.00 frames. ], tot_loss[loss=0.1752, simple_loss=0.2626, pruned_loss=0.0439, over 2373539.56 frames. ], batch size: 34, lr: 4.81e-03, grad_scale: 16.0 2023-05-15 23:39:42,752 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=140259.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 23:39:53,891 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=140275.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 23:40:02,443 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=140286.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 23:40:11,435 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.7651, 5.7205, 5.5432, 5.0668, 4.9570, 5.6362, 5.2712, 5.0812], device='cuda:0'), covar=tensor([0.0664, 0.0825, 0.0595, 0.1315, 0.0671, 0.0635, 0.1329, 0.0993], device='cuda:0'), in_proj_covar=tensor([0.0573, 0.0518, 0.0486, 0.0589, 0.0382, 0.0666, 0.0723, 0.0540], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0004, 0.0004, 0.0003], device='cuda:0') 2023-05-15 23:40:13,470 INFO [finetune.py:992] (0/2) Epoch 4, batch 3600, loss[loss=0.1674, simple_loss=0.2585, pruned_loss=0.03811, over 12107.00 frames. ], tot_loss[loss=0.175, simple_loss=0.2623, pruned_loss=0.04388, over 2367944.01 frames. ], batch size: 33, lr: 4.81e-03, grad_scale: 16.0 2023-05-15 23:40:26,654 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=140320.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 23:40:31,535 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.016e+02 2.818e+02 3.366e+02 3.811e+02 1.666e+03, threshold=6.733e+02, percent-clipped=4.0 2023-05-15 23:40:48,984 INFO [finetune.py:992] (0/2) Epoch 4, batch 3650, loss[loss=0.1628, simple_loss=0.2449, pruned_loss=0.0404, over 12352.00 frames. ], tot_loss[loss=0.1747, simple_loss=0.2618, pruned_loss=0.04382, over 2369407.05 frames. ], batch size: 31, lr: 4.81e-03, grad_scale: 16.0 2023-05-15 23:41:14,686 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=140387.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 23:41:20,557 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 2023-05-15 23:41:25,429 INFO [finetune.py:992] (0/2) Epoch 4, batch 3700, loss[loss=0.1696, simple_loss=0.2483, pruned_loss=0.04544, over 12366.00 frames. ], tot_loss[loss=0.1745, simple_loss=0.2619, pruned_loss=0.04356, over 2376116.18 frames. ], batch size: 30, lr: 4.81e-03, grad_scale: 16.0 2023-05-15 23:41:31,369 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.7642, 2.8443, 4.5897, 4.7255, 3.0605, 2.7926, 3.0419, 2.1484], device='cuda:0'), covar=tensor([0.1295, 0.2554, 0.0387, 0.0355, 0.1003, 0.1844, 0.2305, 0.3436], device='cuda:0'), in_proj_covar=tensor([0.0271, 0.0369, 0.0261, 0.0285, 0.0250, 0.0278, 0.0346, 0.0349], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-15 23:41:41,791 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=140425.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 23:41:43,074 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.199e+02 2.942e+02 3.360e+02 4.212e+02 8.800e+02, threshold=6.720e+02, percent-clipped=3.0 2023-05-15 23:41:56,605 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=140445.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 23:41:58,618 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=140448.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 23:42:01,166 INFO [finetune.py:992] (0/2) Epoch 4, batch 3750, loss[loss=0.1862, simple_loss=0.2747, pruned_loss=0.04886, over 12092.00 frames. ], tot_loss[loss=0.1748, simple_loss=0.2623, pruned_loss=0.04368, over 2367113.09 frames. ], batch size: 33, lr: 4.81e-03, grad_scale: 16.0 2023-05-15 23:42:06,194 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=140459.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 23:42:07,064 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=140460.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 23:42:08,484 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=140462.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 23:42:10,951 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=3.46 vs. limit=5.0 2023-05-15 23:42:11,323 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=140466.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 23:42:16,200 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=140473.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 23:42:23,937 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.9304, 5.9158, 5.6564, 5.1749, 5.0698, 5.8114, 5.3855, 5.2208], device='cuda:0'), covar=tensor([0.0676, 0.0822, 0.0655, 0.1445, 0.0685, 0.0745, 0.1470, 0.1073], device='cuda:0'), in_proj_covar=tensor([0.0570, 0.0514, 0.0483, 0.0590, 0.0381, 0.0662, 0.0721, 0.0538], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0004, 0.0004, 0.0003], device='cuda:0') 2023-05-15 23:42:27,481 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=140489.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 23:42:30,350 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=140493.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 23:42:36,574 INFO [finetune.py:992] (0/2) Epoch 4, batch 3800, loss[loss=0.1992, simple_loss=0.2807, pruned_loss=0.05888, over 8229.00 frames. ], tot_loss[loss=0.1762, simple_loss=0.2637, pruned_loss=0.04438, over 2360051.55 frames. ], batch size: 99, lr: 4.81e-03, grad_scale: 16.0 2023-05-15 23:42:39,602 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=140506.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 23:42:40,898 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=140508.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 23:42:41,764 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=140509.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 23:42:42,367 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=140510.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 23:42:54,127 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.535e+02 2.899e+02 3.467e+02 4.265e+02 8.161e+02, threshold=6.934e+02, percent-clipped=4.0 2023-05-15 23:43:11,466 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=140550.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 23:43:12,580 INFO [finetune.py:992] (0/2) Epoch 4, batch 3850, loss[loss=0.1687, simple_loss=0.2588, pruned_loss=0.03925, over 12099.00 frames. ], tot_loss[loss=0.1759, simple_loss=0.2634, pruned_loss=0.04418, over 2369536.66 frames. ], batch size: 33, lr: 4.81e-03, grad_scale: 16.0 2023-05-15 23:43:14,097 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=140554.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 23:43:14,896 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.9206, 4.5408, 4.6000, 4.8473, 4.4283, 4.8464, 4.6920, 2.6127], device='cuda:0'), covar=tensor([0.0078, 0.0061, 0.0089, 0.0056, 0.0057, 0.0086, 0.0068, 0.0644], device='cuda:0'), in_proj_covar=tensor([0.0063, 0.0071, 0.0075, 0.0067, 0.0056, 0.0085, 0.0072, 0.0088], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-15 23:43:25,440 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=140570.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 23:43:46,106 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.0752, 4.7390, 4.8234, 4.9951, 4.6514, 4.9836, 4.8561, 2.6169], device='cuda:0'), covar=tensor([0.0092, 0.0051, 0.0074, 0.0052, 0.0046, 0.0087, 0.0060, 0.0659], device='cuda:0'), in_proj_covar=tensor([0.0064, 0.0071, 0.0076, 0.0068, 0.0056, 0.0085, 0.0073, 0.0089], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-15 23:43:49,028 INFO [finetune.py:992] (0/2) Epoch 4, batch 3900, loss[loss=0.1864, simple_loss=0.277, pruned_loss=0.0479, over 12380.00 frames. ], tot_loss[loss=0.1758, simple_loss=0.2631, pruned_loss=0.04427, over 2365882.33 frames. ], batch size: 38, lr: 4.81e-03, grad_scale: 16.0 2023-05-15 23:43:49,188 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.3193, 4.8792, 5.3233, 4.6371, 4.9125, 4.6933, 5.3453, 4.9345], device='cuda:0'), covar=tensor([0.0224, 0.0319, 0.0226, 0.0238, 0.0323, 0.0295, 0.0194, 0.0280], device='cuda:0'), in_proj_covar=tensor([0.0231, 0.0239, 0.0257, 0.0232, 0.0231, 0.0232, 0.0211, 0.0194], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-15 23:43:58,487 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=140615.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 23:44:06,951 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.772e+02 2.830e+02 3.285e+02 3.854e+02 6.178e+02, threshold=6.570e+02, percent-clipped=0.0 2023-05-15 23:44:22,353 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.3703, 5.2154, 5.2866, 5.3615, 4.9814, 5.0649, 4.8313, 5.3204], device='cuda:0'), covar=tensor([0.0629, 0.0505, 0.0686, 0.0568, 0.1672, 0.1133, 0.0474, 0.0784], device='cuda:0'), in_proj_covar=tensor([0.0497, 0.0632, 0.0544, 0.0594, 0.0777, 0.0701, 0.0516, 0.0462], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0005, 0.0004, 0.0004, 0.0005, 0.0005, 0.0004, 0.0004], device='cuda:0') 2023-05-15 23:44:24,232 INFO [finetune.py:992] (0/2) Epoch 4, batch 3950, loss[loss=0.1819, simple_loss=0.2716, pruned_loss=0.04604, over 12305.00 frames. ], tot_loss[loss=0.1756, simple_loss=0.2628, pruned_loss=0.04417, over 2372655.19 frames. ], batch size: 33, lr: 4.81e-03, grad_scale: 16.0 2023-05-15 23:44:43,901 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.47 vs. limit=2.0 2023-05-15 23:45:00,299 INFO [finetune.py:992] (0/2) Epoch 4, batch 4000, loss[loss=0.2065, simple_loss=0.2949, pruned_loss=0.05911, over 12020.00 frames. ], tot_loss[loss=0.1758, simple_loss=0.2635, pruned_loss=0.04405, over 2375252.41 frames. ], batch size: 40, lr: 4.81e-03, grad_scale: 16.0 2023-05-15 23:45:15,645 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=140723.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 23:45:18,331 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.962e+02 2.843e+02 3.354e+02 4.129e+02 8.480e+02, threshold=6.707e+02, percent-clipped=3.0 2023-05-15 23:45:29,859 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=140743.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 23:45:32,137 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.6944, 2.6335, 3.4039, 4.5273, 2.6771, 4.6839, 4.6760, 4.7155], device='cuda:0'), covar=tensor([0.0102, 0.1090, 0.0419, 0.0099, 0.1130, 0.0150, 0.0115, 0.0101], device='cuda:0'), in_proj_covar=tensor([0.0118, 0.0199, 0.0185, 0.0113, 0.0186, 0.0174, 0.0167, 0.0118], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-15 23:45:36,189 INFO [finetune.py:992] (0/2) Epoch 4, batch 4050, loss[loss=0.1875, simple_loss=0.2679, pruned_loss=0.05352, over 12365.00 frames. ], tot_loss[loss=0.1752, simple_loss=0.2627, pruned_loss=0.04391, over 2371746.79 frames. ], batch size: 31, lr: 4.81e-03, grad_scale: 16.0 2023-05-15 23:45:41,444 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=140759.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 23:45:46,386 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=140766.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 23:45:57,871 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.0498, 3.9099, 3.9636, 4.2882, 2.9799, 3.7266, 2.5658, 3.8790], device='cuda:0'), covar=tensor([0.1798, 0.0729, 0.0974, 0.0632, 0.1040, 0.0688, 0.1783, 0.1274], device='cuda:0'), in_proj_covar=tensor([0.0223, 0.0259, 0.0291, 0.0346, 0.0234, 0.0234, 0.0253, 0.0381], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-15 23:45:59,177 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=140784.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 23:46:12,229 INFO [finetune.py:992] (0/2) Epoch 4, batch 4100, loss[loss=0.1562, simple_loss=0.2394, pruned_loss=0.03649, over 12002.00 frames. ], tot_loss[loss=0.1754, simple_loss=0.2629, pruned_loss=0.04395, over 2376360.51 frames. ], batch size: 28, lr: 4.81e-03, grad_scale: 16.0 2023-05-15 23:46:15,831 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=140807.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 23:46:17,290 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=140809.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 23:46:20,609 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=140814.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 23:46:29,833 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.948e+02 2.854e+02 3.326e+02 3.979e+02 7.859e+02, threshold=6.651e+02, percent-clipped=1.0 2023-05-15 23:46:43,322 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=140845.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 23:46:48,261 INFO [finetune.py:992] (0/2) Epoch 4, batch 4150, loss[loss=0.1718, simple_loss=0.2575, pruned_loss=0.04312, over 12096.00 frames. ], tot_loss[loss=0.1762, simple_loss=0.2633, pruned_loss=0.04452, over 2361924.27 frames. ], batch size: 32, lr: 4.81e-03, grad_scale: 16.0 2023-05-15 23:46:58,278 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=140865.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 23:47:01,446 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.90 vs. limit=2.0 2023-05-15 23:47:01,828 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=140870.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 23:47:09,578 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.4383, 5.3160, 5.3600, 5.4255, 5.0132, 5.1284, 4.8929, 5.4221], device='cuda:0'), covar=tensor([0.0830, 0.0617, 0.0821, 0.0669, 0.2179, 0.1368, 0.0554, 0.0944], device='cuda:0'), in_proj_covar=tensor([0.0498, 0.0633, 0.0549, 0.0596, 0.0778, 0.0700, 0.0516, 0.0462], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0005, 0.0004, 0.0004, 0.0005, 0.0005, 0.0004, 0.0004], device='cuda:0') 2023-05-15 23:47:18,908 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.0818, 2.2141, 3.6429, 3.0327, 3.4303, 3.1283, 2.5277, 3.5745], device='cuda:0'), covar=tensor([0.0112, 0.0357, 0.0115, 0.0211, 0.0128, 0.0160, 0.0311, 0.0094], device='cuda:0'), in_proj_covar=tensor([0.0160, 0.0191, 0.0169, 0.0173, 0.0195, 0.0149, 0.0184, 0.0163], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-15 23:47:23,050 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.8198, 3.9745, 3.5677, 4.2775, 3.9106, 2.6445, 3.7749, 2.9204], device='cuda:0'), covar=tensor([0.0739, 0.0890, 0.1407, 0.0464, 0.1104, 0.1553, 0.0892, 0.2766], device='cuda:0'), in_proj_covar=tensor([0.0292, 0.0365, 0.0343, 0.0254, 0.0352, 0.0256, 0.0330, 0.0353], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-15 23:47:24,242 INFO [finetune.py:992] (0/2) Epoch 4, batch 4200, loss[loss=0.1767, simple_loss=0.2661, pruned_loss=0.04367, over 12132.00 frames. ], tot_loss[loss=0.1761, simple_loss=0.2631, pruned_loss=0.04455, over 2361277.66 frames. ], batch size: 38, lr: 4.81e-03, grad_scale: 16.0 2023-05-15 23:47:33,484 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=140915.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 23:47:34,860 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=140917.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 23:47:40,698 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2023-05-15 23:47:41,777 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.944e+02 2.886e+02 3.319e+02 3.814e+02 7.969e+02, threshold=6.638e+02, percent-clipped=2.0 2023-05-15 23:47:50,774 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.70 vs. limit=2.0 2023-05-15 23:47:56,307 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=140947.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 23:47:59,688 INFO [finetune.py:992] (0/2) Epoch 4, batch 4250, loss[loss=0.1922, simple_loss=0.2751, pruned_loss=0.05466, over 12121.00 frames. ], tot_loss[loss=0.1762, simple_loss=0.2633, pruned_loss=0.04451, over 2364257.10 frames. ], batch size: 38, lr: 4.81e-03, grad_scale: 16.0 2023-05-15 23:48:07,677 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=140963.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 23:48:18,480 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=140978.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 23:48:36,347 INFO [finetune.py:992] (0/2) Epoch 4, batch 4300, loss[loss=0.2157, simple_loss=0.2834, pruned_loss=0.07394, over 7649.00 frames. ], tot_loss[loss=0.1759, simple_loss=0.2633, pruned_loss=0.04428, over 2365052.05 frames. ], batch size: 99, lr: 4.81e-03, grad_scale: 16.0 2023-05-15 23:48:40,886 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=141008.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 23:48:54,649 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.94 vs. limit=2.0 2023-05-15 23:48:54,924 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.816e+02 2.741e+02 3.149e+02 3.897e+02 7.477e+02, threshold=6.298e+02, percent-clipped=1.0 2023-05-15 23:49:06,355 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=141043.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 23:49:11,314 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([6.0389, 5.9914, 5.7655, 5.3894, 5.0738, 5.9076, 5.5552, 5.3553], device='cuda:0'), covar=tensor([0.0736, 0.0900, 0.0665, 0.1484, 0.0695, 0.0732, 0.1410, 0.0976], device='cuda:0'), in_proj_covar=tensor([0.0572, 0.0508, 0.0482, 0.0588, 0.0380, 0.0662, 0.0722, 0.0534], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0004, 0.0004, 0.0003], device='cuda:0') 2023-05-15 23:49:12,618 INFO [finetune.py:992] (0/2) Epoch 4, batch 4350, loss[loss=0.1696, simple_loss=0.2485, pruned_loss=0.04536, over 11379.00 frames. ], tot_loss[loss=0.1758, simple_loss=0.2631, pruned_loss=0.04422, over 2363055.09 frames. ], batch size: 25, lr: 4.81e-03, grad_scale: 16.0 2023-05-15 23:49:31,768 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=141079.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 23:49:40,180 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=141091.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 23:49:47,698 INFO [finetune.py:992] (0/2) Epoch 4, batch 4400, loss[loss=0.165, simple_loss=0.2426, pruned_loss=0.04369, over 12249.00 frames. ], tot_loss[loss=0.177, simple_loss=0.2643, pruned_loss=0.04484, over 2359609.13 frames. ], batch size: 32, lr: 4.81e-03, grad_scale: 16.0 2023-05-15 23:50:05,756 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.783e+02 3.044e+02 3.595e+02 4.304e+02 1.021e+03, threshold=7.190e+02, percent-clipped=4.0 2023-05-15 23:50:17,182 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=141143.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 23:50:17,822 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.1916, 5.9909, 5.5218, 5.4822, 6.0577, 5.4457, 5.6741, 5.6146], device='cuda:0'), covar=tensor([0.1135, 0.0817, 0.0798, 0.1986, 0.0832, 0.1776, 0.1500, 0.0897], device='cuda:0'), in_proj_covar=tensor([0.0334, 0.0465, 0.0361, 0.0420, 0.0439, 0.0422, 0.0373, 0.0361], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-15 23:50:18,568 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=141145.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 23:50:24,146 INFO [finetune.py:992] (0/2) Epoch 4, batch 4450, loss[loss=0.1803, simple_loss=0.2715, pruned_loss=0.04455, over 11838.00 frames. ], tot_loss[loss=0.1774, simple_loss=0.2651, pruned_loss=0.04487, over 2350854.61 frames. ], batch size: 44, lr: 4.81e-03, grad_scale: 16.0 2023-05-15 23:50:33,431 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=141165.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 23:50:33,536 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=141165.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 23:50:53,127 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=141193.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 23:50:59,540 INFO [finetune.py:992] (0/2) Epoch 4, batch 4500, loss[loss=0.1795, simple_loss=0.2734, pruned_loss=0.04287, over 12369.00 frames. ], tot_loss[loss=0.1773, simple_loss=0.2643, pruned_loss=0.0452, over 2356706.71 frames. ], batch size: 35, lr: 4.81e-03, grad_scale: 16.0 2023-05-15 23:51:01,123 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=141204.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 23:51:07,376 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=141213.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 23:51:17,115 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.837e+02 2.809e+02 3.455e+02 4.408e+02 1.258e+03, threshold=6.910e+02, percent-clipped=4.0 2023-05-15 23:51:34,778 INFO [finetune.py:992] (0/2) Epoch 4, batch 4550, loss[loss=0.1985, simple_loss=0.2836, pruned_loss=0.05665, over 11261.00 frames. ], tot_loss[loss=0.1774, simple_loss=0.2642, pruned_loss=0.04524, over 2362898.06 frames. ], batch size: 55, lr: 4.81e-03, grad_scale: 16.0 2023-05-15 23:51:50,397 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=141273.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 23:52:11,294 INFO [finetune.py:992] (0/2) Epoch 4, batch 4600, loss[loss=0.157, simple_loss=0.2367, pruned_loss=0.03861, over 12335.00 frames. ], tot_loss[loss=0.1774, simple_loss=0.2642, pruned_loss=0.04529, over 2373148.34 frames. ], batch size: 30, lr: 4.81e-03, grad_scale: 16.0 2023-05-15 23:52:12,004 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=141303.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 23:52:28,860 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.677e+02 2.669e+02 3.274e+02 3.911e+02 6.807e+02, threshold=6.547e+02, percent-clipped=0.0 2023-05-15 23:52:30,406 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=141329.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 23:52:46,735 INFO [finetune.py:992] (0/2) Epoch 4, batch 4650, loss[loss=0.141, simple_loss=0.2208, pruned_loss=0.0306, over 12353.00 frames. ], tot_loss[loss=0.1773, simple_loss=0.2643, pruned_loss=0.04517, over 2372690.74 frames. ], batch size: 30, lr: 4.81e-03, grad_scale: 16.0 2023-05-15 23:53:05,791 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=141379.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 23:53:13,694 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=141390.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 23:53:16,640 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.3008, 4.3334, 4.1290, 4.4627, 3.0688, 3.9643, 2.5048, 4.2060], device='cuda:0'), covar=tensor([0.1617, 0.0524, 0.0881, 0.0600, 0.1041, 0.0571, 0.1791, 0.1167], device='cuda:0'), in_proj_covar=tensor([0.0228, 0.0264, 0.0296, 0.0353, 0.0239, 0.0237, 0.0257, 0.0389], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-15 23:53:22,573 INFO [finetune.py:992] (0/2) Epoch 4, batch 4700, loss[loss=0.183, simple_loss=0.2737, pruned_loss=0.04618, over 12141.00 frames. ], tot_loss[loss=0.1769, simple_loss=0.2636, pruned_loss=0.04506, over 2368104.45 frames. ], batch size: 38, lr: 4.81e-03, grad_scale: 16.0 2023-05-15 23:53:26,313 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.2799, 5.1289, 5.2298, 5.2869, 4.8873, 4.9733, 4.8013, 5.2194], device='cuda:0'), covar=tensor([0.0720, 0.0529, 0.0633, 0.0594, 0.1810, 0.1233, 0.0462, 0.0919], device='cuda:0'), in_proj_covar=tensor([0.0494, 0.0635, 0.0549, 0.0588, 0.0776, 0.0700, 0.0512, 0.0461], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0005, 0.0004, 0.0004, 0.0005, 0.0005, 0.0004, 0.0004], device='cuda:0') 2023-05-15 23:53:33,512 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=141417.0, num_to_drop=1, layers_to_drop={1} 2023-05-15 23:53:41,004 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.022e+02 2.952e+02 3.312e+02 4.023e+02 9.384e+02, threshold=6.625e+02, percent-clipped=1.0 2023-05-15 23:53:41,093 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=141427.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 23:53:48,939 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.0243, 4.8839, 4.9320, 5.0151, 4.6325, 4.6905, 4.5994, 4.9862], device='cuda:0'), covar=tensor([0.0720, 0.0562, 0.0765, 0.0541, 0.1759, 0.1269, 0.0473, 0.0812], device='cuda:0'), in_proj_covar=tensor([0.0493, 0.0632, 0.0547, 0.0585, 0.0774, 0.0697, 0.0510, 0.0458], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0005, 0.0004, 0.0004, 0.0005, 0.0005, 0.0004, 0.0004], device='cuda:0') 2023-05-15 23:53:51,075 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=141441.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 23:53:58,170 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.9481, 5.6865, 5.3336, 5.2523, 5.7955, 5.0803, 5.4297, 5.2779], device='cuda:0'), covar=tensor([0.1438, 0.0918, 0.0772, 0.2025, 0.0971, 0.2230, 0.1403, 0.1091], device='cuda:0'), in_proj_covar=tensor([0.0335, 0.0465, 0.0361, 0.0415, 0.0443, 0.0424, 0.0372, 0.0363], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-15 23:53:59,440 INFO [finetune.py:992] (0/2) Epoch 4, batch 4750, loss[loss=0.1829, simple_loss=0.2707, pruned_loss=0.04751, over 12162.00 frames. ], tot_loss[loss=0.1776, simple_loss=0.2647, pruned_loss=0.04523, over 2367787.27 frames. ], batch size: 34, lr: 4.80e-03, grad_scale: 16.0 2023-05-15 23:54:08,714 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=141465.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 23:54:17,843 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=141478.0, num_to_drop=1, layers_to_drop={2} 2023-05-15 23:54:32,533 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=141499.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 23:54:34,602 INFO [finetune.py:992] (0/2) Epoch 4, batch 4800, loss[loss=0.1774, simple_loss=0.2676, pruned_loss=0.04359, over 11582.00 frames. ], tot_loss[loss=0.1777, simple_loss=0.2649, pruned_loss=0.04522, over 2372540.94 frames. ], batch size: 48, lr: 4.80e-03, grad_scale: 16.0 2023-05-15 23:54:34,801 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=141502.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 23:54:42,275 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=141513.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 23:54:52,288 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.822e+02 2.852e+02 3.329e+02 3.958e+02 9.403e+02, threshold=6.658e+02, percent-clipped=2.0 2023-05-15 23:55:02,135 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.42 vs. limit=2.0 2023-05-15 23:55:09,976 INFO [finetune.py:992] (0/2) Epoch 4, batch 4850, loss[loss=0.1933, simple_loss=0.2875, pruned_loss=0.04956, over 12369.00 frames. ], tot_loss[loss=0.1772, simple_loss=0.2643, pruned_loss=0.04501, over 2371032.31 frames. ], batch size: 38, lr: 4.80e-03, grad_scale: 16.0 2023-05-15 23:55:25,134 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=141573.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 23:55:46,765 INFO [finetune.py:992] (0/2) Epoch 4, batch 4900, loss[loss=0.1808, simple_loss=0.281, pruned_loss=0.04026, over 12035.00 frames. ], tot_loss[loss=0.1768, simple_loss=0.2639, pruned_loss=0.04487, over 2363991.79 frames. ], batch size: 40, lr: 4.80e-03, grad_scale: 16.0 2023-05-15 23:55:47,596 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=141603.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 23:55:55,835 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.87 vs. limit=2.0 2023-05-15 23:56:00,189 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=141621.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 23:56:03,246 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.7731, 2.9869, 4.7676, 4.9738, 2.9288, 2.7566, 3.0096, 2.2851], device='cuda:0'), covar=tensor([0.1305, 0.2815, 0.0364, 0.0356, 0.1132, 0.1955, 0.2419, 0.3377], device='cuda:0'), in_proj_covar=tensor([0.0269, 0.0366, 0.0260, 0.0284, 0.0248, 0.0275, 0.0344, 0.0347], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-15 23:56:04,352 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.752e+02 2.647e+02 3.128e+02 3.697e+02 6.289e+02, threshold=6.257e+02, percent-clipped=0.0 2023-05-15 23:56:21,177 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=4.75 vs. limit=5.0 2023-05-15 23:56:21,357 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=141651.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 23:56:22,018 INFO [finetune.py:992] (0/2) Epoch 4, batch 4950, loss[loss=0.1763, simple_loss=0.2591, pruned_loss=0.04669, over 12086.00 frames. ], tot_loss[loss=0.177, simple_loss=0.2642, pruned_loss=0.04495, over 2367319.86 frames. ], batch size: 32, lr: 4.80e-03, grad_scale: 16.0 2023-05-15 23:56:45,263 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=141685.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 23:56:57,378 INFO [finetune.py:992] (0/2) Epoch 4, batch 5000, loss[loss=0.1362, simple_loss=0.2201, pruned_loss=0.02618, over 12312.00 frames. ], tot_loss[loss=0.1765, simple_loss=0.2634, pruned_loss=0.0448, over 2365671.25 frames. ], batch size: 30, lr: 4.80e-03, grad_scale: 16.0 2023-05-15 23:57:15,458 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.743e+02 2.970e+02 3.491e+02 4.383e+02 8.620e+02, threshold=6.983e+02, percent-clipped=4.0 2023-05-15 23:57:33,775 INFO [finetune.py:992] (0/2) Epoch 4, batch 5050, loss[loss=0.1605, simple_loss=0.2431, pruned_loss=0.03897, over 12249.00 frames. ], tot_loss[loss=0.1752, simple_loss=0.2622, pruned_loss=0.04411, over 2375570.72 frames. ], batch size: 32, lr: 4.80e-03, grad_scale: 16.0 2023-05-15 23:57:48,592 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=141773.0, num_to_drop=1, layers_to_drop={0} 2023-05-15 23:58:05,508 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=141797.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 23:58:06,923 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=141799.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 23:58:09,325 INFO [finetune.py:992] (0/2) Epoch 4, batch 5100, loss[loss=0.204, simple_loss=0.2879, pruned_loss=0.06008, over 12352.00 frames. ], tot_loss[loss=0.1752, simple_loss=0.2625, pruned_loss=0.04398, over 2370395.16 frames. ], batch size: 35, lr: 4.80e-03, grad_scale: 16.0 2023-05-15 23:58:27,076 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.257e+02 2.764e+02 3.422e+02 4.267e+02 1.010e+03, threshold=6.843e+02, percent-clipped=2.0 2023-05-15 23:58:40,770 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=141847.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 23:58:44,285 INFO [finetune.py:992] (0/2) Epoch 4, batch 5150, loss[loss=0.1724, simple_loss=0.2684, pruned_loss=0.03823, over 11984.00 frames. ], tot_loss[loss=0.176, simple_loss=0.2636, pruned_loss=0.04426, over 2364449.04 frames. ], batch size: 40, lr: 4.80e-03, grad_scale: 16.0 2023-05-15 23:59:21,245 INFO [finetune.py:992] (0/2) Epoch 4, batch 5200, loss[loss=0.191, simple_loss=0.284, pruned_loss=0.04897, over 12297.00 frames. ], tot_loss[loss=0.1756, simple_loss=0.2631, pruned_loss=0.04409, over 2370849.70 frames. ], batch size: 33, lr: 4.80e-03, grad_scale: 16.0 2023-05-15 23:59:31,092 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=141916.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 23:59:38,867 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.200e+02 2.841e+02 3.341e+02 4.056e+02 6.577e+02, threshold=6.683e+02, percent-clipped=0.0 2023-05-15 23:59:39,747 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=141928.0, num_to_drop=0, layers_to_drop=set() 2023-05-15 23:59:56,516 INFO [finetune.py:992] (0/2) Epoch 4, batch 5250, loss[loss=0.1876, simple_loss=0.2736, pruned_loss=0.05079, over 10371.00 frames. ], tot_loss[loss=0.1748, simple_loss=0.2624, pruned_loss=0.04365, over 2377495.00 frames. ], batch size: 68, lr: 4.80e-03, grad_scale: 16.0 2023-05-16 00:00:14,672 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=141977.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 00:00:20,286 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=141985.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 00:00:23,190 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=141989.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 00:00:31,061 INFO [checkpoint.py:75] (0/2) Saving checkpoint to pruned_transducer_stateless7/exp_giga_finetune/checkpoint-42000.pt 2023-05-16 00:00:35,452 INFO [finetune.py:992] (0/2) Epoch 4, batch 5300, loss[loss=0.1805, simple_loss=0.2666, pruned_loss=0.04717, over 12299.00 frames. ], tot_loss[loss=0.1745, simple_loss=0.262, pruned_loss=0.0435, over 2369574.67 frames. ], batch size: 33, lr: 4.80e-03, grad_scale: 16.0 2023-05-16 00:00:54,339 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.747e+02 2.901e+02 3.592e+02 4.199e+02 7.190e+02, threshold=7.183e+02, percent-clipped=1.0 2023-05-16 00:00:57,106 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.34 vs. limit=2.0 2023-05-16 00:00:58,587 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=142033.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 00:01:12,237 INFO [finetune.py:992] (0/2) Epoch 4, batch 5350, loss[loss=0.1677, simple_loss=0.2555, pruned_loss=0.03997, over 12296.00 frames. ], tot_loss[loss=0.1757, simple_loss=0.2629, pruned_loss=0.04426, over 2363022.15 frames. ], batch size: 33, lr: 4.80e-03, grad_scale: 16.0 2023-05-16 00:01:24,867 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.46 vs. limit=2.0 2023-05-16 00:01:27,258 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=142073.0, num_to_drop=1, layers_to_drop={1} 2023-05-16 00:01:44,300 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=142097.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 00:01:47,672 INFO [finetune.py:992] (0/2) Epoch 4, batch 5400, loss[loss=0.1815, simple_loss=0.2651, pruned_loss=0.04897, over 12127.00 frames. ], tot_loss[loss=0.1758, simple_loss=0.2633, pruned_loss=0.04413, over 2364810.76 frames. ], batch size: 39, lr: 4.80e-03, grad_scale: 16.0 2023-05-16 00:01:49,740 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.39 vs. limit=2.0 2023-05-16 00:02:01,161 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=142121.0, num_to_drop=1, layers_to_drop={1} 2023-05-16 00:02:05,100 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.087e+02 2.960e+02 3.619e+02 4.165e+02 8.260e+02, threshold=7.238e+02, percent-clipped=1.0 2023-05-16 00:02:17,853 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=142145.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 00:02:23,468 INFO [finetune.py:992] (0/2) Epoch 4, batch 5450, loss[loss=0.1555, simple_loss=0.2415, pruned_loss=0.03478, over 12103.00 frames. ], tot_loss[loss=0.1764, simple_loss=0.2638, pruned_loss=0.04446, over 2369135.02 frames. ], batch size: 32, lr: 4.80e-03, grad_scale: 16.0 2023-05-16 00:02:59,276 INFO [finetune.py:992] (0/2) Epoch 4, batch 5500, loss[loss=0.1778, simple_loss=0.2649, pruned_loss=0.04535, over 12259.00 frames. ], tot_loss[loss=0.1769, simple_loss=0.2645, pruned_loss=0.04466, over 2371869.13 frames. ], batch size: 32, lr: 4.80e-03, grad_scale: 16.0 2023-05-16 00:03:01,490 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.9064, 4.5324, 4.1089, 4.1097, 4.6380, 3.9907, 4.2549, 3.9705], device='cuda:0'), covar=tensor([0.1426, 0.1192, 0.1371, 0.2127, 0.1179, 0.2242, 0.1747, 0.1425], device='cuda:0'), in_proj_covar=tensor([0.0334, 0.0468, 0.0362, 0.0416, 0.0440, 0.0422, 0.0375, 0.0362], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 00:03:16,920 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.067e+02 2.754e+02 3.312e+02 4.084e+02 8.741e+02, threshold=6.624e+02, percent-clipped=1.0 2023-05-16 00:03:34,601 INFO [finetune.py:992] (0/2) Epoch 4, batch 5550, loss[loss=0.1667, simple_loss=0.264, pruned_loss=0.03468, over 12145.00 frames. ], tot_loss[loss=0.1778, simple_loss=0.2651, pruned_loss=0.04524, over 2358333.28 frames. ], batch size: 36, lr: 4.80e-03, grad_scale: 32.0 2023-05-16 00:03:48,783 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=142272.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 00:03:57,366 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=142284.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 00:04:10,779 INFO [finetune.py:992] (0/2) Epoch 4, batch 5600, loss[loss=0.1674, simple_loss=0.2569, pruned_loss=0.03894, over 12120.00 frames. ], tot_loss[loss=0.1784, simple_loss=0.2657, pruned_loss=0.04555, over 2358363.18 frames. ], batch size: 33, lr: 4.80e-03, grad_scale: 32.0 2023-05-16 00:04:29,386 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.880e+02 2.833e+02 3.192e+02 3.756e+02 1.144e+03, threshold=6.384e+02, percent-clipped=4.0 2023-05-16 00:04:47,039 INFO [finetune.py:992] (0/2) Epoch 4, batch 5650, loss[loss=0.1859, simple_loss=0.2768, pruned_loss=0.04753, over 11953.00 frames. ], tot_loss[loss=0.1791, simple_loss=0.2659, pruned_loss=0.0461, over 2353406.44 frames. ], batch size: 42, lr: 4.80e-03, grad_scale: 32.0 2023-05-16 00:04:53,612 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.9764, 3.1924, 5.1522, 2.8877, 2.9768, 3.9821, 3.3997, 3.9996], device='cuda:0'), covar=tensor([0.0357, 0.1311, 0.0427, 0.1082, 0.1843, 0.1355, 0.1239, 0.1004], device='cuda:0'), in_proj_covar=tensor([0.0230, 0.0228, 0.0231, 0.0179, 0.0232, 0.0279, 0.0223, 0.0252], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-16 00:05:01,095 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.8301, 5.8375, 5.6109, 5.1896, 4.9687, 5.7817, 5.3560, 5.1413], device='cuda:0'), covar=tensor([0.0723, 0.0776, 0.0589, 0.1440, 0.0721, 0.0720, 0.1538, 0.0992], device='cuda:0'), in_proj_covar=tensor([0.0569, 0.0503, 0.0476, 0.0586, 0.0378, 0.0657, 0.0713, 0.0532], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0004, 0.0004, 0.0003], device='cuda:0') 2023-05-16 00:05:22,318 INFO [finetune.py:992] (0/2) Epoch 4, batch 5700, loss[loss=0.1748, simple_loss=0.2704, pruned_loss=0.03964, over 12189.00 frames. ], tot_loss[loss=0.1793, simple_loss=0.2658, pruned_loss=0.04642, over 2357857.44 frames. ], batch size: 35, lr: 4.80e-03, grad_scale: 32.0 2023-05-16 00:05:39,847 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.963e+02 3.063e+02 3.770e+02 4.357e+02 1.028e+03, threshold=7.540e+02, percent-clipped=5.0 2023-05-16 00:05:43,625 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.0479, 2.5198, 3.5328, 3.0348, 3.3554, 3.1292, 2.4543, 3.4975], device='cuda:0'), covar=tensor([0.0113, 0.0284, 0.0115, 0.0232, 0.0136, 0.0168, 0.0345, 0.0107], device='cuda:0'), in_proj_covar=tensor([0.0160, 0.0191, 0.0170, 0.0172, 0.0195, 0.0149, 0.0184, 0.0165], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 00:05:52,834 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=2.33 vs. limit=5.0 2023-05-16 00:05:58,118 INFO [finetune.py:992] (0/2) Epoch 4, batch 5750, loss[loss=0.1584, simple_loss=0.239, pruned_loss=0.03891, over 12277.00 frames. ], tot_loss[loss=0.1778, simple_loss=0.2644, pruned_loss=0.04554, over 2366202.99 frames. ], batch size: 28, lr: 4.80e-03, grad_scale: 32.0 2023-05-16 00:06:31,681 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.8865, 3.3962, 5.1857, 2.6961, 2.8509, 3.8374, 3.4355, 3.9701], device='cuda:0'), covar=tensor([0.0409, 0.1068, 0.0220, 0.1148, 0.1730, 0.1433, 0.1193, 0.0960], device='cuda:0'), in_proj_covar=tensor([0.0231, 0.0229, 0.0231, 0.0179, 0.0233, 0.0281, 0.0223, 0.0252], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-16 00:06:33,546 INFO [finetune.py:992] (0/2) Epoch 4, batch 5800, loss[loss=0.1885, simple_loss=0.2899, pruned_loss=0.04349, over 12358.00 frames. ], tot_loss[loss=0.178, simple_loss=0.2648, pruned_loss=0.04564, over 2368115.27 frames. ], batch size: 36, lr: 4.80e-03, grad_scale: 32.0 2023-05-16 00:06:49,074 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.98 vs. limit=2.0 2023-05-16 00:06:51,332 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.283e+02 2.997e+02 3.428e+02 4.047e+02 7.003e+02, threshold=6.857e+02, percent-clipped=0.0 2023-05-16 00:07:01,738 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-05-16 00:07:08,938 INFO [finetune.py:992] (0/2) Epoch 4, batch 5850, loss[loss=0.2204, simple_loss=0.3045, pruned_loss=0.06817, over 10432.00 frames. ], tot_loss[loss=0.1769, simple_loss=0.2636, pruned_loss=0.04508, over 2370161.68 frames. ], batch size: 68, lr: 4.80e-03, grad_scale: 32.0 2023-05-16 00:07:23,352 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=142572.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 00:07:24,076 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=142573.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 00:07:31,922 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=142584.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 00:07:38,334 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.6579, 2.9526, 4.6041, 4.8332, 2.9924, 2.7209, 2.8717, 2.1271], device='cuda:0'), covar=tensor([0.1483, 0.2737, 0.0426, 0.0337, 0.1177, 0.2051, 0.2480, 0.3736], device='cuda:0'), in_proj_covar=tensor([0.0273, 0.0373, 0.0263, 0.0289, 0.0252, 0.0280, 0.0349, 0.0352], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-16 00:07:41,852 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.1568, 2.6050, 3.7786, 3.1641, 3.5146, 3.2242, 2.5435, 3.6947], device='cuda:0'), covar=tensor([0.0102, 0.0278, 0.0103, 0.0198, 0.0100, 0.0154, 0.0302, 0.0077], device='cuda:0'), in_proj_covar=tensor([0.0162, 0.0194, 0.0173, 0.0175, 0.0198, 0.0151, 0.0188, 0.0167], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 00:07:45,479 INFO [finetune.py:992] (0/2) Epoch 4, batch 5900, loss[loss=0.1739, simple_loss=0.2673, pruned_loss=0.04022, over 12149.00 frames. ], tot_loss[loss=0.1766, simple_loss=0.2636, pruned_loss=0.04477, over 2373352.94 frames. ], batch size: 36, lr: 4.79e-03, grad_scale: 16.0 2023-05-16 00:07:51,620 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 2023-05-16 00:07:58,823 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=142620.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 00:08:04,596 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.967e+02 2.871e+02 3.403e+02 4.114e+02 6.329e+02, threshold=6.806e+02, percent-clipped=0.0 2023-05-16 00:08:07,555 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=142632.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 00:08:09,046 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=142634.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 00:08:21,689 INFO [finetune.py:992] (0/2) Epoch 4, batch 5950, loss[loss=0.15, simple_loss=0.2257, pruned_loss=0.03719, over 12008.00 frames. ], tot_loss[loss=0.1765, simple_loss=0.2636, pruned_loss=0.04469, over 2372118.30 frames. ], batch size: 28, lr: 4.79e-03, grad_scale: 16.0 2023-05-16 00:08:29,744 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.2915, 2.6263, 3.7847, 3.2982, 3.5496, 3.3768, 2.6448, 3.7067], device='cuda:0'), covar=tensor([0.0127, 0.0288, 0.0126, 0.0191, 0.0128, 0.0144, 0.0323, 0.0108], device='cuda:0'), in_proj_covar=tensor([0.0161, 0.0192, 0.0171, 0.0174, 0.0196, 0.0150, 0.0186, 0.0165], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 00:08:57,080 INFO [finetune.py:992] (0/2) Epoch 4, batch 6000, loss[loss=0.1975, simple_loss=0.2841, pruned_loss=0.0555, over 12408.00 frames. ], tot_loss[loss=0.1773, simple_loss=0.2647, pruned_loss=0.04492, over 2368978.15 frames. ], batch size: 32, lr: 4.79e-03, grad_scale: 16.0 2023-05-16 00:08:57,081 INFO [finetune.py:1017] (0/2) Computing validation loss 2023-05-16 00:09:02,075 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.6039, 3.2147, 2.2394, 1.8173, 2.9951, 2.1094, 3.1560, 2.3021], device='cuda:0'), covar=tensor([0.0595, 0.0563, 0.1009, 0.1656, 0.0258, 0.1216, 0.0401, 0.0865], device='cuda:0'), in_proj_covar=tensor([0.0182, 0.0243, 0.0170, 0.0194, 0.0136, 0.0175, 0.0188, 0.0167], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 00:09:13,091 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.0691, 1.7587, 2.0915, 1.9115, 2.0479, 2.2068, 1.5615, 2.0712], device='cuda:0'), covar=tensor([0.0084, 0.0245, 0.0095, 0.0168, 0.0117, 0.0113, 0.0239, 0.0095], device='cuda:0'), in_proj_covar=tensor([0.0162, 0.0193, 0.0172, 0.0174, 0.0197, 0.0151, 0.0187, 0.0166], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 00:09:15,291 INFO [finetune.py:1026] (0/2) Epoch 4, validation: loss=0.3275, simple_loss=0.4012, pruned_loss=0.1269, over 1020973.00 frames. 2023-05-16 00:09:15,292 INFO [finetune.py:1027] (0/2) Maximum memory allocated so far is 12745MB 2023-05-16 00:09:23,930 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.3335, 5.1394, 5.2695, 5.2887, 4.8871, 4.9841, 4.8083, 5.2302], device='cuda:0'), covar=tensor([0.0540, 0.0521, 0.0680, 0.0497, 0.1886, 0.1140, 0.0438, 0.0954], device='cuda:0'), in_proj_covar=tensor([0.0487, 0.0630, 0.0544, 0.0585, 0.0775, 0.0698, 0.0513, 0.0458], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0005, 0.0004, 0.0004, 0.0005, 0.0005, 0.0004, 0.0004], device='cuda:0') 2023-05-16 00:09:30,961 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=142724.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 00:09:33,671 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.236e+02 2.858e+02 3.331e+02 4.161e+02 8.424e+02, threshold=6.661e+02, percent-clipped=1.0 2023-05-16 00:09:40,269 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.5498, 4.7330, 4.3270, 5.2698, 4.8343, 2.6405, 4.3842, 3.1873], device='cuda:0'), covar=tensor([0.0580, 0.0722, 0.1123, 0.0330, 0.0772, 0.1562, 0.0910, 0.2741], device='cuda:0'), in_proj_covar=tensor([0.0293, 0.0367, 0.0346, 0.0259, 0.0354, 0.0259, 0.0332, 0.0358], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 00:09:49,757 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.3090, 4.6862, 2.7120, 1.9141, 4.0636, 2.0568, 3.8760, 2.9338], device='cuda:0'), covar=tensor([0.0552, 0.0459, 0.1266, 0.2338, 0.0317, 0.1845, 0.0510, 0.1004], device='cuda:0'), in_proj_covar=tensor([0.0180, 0.0242, 0.0168, 0.0193, 0.0135, 0.0174, 0.0186, 0.0166], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 00:09:50,257 INFO [finetune.py:992] (0/2) Epoch 4, batch 6050, loss[loss=0.1765, simple_loss=0.2757, pruned_loss=0.03861, over 12203.00 frames. ], tot_loss[loss=0.1781, simple_loss=0.2656, pruned_loss=0.04533, over 2363351.78 frames. ], batch size: 35, lr: 4.79e-03, grad_scale: 16.0 2023-05-16 00:10:08,859 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.3979, 4.8145, 4.1371, 5.1662, 4.6277, 3.0146, 4.4250, 3.1737], device='cuda:0'), covar=tensor([0.0685, 0.0579, 0.1303, 0.0311, 0.0987, 0.1397, 0.0812, 0.2839], device='cuda:0'), in_proj_covar=tensor([0.0292, 0.0364, 0.0344, 0.0258, 0.0353, 0.0258, 0.0331, 0.0356], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 00:10:13,754 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=142785.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 00:10:22,647 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.69 vs. limit=2.0 2023-05-16 00:10:26,120 INFO [finetune.py:992] (0/2) Epoch 4, batch 6100, loss[loss=0.1508, simple_loss=0.2352, pruned_loss=0.03316, over 12125.00 frames. ], tot_loss[loss=0.1781, simple_loss=0.2653, pruned_loss=0.04546, over 2363464.16 frames. ], batch size: 30, lr: 4.79e-03, grad_scale: 16.0 2023-05-16 00:10:27,284 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2023-05-16 00:10:44,522 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.973e+02 2.941e+02 3.348e+02 3.892e+02 7.445e+02, threshold=6.695e+02, percent-clipped=1.0 2023-05-16 00:10:51,418 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.5055, 3.7805, 3.3425, 3.2797, 3.0335, 2.8301, 3.5418, 2.4019], device='cuda:0'), covar=tensor([0.0332, 0.0080, 0.0141, 0.0150, 0.0295, 0.0281, 0.0119, 0.0394], device='cuda:0'), in_proj_covar=tensor([0.0183, 0.0157, 0.0151, 0.0175, 0.0195, 0.0191, 0.0156, 0.0196], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 00:11:02,351 INFO [finetune.py:992] (0/2) Epoch 4, batch 6150, loss[loss=0.1409, simple_loss=0.2237, pruned_loss=0.02908, over 12283.00 frames. ], tot_loss[loss=0.1769, simple_loss=0.2643, pruned_loss=0.04477, over 2375061.92 frames. ], batch size: 28, lr: 4.79e-03, grad_scale: 16.0 2023-05-16 00:11:18,002 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.2851, 2.6616, 3.8126, 3.2943, 3.5842, 3.4297, 2.7510, 3.7145], device='cuda:0'), covar=tensor([0.0124, 0.0289, 0.0130, 0.0189, 0.0149, 0.0131, 0.0291, 0.0096], device='cuda:0'), in_proj_covar=tensor([0.0163, 0.0193, 0.0171, 0.0174, 0.0197, 0.0150, 0.0186, 0.0166], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 00:11:38,069 INFO [finetune.py:992] (0/2) Epoch 4, batch 6200, loss[loss=0.1738, simple_loss=0.271, pruned_loss=0.03833, over 11576.00 frames. ], tot_loss[loss=0.1769, simple_loss=0.2645, pruned_loss=0.04463, over 2378031.88 frames. ], batch size: 48, lr: 4.79e-03, grad_scale: 16.0 2023-05-16 00:11:56,735 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.973e+02 2.738e+02 3.128e+02 3.851e+02 9.110e+02, threshold=6.257e+02, percent-clipped=2.0 2023-05-16 00:11:57,581 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=142929.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 00:12:01,431 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.1889, 4.0176, 4.1558, 4.3978, 2.9519, 3.9524, 2.6495, 4.0166], device='cuda:0'), covar=tensor([0.1611, 0.0678, 0.0798, 0.0620, 0.1095, 0.0576, 0.1691, 0.1329], device='cuda:0'), in_proj_covar=tensor([0.0227, 0.0263, 0.0294, 0.0349, 0.0239, 0.0238, 0.0257, 0.0383], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 00:12:14,012 INFO [finetune.py:992] (0/2) Epoch 4, batch 6250, loss[loss=0.1897, simple_loss=0.2816, pruned_loss=0.04888, over 12275.00 frames. ], tot_loss[loss=0.1766, simple_loss=0.2642, pruned_loss=0.04447, over 2382883.62 frames. ], batch size: 37, lr: 4.79e-03, grad_scale: 16.0 2023-05-16 00:12:16,456 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.64 vs. limit=2.0 2023-05-16 00:12:49,953 INFO [finetune.py:992] (0/2) Epoch 4, batch 6300, loss[loss=0.1601, simple_loss=0.2588, pruned_loss=0.03074, over 12307.00 frames. ], tot_loss[loss=0.1757, simple_loss=0.2631, pruned_loss=0.04416, over 2386443.99 frames. ], batch size: 34, lr: 4.79e-03, grad_scale: 16.0 2023-05-16 00:13:08,924 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.904e+02 2.720e+02 3.349e+02 3.998e+02 8.545e+02, threshold=6.697e+02, percent-clipped=2.0 2023-05-16 00:13:23,127 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.48 vs. limit=2.0 2023-05-16 00:13:26,002 INFO [finetune.py:992] (0/2) Epoch 4, batch 6350, loss[loss=0.188, simple_loss=0.2791, pruned_loss=0.04847, over 12163.00 frames. ], tot_loss[loss=0.1763, simple_loss=0.2635, pruned_loss=0.04458, over 2381115.84 frames. ], batch size: 36, lr: 4.79e-03, grad_scale: 16.0 2023-05-16 00:13:45,782 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=143080.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 00:13:47,473 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.6107, 2.8624, 4.6132, 4.8249, 2.9856, 2.6528, 2.9371, 2.0437], device='cuda:0'), covar=tensor([0.1416, 0.2734, 0.0418, 0.0343, 0.1089, 0.1939, 0.2429, 0.3817], device='cuda:0'), in_proj_covar=tensor([0.0274, 0.0374, 0.0264, 0.0291, 0.0254, 0.0282, 0.0351, 0.0355], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-16 00:14:01,322 INFO [finetune.py:992] (0/2) Epoch 4, batch 6400, loss[loss=0.1797, simple_loss=0.2603, pruned_loss=0.04953, over 12182.00 frames. ], tot_loss[loss=0.1776, simple_loss=0.2645, pruned_loss=0.04534, over 2374454.75 frames. ], batch size: 31, lr: 4.79e-03, grad_scale: 16.0 2023-05-16 00:14:19,844 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.967e+02 2.964e+02 3.528e+02 4.352e+02 1.519e+03, threshold=7.056e+02, percent-clipped=5.0 2023-05-16 00:14:38,291 INFO [finetune.py:992] (0/2) Epoch 4, batch 6450, loss[loss=0.151, simple_loss=0.2365, pruned_loss=0.03281, over 12186.00 frames. ], tot_loss[loss=0.1773, simple_loss=0.2645, pruned_loss=0.04509, over 2379527.03 frames. ], batch size: 29, lr: 4.79e-03, grad_scale: 16.0 2023-05-16 00:14:41,397 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.6077, 2.7086, 4.4385, 4.6457, 3.1560, 2.6362, 2.9062, 2.1859], device='cuda:0'), covar=tensor([0.1432, 0.3082, 0.0508, 0.0394, 0.1029, 0.2056, 0.2678, 0.3824], device='cuda:0'), in_proj_covar=tensor([0.0275, 0.0374, 0.0264, 0.0291, 0.0253, 0.0282, 0.0351, 0.0355], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-16 00:15:01,062 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.58 vs. limit=2.0 2023-05-16 00:15:14,520 INFO [finetune.py:992] (0/2) Epoch 4, batch 6500, loss[loss=0.1679, simple_loss=0.2493, pruned_loss=0.04331, over 12190.00 frames. ], tot_loss[loss=0.1781, simple_loss=0.2653, pruned_loss=0.04543, over 2378717.68 frames. ], batch size: 31, lr: 4.79e-03, grad_scale: 16.0 2023-05-16 00:15:32,396 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.4721, 3.6602, 3.2673, 3.2216, 2.9167, 2.6894, 3.5855, 2.3079], device='cuda:0'), covar=tensor([0.0315, 0.0118, 0.0159, 0.0144, 0.0327, 0.0309, 0.0105, 0.0380], device='cuda:0'), in_proj_covar=tensor([0.0186, 0.0160, 0.0152, 0.0177, 0.0198, 0.0193, 0.0158, 0.0197], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 00:15:32,843 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.055e+02 2.770e+02 3.120e+02 3.791e+02 9.865e+02, threshold=6.240e+02, percent-clipped=3.0 2023-05-16 00:15:33,700 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=143229.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 00:15:43,678 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.3516, 4.9421, 5.3447, 4.6495, 5.0001, 4.7268, 5.3674, 5.0096], device='cuda:0'), covar=tensor([0.0217, 0.0315, 0.0228, 0.0240, 0.0296, 0.0266, 0.0176, 0.0285], device='cuda:0'), in_proj_covar=tensor([0.0236, 0.0242, 0.0260, 0.0237, 0.0233, 0.0235, 0.0214, 0.0196], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-05-16 00:15:46,325 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=4.60 vs. limit=5.0 2023-05-16 00:15:49,996 INFO [finetune.py:992] (0/2) Epoch 4, batch 6550, loss[loss=0.1862, simple_loss=0.2685, pruned_loss=0.05189, over 12108.00 frames. ], tot_loss[loss=0.1785, simple_loss=0.266, pruned_loss=0.04551, over 2379234.90 frames. ], batch size: 42, lr: 4.79e-03, grad_scale: 16.0 2023-05-16 00:16:08,530 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=143277.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 00:16:26,977 INFO [finetune.py:992] (0/2) Epoch 4, batch 6600, loss[loss=0.1664, simple_loss=0.2633, pruned_loss=0.03473, over 12182.00 frames. ], tot_loss[loss=0.1774, simple_loss=0.2648, pruned_loss=0.04502, over 2378794.81 frames. ], batch size: 35, lr: 4.79e-03, grad_scale: 16.0 2023-05-16 00:16:45,588 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.893e+02 2.783e+02 3.345e+02 4.038e+02 6.670e+02, threshold=6.691e+02, percent-clipped=2.0 2023-05-16 00:16:45,862 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.7424, 2.5244, 3.5317, 3.6912, 2.8916, 2.6767, 2.6009, 2.3466], device='cuda:0'), covar=tensor([0.1069, 0.2572, 0.0568, 0.0448, 0.0841, 0.1705, 0.2218, 0.3195], device='cuda:0'), in_proj_covar=tensor([0.0274, 0.0373, 0.0263, 0.0290, 0.0252, 0.0281, 0.0351, 0.0354], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-16 00:17:02,636 INFO [finetune.py:992] (0/2) Epoch 4, batch 6650, loss[loss=0.1415, simple_loss=0.2293, pruned_loss=0.02682, over 11977.00 frames. ], tot_loss[loss=0.1785, simple_loss=0.2655, pruned_loss=0.04571, over 2360261.68 frames. ], batch size: 28, lr: 4.79e-03, grad_scale: 16.0 2023-05-16 00:17:22,847 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=143380.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 00:17:37,118 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=143400.0, num_to_drop=1, layers_to_drop={1} 2023-05-16 00:17:38,342 INFO [finetune.py:992] (0/2) Epoch 4, batch 6700, loss[loss=0.1596, simple_loss=0.242, pruned_loss=0.03855, over 12139.00 frames. ], tot_loss[loss=0.1781, simple_loss=0.265, pruned_loss=0.04559, over 2362752.87 frames. ], batch size: 30, lr: 4.79e-03, grad_scale: 8.0 2023-05-16 00:17:45,562 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.4873, 4.5034, 4.2636, 4.6592, 3.3716, 4.2430, 2.8519, 4.3771], device='cuda:0'), covar=tensor([0.1454, 0.0510, 0.0839, 0.0568, 0.0955, 0.0492, 0.1590, 0.1211], device='cuda:0'), in_proj_covar=tensor([0.0223, 0.0259, 0.0290, 0.0346, 0.0236, 0.0235, 0.0253, 0.0380], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 00:17:57,419 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=143428.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 00:17:57,977 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.715e+02 2.936e+02 3.309e+02 3.963e+02 6.958e+02, threshold=6.618e+02, percent-clipped=1.0 2023-05-16 00:18:13,768 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.8277, 4.6522, 4.6527, 4.7049, 4.5697, 4.8397, 4.6879, 2.7861], device='cuda:0'), covar=tensor([0.0142, 0.0068, 0.0100, 0.0071, 0.0058, 0.0094, 0.0072, 0.0663], device='cuda:0'), in_proj_covar=tensor([0.0065, 0.0073, 0.0077, 0.0071, 0.0057, 0.0088, 0.0075, 0.0091], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 00:18:14,953 INFO [finetune.py:992] (0/2) Epoch 4, batch 6750, loss[loss=0.1673, simple_loss=0.2625, pruned_loss=0.03607, over 12050.00 frames. ], tot_loss[loss=0.1781, simple_loss=0.2652, pruned_loss=0.04546, over 2365760.32 frames. ], batch size: 40, lr: 4.79e-03, grad_scale: 8.0 2023-05-16 00:18:21,619 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=143461.0, num_to_drop=1, layers_to_drop={3} 2023-05-16 00:18:30,126 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=143473.0, num_to_drop=1, layers_to_drop={1} 2023-05-16 00:18:50,462 INFO [finetune.py:992] (0/2) Epoch 4, batch 6800, loss[loss=0.1441, simple_loss=0.2292, pruned_loss=0.02948, over 12135.00 frames. ], tot_loss[loss=0.1781, simple_loss=0.2653, pruned_loss=0.04549, over 2366852.84 frames. ], batch size: 30, lr: 4.79e-03, grad_scale: 8.0 2023-05-16 00:18:55,120 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.71 vs. limit=2.0 2023-05-16 00:19:09,477 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.145e+02 2.927e+02 3.459e+02 4.200e+02 8.756e+02, threshold=6.918e+02, percent-clipped=3.0 2023-05-16 00:19:13,214 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=143534.0, num_to_drop=1, layers_to_drop={2} 2023-05-16 00:19:25,826 INFO [finetune.py:992] (0/2) Epoch 4, batch 6850, loss[loss=0.1807, simple_loss=0.2762, pruned_loss=0.04265, over 12345.00 frames. ], tot_loss[loss=0.1782, simple_loss=0.2657, pruned_loss=0.04534, over 2374250.24 frames. ], batch size: 35, lr: 4.79e-03, grad_scale: 8.0 2023-05-16 00:19:54,202 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.1226, 5.7800, 5.3086, 5.3612, 5.8940, 5.2195, 5.4928, 5.3729], device='cuda:0'), covar=tensor([0.1241, 0.1082, 0.1153, 0.1997, 0.0921, 0.2212, 0.1703, 0.1134], device='cuda:0'), in_proj_covar=tensor([0.0330, 0.0462, 0.0361, 0.0414, 0.0440, 0.0417, 0.0373, 0.0360], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 00:20:02,973 INFO [finetune.py:992] (0/2) Epoch 4, batch 6900, loss[loss=0.1679, simple_loss=0.2612, pruned_loss=0.03732, over 12310.00 frames. ], tot_loss[loss=0.1775, simple_loss=0.2648, pruned_loss=0.04507, over 2375177.63 frames. ], batch size: 33, lr: 4.79e-03, grad_scale: 8.0 2023-05-16 00:20:09,628 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.8511, 5.8906, 5.6398, 5.1257, 4.9860, 5.7454, 5.2737, 5.1943], device='cuda:0'), covar=tensor([0.0738, 0.0752, 0.0607, 0.1424, 0.0765, 0.0690, 0.1518, 0.0959], device='cuda:0'), in_proj_covar=tensor([0.0571, 0.0511, 0.0478, 0.0590, 0.0382, 0.0662, 0.0728, 0.0535], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0004, 0.0004, 0.0003], device='cuda:0') 2023-05-16 00:20:22,135 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.944e+02 2.832e+02 3.237e+02 3.973e+02 1.176e+03, threshold=6.474e+02, percent-clipped=4.0 2023-05-16 00:20:38,458 INFO [finetune.py:992] (0/2) Epoch 4, batch 6950, loss[loss=0.2002, simple_loss=0.2856, pruned_loss=0.05741, over 10436.00 frames. ], tot_loss[loss=0.1777, simple_loss=0.2652, pruned_loss=0.04515, over 2358755.02 frames. ], batch size: 69, lr: 4.79e-03, grad_scale: 8.0 2023-05-16 00:21:00,385 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.3482, 2.4489, 3.0310, 4.1275, 2.5022, 4.2904, 4.2731, 4.4297], device='cuda:0'), covar=tensor([0.0139, 0.1150, 0.0498, 0.0167, 0.1131, 0.0230, 0.0179, 0.0089], device='cuda:0'), in_proj_covar=tensor([0.0121, 0.0204, 0.0188, 0.0119, 0.0190, 0.0177, 0.0170, 0.0122], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 00:21:10,351 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.1819, 2.2053, 2.6256, 3.0994, 2.2088, 3.2415, 3.1689, 3.3297], device='cuda:0'), covar=tensor([0.0160, 0.0945, 0.0446, 0.0179, 0.0994, 0.0284, 0.0267, 0.0107], device='cuda:0'), in_proj_covar=tensor([0.0121, 0.0204, 0.0188, 0.0119, 0.0190, 0.0177, 0.0171, 0.0122], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 00:21:14,357 INFO [finetune.py:992] (0/2) Epoch 4, batch 7000, loss[loss=0.143, simple_loss=0.2211, pruned_loss=0.03242, over 12295.00 frames. ], tot_loss[loss=0.1775, simple_loss=0.2648, pruned_loss=0.04511, over 2355736.06 frames. ], batch size: 28, lr: 4.79e-03, grad_scale: 8.0 2023-05-16 00:21:34,174 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.121e+02 2.902e+02 3.396e+02 4.384e+02 9.697e+02, threshold=6.791e+02, percent-clipped=7.0 2023-05-16 00:21:35,019 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.3108, 6.0922, 5.5983, 5.6640, 6.1162, 5.4567, 5.7708, 5.5968], device='cuda:0'), covar=tensor([0.1315, 0.0820, 0.1029, 0.1677, 0.0881, 0.2267, 0.1465, 0.1003], device='cuda:0'), in_proj_covar=tensor([0.0332, 0.0461, 0.0361, 0.0416, 0.0442, 0.0420, 0.0374, 0.0360], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 00:21:50,590 INFO [finetune.py:992] (0/2) Epoch 4, batch 7050, loss[loss=0.1681, simple_loss=0.2515, pruned_loss=0.04236, over 12417.00 frames. ], tot_loss[loss=0.1764, simple_loss=0.2637, pruned_loss=0.04454, over 2363804.17 frames. ], batch size: 32, lr: 4.78e-03, grad_scale: 8.0 2023-05-16 00:21:53,616 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=143756.0, num_to_drop=1, layers_to_drop={2} 2023-05-16 00:22:06,919 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 2023-05-16 00:22:26,498 INFO [finetune.py:992] (0/2) Epoch 4, batch 7100, loss[loss=0.168, simple_loss=0.2493, pruned_loss=0.04339, over 12273.00 frames. ], tot_loss[loss=0.1765, simple_loss=0.2639, pruned_loss=0.04454, over 2370808.88 frames. ], batch size: 28, lr: 4.78e-03, grad_scale: 8.0 2023-05-16 00:22:45,496 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.139e+02 2.928e+02 3.425e+02 3.906e+02 8.366e+02, threshold=6.851e+02, percent-clipped=2.0 2023-05-16 00:22:45,625 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=143829.0, num_to_drop=1, layers_to_drop={3} 2023-05-16 00:23:02,628 INFO [finetune.py:992] (0/2) Epoch 4, batch 7150, loss[loss=0.1722, simple_loss=0.2672, pruned_loss=0.03865, over 12153.00 frames. ], tot_loss[loss=0.1773, simple_loss=0.2647, pruned_loss=0.04494, over 2367377.53 frames. ], batch size: 36, lr: 4.78e-03, grad_scale: 8.0 2023-05-16 00:23:38,816 INFO [finetune.py:992] (0/2) Epoch 4, batch 7200, loss[loss=0.1997, simple_loss=0.2882, pruned_loss=0.05563, over 12052.00 frames. ], tot_loss[loss=0.1779, simple_loss=0.2651, pruned_loss=0.04531, over 2358533.76 frames. ], batch size: 37, lr: 4.78e-03, grad_scale: 8.0 2023-05-16 00:23:52,460 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.9866, 4.8973, 5.0054, 4.9732, 4.6420, 4.6946, 4.5596, 4.9631], device='cuda:0'), covar=tensor([0.0680, 0.0495, 0.0598, 0.0546, 0.1639, 0.1186, 0.0487, 0.0814], device='cuda:0'), in_proj_covar=tensor([0.0488, 0.0627, 0.0546, 0.0583, 0.0774, 0.0694, 0.0510, 0.0458], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0005, 0.0004, 0.0004, 0.0005, 0.0005, 0.0004, 0.0004], device='cuda:0') 2023-05-16 00:23:55,359 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.5348, 3.6443, 3.3474, 3.2226, 2.9328, 2.7173, 3.6378, 2.2190], device='cuda:0'), covar=tensor([0.0353, 0.0136, 0.0162, 0.0161, 0.0384, 0.0380, 0.0126, 0.0460], device='cuda:0'), in_proj_covar=tensor([0.0190, 0.0163, 0.0154, 0.0180, 0.0203, 0.0198, 0.0161, 0.0201], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 00:23:57,996 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.823e+02 2.800e+02 3.188e+02 3.874e+02 1.813e+03, threshold=6.376e+02, percent-clipped=2.0 2023-05-16 00:23:59,972 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.40 vs. limit=2.0 2023-05-16 00:24:14,529 INFO [finetune.py:992] (0/2) Epoch 4, batch 7250, loss[loss=0.1514, simple_loss=0.2377, pruned_loss=0.03258, over 12120.00 frames. ], tot_loss[loss=0.1769, simple_loss=0.2646, pruned_loss=0.04462, over 2366629.96 frames. ], batch size: 30, lr: 4.78e-03, grad_scale: 8.0 2023-05-16 00:24:49,607 INFO [checkpoint.py:75] (0/2) Saving checkpoint to pruned_transducer_stateless7/exp_giga_finetune/checkpoint-44000.pt 2023-05-16 00:24:54,151 INFO [finetune.py:992] (0/2) Epoch 4, batch 7300, loss[loss=0.1819, simple_loss=0.2691, pruned_loss=0.04735, over 12250.00 frames. ], tot_loss[loss=0.176, simple_loss=0.2635, pruned_loss=0.04424, over 2371943.41 frames. ], batch size: 32, lr: 4.78e-03, grad_scale: 8.0 2023-05-16 00:25:13,983 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.853e+02 2.750e+02 3.191e+02 3.884e+02 1.605e+03, threshold=6.381e+02, percent-clipped=3.0 2023-05-16 00:25:30,364 INFO [finetune.py:992] (0/2) Epoch 4, batch 7350, loss[loss=0.1455, simple_loss=0.2349, pruned_loss=0.02804, over 12295.00 frames. ], tot_loss[loss=0.1765, simple_loss=0.2637, pruned_loss=0.04463, over 2371890.07 frames. ], batch size: 33, lr: 4.78e-03, grad_scale: 8.0 2023-05-16 00:25:31,453 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.15 vs. limit=2.0 2023-05-16 00:25:33,412 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=144056.0, num_to_drop=1, layers_to_drop={0} 2023-05-16 00:25:53,768 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.67 vs. limit=2.0 2023-05-16 00:26:06,064 INFO [finetune.py:992] (0/2) Epoch 4, batch 7400, loss[loss=0.1767, simple_loss=0.2671, pruned_loss=0.04313, over 12087.00 frames. ], tot_loss[loss=0.1762, simple_loss=0.2634, pruned_loss=0.04451, over 2369221.15 frames. ], batch size: 38, lr: 4.78e-03, grad_scale: 8.0 2023-05-16 00:26:07,439 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=144104.0, num_to_drop=1, layers_to_drop={1} 2023-05-16 00:26:24,675 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.875e+02 2.946e+02 3.343e+02 3.974e+02 8.907e+02, threshold=6.685e+02, percent-clipped=2.0 2023-05-16 00:26:24,869 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=144129.0, num_to_drop=1, layers_to_drop={0} 2023-05-16 00:26:42,217 INFO [finetune.py:992] (0/2) Epoch 4, batch 7450, loss[loss=0.1803, simple_loss=0.2506, pruned_loss=0.05498, over 11797.00 frames. ], tot_loss[loss=0.1777, simple_loss=0.2646, pruned_loss=0.04535, over 2370676.57 frames. ], batch size: 26, lr: 4.78e-03, grad_scale: 8.0 2023-05-16 00:26:43,157 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.1207, 4.3227, 4.2011, 4.6222, 3.0582, 4.1473, 2.6340, 4.1844], device='cuda:0'), covar=tensor([0.1625, 0.0542, 0.0844, 0.0536, 0.1116, 0.0484, 0.1729, 0.1320], device='cuda:0'), in_proj_covar=tensor([0.0227, 0.0261, 0.0293, 0.0351, 0.0240, 0.0236, 0.0258, 0.0386], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 00:27:00,262 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=144177.0, num_to_drop=1, layers_to_drop={1} 2023-05-16 00:27:02,521 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.5787, 4.8237, 4.2636, 4.9908, 4.8484, 3.0578, 4.5130, 3.1547], device='cuda:0'), covar=tensor([0.0584, 0.0704, 0.1122, 0.0406, 0.0813, 0.1476, 0.0849, 0.3161], device='cuda:0'), in_proj_covar=tensor([0.0291, 0.0365, 0.0342, 0.0259, 0.0351, 0.0257, 0.0330, 0.0354], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 00:27:04,572 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.5875, 2.5366, 3.2772, 4.3561, 2.4104, 4.5269, 4.5051, 4.7528], device='cuda:0'), covar=tensor([0.0139, 0.1121, 0.0395, 0.0163, 0.1197, 0.0162, 0.0143, 0.0083], device='cuda:0'), in_proj_covar=tensor([0.0119, 0.0201, 0.0186, 0.0118, 0.0189, 0.0175, 0.0170, 0.0120], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 00:27:17,983 INFO [finetune.py:992] (0/2) Epoch 4, batch 7500, loss[loss=0.1528, simple_loss=0.2424, pruned_loss=0.03158, over 12079.00 frames. ], tot_loss[loss=0.1774, simple_loss=0.2643, pruned_loss=0.04525, over 2373807.70 frames. ], batch size: 32, lr: 4.78e-03, grad_scale: 8.0 2023-05-16 00:27:32,520 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.0611, 4.7953, 4.9304, 5.0107, 4.7638, 4.9837, 4.9160, 2.7408], device='cuda:0'), covar=tensor([0.0094, 0.0056, 0.0077, 0.0055, 0.0040, 0.0088, 0.0068, 0.0641], device='cuda:0'), in_proj_covar=tensor([0.0064, 0.0072, 0.0077, 0.0069, 0.0057, 0.0087, 0.0074, 0.0089], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 00:27:37,167 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.000e+02 2.942e+02 3.365e+02 3.938e+02 8.869e+02, threshold=6.730e+02, percent-clipped=6.0 2023-05-16 00:27:53,733 INFO [finetune.py:992] (0/2) Epoch 4, batch 7550, loss[loss=0.1576, simple_loss=0.2411, pruned_loss=0.03703, over 12266.00 frames. ], tot_loss[loss=0.1766, simple_loss=0.264, pruned_loss=0.04462, over 2383764.89 frames. ], batch size: 28, lr: 4.78e-03, grad_scale: 8.0 2023-05-16 00:28:30,128 INFO [finetune.py:992] (0/2) Epoch 4, batch 7600, loss[loss=0.2028, simple_loss=0.2801, pruned_loss=0.06274, over 12282.00 frames. ], tot_loss[loss=0.1775, simple_loss=0.2646, pruned_loss=0.04526, over 2377790.00 frames. ], batch size: 33, lr: 4.78e-03, grad_scale: 8.0 2023-05-16 00:28:33,093 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.5711, 5.3828, 5.4636, 5.5270, 5.0847, 5.1454, 5.0546, 5.4666], device='cuda:0'), covar=tensor([0.0557, 0.0502, 0.0728, 0.0495, 0.1879, 0.1208, 0.0459, 0.0864], device='cuda:0'), in_proj_covar=tensor([0.0491, 0.0631, 0.0549, 0.0584, 0.0778, 0.0698, 0.0515, 0.0458], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0005, 0.0004, 0.0004, 0.0005, 0.0005, 0.0004, 0.0004], device='cuda:0') 2023-05-16 00:28:33,129 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.2216, 5.2291, 5.0944, 5.1587, 4.6966, 5.1460, 5.2292, 5.4125], device='cuda:0'), covar=tensor([0.0190, 0.0102, 0.0153, 0.0224, 0.0665, 0.0227, 0.0118, 0.0113], device='cuda:0'), in_proj_covar=tensor([0.0181, 0.0183, 0.0182, 0.0235, 0.0233, 0.0201, 0.0166, 0.0214], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0005, 0.0005, 0.0005, 0.0004, 0.0005], device='cuda:0') 2023-05-16 00:28:45,755 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.6877, 2.7137, 3.7747, 4.6862, 4.0111, 4.5598, 3.8791, 3.3945], device='cuda:0'), covar=tensor([0.0027, 0.0335, 0.0127, 0.0035, 0.0101, 0.0074, 0.0105, 0.0290], device='cuda:0'), in_proj_covar=tensor([0.0080, 0.0119, 0.0101, 0.0074, 0.0099, 0.0110, 0.0088, 0.0135], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 00:28:49,091 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.780e+02 2.903e+02 3.452e+02 4.012e+02 8.262e+02, threshold=6.904e+02, percent-clipped=5.0 2023-05-16 00:29:05,350 INFO [finetune.py:992] (0/2) Epoch 4, batch 7650, loss[loss=0.1875, simple_loss=0.2726, pruned_loss=0.05123, over 11081.00 frames. ], tot_loss[loss=0.1776, simple_loss=0.2644, pruned_loss=0.04538, over 2368676.50 frames. ], batch size: 55, lr: 4.78e-03, grad_scale: 8.0 2023-05-16 00:29:41,340 INFO [finetune.py:992] (0/2) Epoch 4, batch 7700, loss[loss=0.1927, simple_loss=0.2926, pruned_loss=0.04635, over 12137.00 frames. ], tot_loss[loss=0.1766, simple_loss=0.2635, pruned_loss=0.04483, over 2369912.28 frames. ], batch size: 34, lr: 4.78e-03, grad_scale: 8.0 2023-05-16 00:29:43,799 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 2023-05-16 00:29:47,777 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=144411.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 00:30:01,776 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.609e+02 2.806e+02 3.415e+02 3.909e+02 5.886e+02, threshold=6.830e+02, percent-clipped=0.0 2023-05-16 00:30:18,034 INFO [finetune.py:992] (0/2) Epoch 4, batch 7750, loss[loss=0.148, simple_loss=0.2325, pruned_loss=0.03171, over 12250.00 frames. ], tot_loss[loss=0.1768, simple_loss=0.2638, pruned_loss=0.04492, over 2376826.99 frames. ], batch size: 32, lr: 4.78e-03, grad_scale: 8.0 2023-05-16 00:30:32,129 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=144472.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 00:30:53,342 INFO [finetune.py:992] (0/2) Epoch 4, batch 7800, loss[loss=0.1627, simple_loss=0.2562, pruned_loss=0.03462, over 12315.00 frames. ], tot_loss[loss=0.1761, simple_loss=0.2633, pruned_loss=0.04448, over 2378829.74 frames. ], batch size: 34, lr: 4.78e-03, grad_scale: 8.0 2023-05-16 00:31:12,466 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.967e+02 2.788e+02 3.406e+02 4.091e+02 6.144e+02, threshold=6.812e+02, percent-clipped=0.0 2023-05-16 00:31:28,698 INFO [finetune.py:992] (0/2) Epoch 4, batch 7850, loss[loss=0.2074, simple_loss=0.2918, pruned_loss=0.06147, over 10348.00 frames. ], tot_loss[loss=0.1756, simple_loss=0.2629, pruned_loss=0.04417, over 2383158.48 frames. ], batch size: 68, lr: 4.78e-03, grad_scale: 8.0 2023-05-16 00:31:40,595 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.41 vs. limit=2.0 2023-05-16 00:32:05,923 INFO [finetune.py:992] (0/2) Epoch 4, batch 7900, loss[loss=0.1638, simple_loss=0.2531, pruned_loss=0.03724, over 12384.00 frames. ], tot_loss[loss=0.1763, simple_loss=0.2638, pruned_loss=0.04442, over 2378996.11 frames. ], batch size: 36, lr: 4.78e-03, grad_scale: 8.0 2023-05-16 00:32:25,024 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.084e+02 3.076e+02 3.537e+02 4.366e+02 9.547e+02, threshold=7.073e+02, percent-clipped=4.0 2023-05-16 00:32:41,093 INFO [finetune.py:992] (0/2) Epoch 4, batch 7950, loss[loss=0.1757, simple_loss=0.2656, pruned_loss=0.04295, over 12287.00 frames. ], tot_loss[loss=0.1782, simple_loss=0.2657, pruned_loss=0.04528, over 2369002.61 frames. ], batch size: 37, lr: 4.78e-03, grad_scale: 8.0 2023-05-16 00:33:11,247 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.6569, 4.6972, 4.3568, 5.0507, 4.7415, 2.9437, 4.3662, 3.1815], device='cuda:0'), covar=tensor([0.0583, 0.0750, 0.1043, 0.0434, 0.0804, 0.1464, 0.0872, 0.3043], device='cuda:0'), in_proj_covar=tensor([0.0296, 0.0372, 0.0349, 0.0263, 0.0356, 0.0262, 0.0335, 0.0358], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 00:33:16,684 INFO [finetune.py:992] (0/2) Epoch 4, batch 8000, loss[loss=0.1854, simple_loss=0.2785, pruned_loss=0.04616, over 12082.00 frames. ], tot_loss[loss=0.1772, simple_loss=0.2648, pruned_loss=0.04485, over 2374928.16 frames. ], batch size: 38, lr: 4.78e-03, grad_scale: 8.0 2023-05-16 00:33:16,917 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.3114, 4.7436, 2.9639, 2.5543, 4.0732, 2.4049, 4.0277, 3.2408], device='cuda:0'), covar=tensor([0.0673, 0.0445, 0.0986, 0.1560, 0.0242, 0.1356, 0.0413, 0.0765], device='cuda:0'), in_proj_covar=tensor([0.0185, 0.0250, 0.0172, 0.0198, 0.0138, 0.0179, 0.0193, 0.0172], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 00:33:36,351 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.5199, 2.5242, 3.3192, 4.3442, 2.2740, 4.5724, 4.4878, 4.5324], device='cuda:0'), covar=tensor([0.0127, 0.1080, 0.0378, 0.0133, 0.1242, 0.0152, 0.0153, 0.0117], device='cuda:0'), in_proj_covar=tensor([0.0120, 0.0203, 0.0187, 0.0120, 0.0190, 0.0175, 0.0171, 0.0122], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 00:33:37,513 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.804e+02 2.930e+02 3.311e+02 4.094e+02 7.677e+02, threshold=6.621e+02, percent-clipped=2.0 2023-05-16 00:33:53,706 INFO [finetune.py:992] (0/2) Epoch 4, batch 8050, loss[loss=0.1715, simple_loss=0.2681, pruned_loss=0.0375, over 11526.00 frames. ], tot_loss[loss=0.1769, simple_loss=0.2642, pruned_loss=0.04481, over 2375806.71 frames. ], batch size: 48, lr: 4.78e-03, grad_scale: 8.0 2023-05-16 00:34:04,563 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=144767.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 00:34:27,194 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.3937, 5.1955, 5.3193, 5.3400, 4.9510, 5.0270, 4.8530, 5.3097], device='cuda:0'), covar=tensor([0.0614, 0.0553, 0.0686, 0.0581, 0.1777, 0.1171, 0.0518, 0.0822], device='cuda:0'), in_proj_covar=tensor([0.0487, 0.0630, 0.0548, 0.0582, 0.0774, 0.0694, 0.0513, 0.0458], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0005, 0.0004, 0.0004, 0.0005, 0.0005, 0.0004, 0.0004], device='cuda:0') 2023-05-16 00:34:29,526 INFO [finetune.py:992] (0/2) Epoch 4, batch 8100, loss[loss=0.1359, simple_loss=0.2249, pruned_loss=0.02348, over 12287.00 frames. ], tot_loss[loss=0.1759, simple_loss=0.2632, pruned_loss=0.04435, over 2375659.78 frames. ], batch size: 28, lr: 4.78e-03, grad_scale: 8.0 2023-05-16 00:34:41,305 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.1077, 3.8264, 3.8642, 4.1719, 2.7894, 3.7278, 2.3353, 3.8929], device='cuda:0'), covar=tensor([0.1611, 0.0798, 0.1069, 0.0905, 0.1268, 0.0701, 0.1936, 0.1263], device='cuda:0'), in_proj_covar=tensor([0.0222, 0.0256, 0.0287, 0.0346, 0.0235, 0.0234, 0.0252, 0.0378], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 00:34:48,720 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.835e+02 2.810e+02 3.253e+02 3.981e+02 7.251e+02, threshold=6.507e+02, percent-clipped=2.0 2023-05-16 00:35:04,979 INFO [finetune.py:992] (0/2) Epoch 4, batch 8150, loss[loss=0.1776, simple_loss=0.2549, pruned_loss=0.05018, over 12351.00 frames. ], tot_loss[loss=0.1766, simple_loss=0.2638, pruned_loss=0.04468, over 2367075.00 frames. ], batch size: 31, lr: 4.77e-03, grad_scale: 8.0 2023-05-16 00:35:42,018 INFO [finetune.py:992] (0/2) Epoch 4, batch 8200, loss[loss=0.1794, simple_loss=0.2624, pruned_loss=0.0482, over 12309.00 frames. ], tot_loss[loss=0.1765, simple_loss=0.2637, pruned_loss=0.04462, over 2372796.30 frames. ], batch size: 34, lr: 4.77e-03, grad_scale: 8.0 2023-05-16 00:36:01,028 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.946e+02 2.917e+02 3.512e+02 3.999e+02 6.164e+02, threshold=7.024e+02, percent-clipped=0.0 2023-05-16 00:36:12,716 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([1.8946, 3.9412, 4.0847, 4.3372, 2.9342, 3.8100, 2.4336, 3.9593], device='cuda:0'), covar=tensor([0.1843, 0.0787, 0.0883, 0.0663, 0.1204, 0.0699, 0.1925, 0.1443], device='cuda:0'), in_proj_covar=tensor([0.0221, 0.0256, 0.0287, 0.0344, 0.0234, 0.0233, 0.0250, 0.0377], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 00:36:16,457 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.95 vs. limit=2.0 2023-05-16 00:36:17,396 INFO [finetune.py:992] (0/2) Epoch 4, batch 8250, loss[loss=0.1676, simple_loss=0.2524, pruned_loss=0.0414, over 12105.00 frames. ], tot_loss[loss=0.177, simple_loss=0.264, pruned_loss=0.04495, over 2372729.43 frames. ], batch size: 33, lr: 4.77e-03, grad_scale: 8.0 2023-05-16 00:36:48,811 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=144996.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 00:36:53,258 INFO [finetune.py:992] (0/2) Epoch 4, batch 8300, loss[loss=0.1696, simple_loss=0.259, pruned_loss=0.04007, over 12116.00 frames. ], tot_loss[loss=0.1765, simple_loss=0.2636, pruned_loss=0.04471, over 2370647.53 frames. ], batch size: 33, lr: 4.77e-03, grad_scale: 8.0 2023-05-16 00:37:12,490 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=145027.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 00:37:13,771 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.933e+02 2.880e+02 3.303e+02 3.913e+02 1.017e+03, threshold=6.606e+02, percent-clipped=4.0 2023-05-16 00:37:30,151 INFO [finetune.py:992] (0/2) Epoch 4, batch 8350, loss[loss=0.2206, simple_loss=0.2968, pruned_loss=0.07222, over 12117.00 frames. ], tot_loss[loss=0.1756, simple_loss=0.2632, pruned_loss=0.04402, over 2378122.23 frames. ], batch size: 38, lr: 4.77e-03, grad_scale: 8.0 2023-05-16 00:37:33,875 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=145057.0, num_to_drop=1, layers_to_drop={3} 2023-05-16 00:37:39,860 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=145066.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 00:37:40,553 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=145067.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 00:37:50,564 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.3011, 4.7736, 2.8371, 2.3772, 4.2029, 2.1001, 4.0863, 3.0027], device='cuda:0'), covar=tensor([0.0528, 0.0350, 0.0982, 0.1664, 0.0217, 0.1521, 0.0345, 0.0913], device='cuda:0'), in_proj_covar=tensor([0.0184, 0.0248, 0.0172, 0.0196, 0.0137, 0.0178, 0.0192, 0.0172], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 00:37:55,333 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=145088.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 00:38:05,109 INFO [finetune.py:992] (0/2) Epoch 4, batch 8400, loss[loss=0.1676, simple_loss=0.2553, pruned_loss=0.03996, over 12104.00 frames. ], tot_loss[loss=0.1758, simple_loss=0.2632, pruned_loss=0.04423, over 2374980.07 frames. ], batch size: 33, lr: 4.77e-03, grad_scale: 8.0 2023-05-16 00:38:14,185 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=145115.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 00:38:22,658 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=145127.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 00:38:23,890 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.875e+02 2.792e+02 3.296e+02 4.039e+02 8.655e+02, threshold=6.593e+02, percent-clipped=1.0 2023-05-16 00:38:40,184 INFO [finetune.py:992] (0/2) Epoch 4, batch 8450, loss[loss=0.1733, simple_loss=0.2672, pruned_loss=0.03976, over 12354.00 frames. ], tot_loss[loss=0.1779, simple_loss=0.2648, pruned_loss=0.04546, over 2368066.76 frames. ], batch size: 35, lr: 4.77e-03, grad_scale: 8.0 2023-05-16 00:38:40,701 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.84 vs. limit=2.0 2023-05-16 00:38:58,248 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.78 vs. limit=2.0 2023-05-16 00:39:16,980 INFO [finetune.py:992] (0/2) Epoch 4, batch 8500, loss[loss=0.1853, simple_loss=0.281, pruned_loss=0.04482, over 12146.00 frames. ], tot_loss[loss=0.1768, simple_loss=0.2639, pruned_loss=0.04482, over 2374840.00 frames. ], batch size: 36, lr: 4.77e-03, grad_scale: 8.0 2023-05-16 00:39:36,203 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.010e+02 2.838e+02 3.322e+02 4.007e+02 1.046e+03, threshold=6.644e+02, percent-clipped=1.0 2023-05-16 00:39:52,529 INFO [finetune.py:992] (0/2) Epoch 4, batch 8550, loss[loss=0.1831, simple_loss=0.2722, pruned_loss=0.04698, over 11617.00 frames. ], tot_loss[loss=0.1774, simple_loss=0.2645, pruned_loss=0.04511, over 2372589.39 frames. ], batch size: 48, lr: 4.77e-03, grad_scale: 8.0 2023-05-16 00:40:28,891 INFO [finetune.py:992] (0/2) Epoch 4, batch 8600, loss[loss=0.1902, simple_loss=0.2768, pruned_loss=0.05186, over 12096.00 frames. ], tot_loss[loss=0.1776, simple_loss=0.2643, pruned_loss=0.04545, over 2374826.66 frames. ], batch size: 38, lr: 4.77e-03, grad_scale: 8.0 2023-05-16 00:40:48,704 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.875e+02 3.182e+02 3.638e+02 4.401e+02 1.025e+03, threshold=7.277e+02, percent-clipped=3.0 2023-05-16 00:41:05,251 INFO [finetune.py:992] (0/2) Epoch 4, batch 8650, loss[loss=0.157, simple_loss=0.2487, pruned_loss=0.03262, over 12028.00 frames. ], tot_loss[loss=0.1761, simple_loss=0.2631, pruned_loss=0.04451, over 2379769.90 frames. ], batch size: 31, lr: 4.77e-03, grad_scale: 8.0 2023-05-16 00:41:05,357 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=145352.0, num_to_drop=1, layers_to_drop={0} 2023-05-16 00:41:10,504 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.9842, 4.2539, 3.7546, 4.6460, 4.2561, 2.7428, 3.9349, 2.8525], device='cuda:0'), covar=tensor([0.0828, 0.0870, 0.1453, 0.0421, 0.1065, 0.1555, 0.0954, 0.3081], device='cuda:0'), in_proj_covar=tensor([0.0288, 0.0361, 0.0338, 0.0254, 0.0346, 0.0254, 0.0325, 0.0348], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 00:41:17,467 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.8719, 2.5632, 4.8376, 5.1711, 3.3653, 2.5550, 2.8793, 1.9244], device='cuda:0'), covar=tensor([0.1351, 0.3481, 0.0387, 0.0258, 0.0936, 0.2276, 0.2683, 0.4627], device='cuda:0'), in_proj_covar=tensor([0.0272, 0.0369, 0.0262, 0.0287, 0.0251, 0.0280, 0.0346, 0.0351], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-16 00:41:18,824 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=145371.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 00:41:27,235 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=145383.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 00:41:40,865 INFO [finetune.py:992] (0/2) Epoch 4, batch 8700, loss[loss=0.2043, simple_loss=0.3038, pruned_loss=0.05235, over 11817.00 frames. ], tot_loss[loss=0.1759, simple_loss=0.2628, pruned_loss=0.0445, over 2375408.53 frames. ], batch size: 44, lr: 4.77e-03, grad_scale: 16.0 2023-05-16 00:41:41,809 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.1229, 2.3627, 3.7108, 3.1173, 3.4997, 3.2037, 2.5294, 3.5914], device='cuda:0'), covar=tensor([0.0100, 0.0305, 0.0122, 0.0184, 0.0114, 0.0133, 0.0279, 0.0095], device='cuda:0'), in_proj_covar=tensor([0.0161, 0.0194, 0.0171, 0.0175, 0.0197, 0.0151, 0.0184, 0.0166], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 00:41:52,469 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.83 vs. limit=2.0 2023-05-16 00:41:54,898 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=145422.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 00:41:57,375 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2023-05-16 00:41:59,822 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.107e+02 3.044e+02 3.423e+02 4.062e+02 6.784e+02, threshold=6.847e+02, percent-clipped=1.0 2023-05-16 00:42:02,103 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=145432.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 00:42:06,371 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.50 vs. limit=2.0 2023-05-16 00:42:16,693 INFO [finetune.py:992] (0/2) Epoch 4, batch 8750, loss[loss=0.1862, simple_loss=0.2727, pruned_loss=0.04988, over 12041.00 frames. ], tot_loss[loss=0.1757, simple_loss=0.2629, pruned_loss=0.04421, over 2383671.99 frames. ], batch size: 40, lr: 4.77e-03, grad_scale: 16.0 2023-05-16 00:42:25,525 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.2667, 4.2683, 4.0543, 4.4209, 3.2613, 3.9804, 2.5590, 4.0732], device='cuda:0'), covar=tensor([0.1510, 0.0550, 0.0884, 0.0554, 0.1003, 0.0548, 0.1736, 0.1225], device='cuda:0'), in_proj_covar=tensor([0.0222, 0.0255, 0.0287, 0.0346, 0.0235, 0.0233, 0.0251, 0.0376], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 00:42:40,706 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.0433, 5.0191, 4.8586, 5.0017, 3.9491, 5.1298, 5.0105, 5.1495], device='cuda:0'), covar=tensor([0.0179, 0.0142, 0.0173, 0.0280, 0.1067, 0.0258, 0.0165, 0.0196], device='cuda:0'), in_proj_covar=tensor([0.0179, 0.0182, 0.0180, 0.0234, 0.0231, 0.0200, 0.0166, 0.0214], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0005, 0.0005, 0.0005, 0.0004, 0.0005], device='cuda:0') 2023-05-16 00:42:53,180 INFO [finetune.py:992] (0/2) Epoch 4, batch 8800, loss[loss=0.1868, simple_loss=0.2842, pruned_loss=0.04469, over 11619.00 frames. ], tot_loss[loss=0.177, simple_loss=0.2642, pruned_loss=0.0449, over 2374310.00 frames. ], batch size: 48, lr: 4.77e-03, grad_scale: 16.0 2023-05-16 00:43:06,078 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.2344, 4.8154, 2.9018, 2.9188, 4.0094, 2.7437, 4.0137, 3.2571], device='cuda:0'), covar=tensor([0.0735, 0.0496, 0.1109, 0.1329, 0.0339, 0.1212, 0.0505, 0.0813], device='cuda:0'), in_proj_covar=tensor([0.0185, 0.0251, 0.0173, 0.0198, 0.0139, 0.0178, 0.0194, 0.0173], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 00:43:08,047 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.8552, 5.6486, 5.2357, 5.1931, 5.6997, 5.0600, 5.3028, 5.1678], device='cuda:0'), covar=tensor([0.1440, 0.0879, 0.0895, 0.1888, 0.0883, 0.1990, 0.1473, 0.1090], device='cuda:0'), in_proj_covar=tensor([0.0335, 0.0456, 0.0361, 0.0410, 0.0443, 0.0415, 0.0372, 0.0362], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 00:43:12,279 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.939e+02 2.956e+02 3.531e+02 4.140e+02 9.183e+02, threshold=7.062e+02, percent-clipped=2.0 2023-05-16 00:43:28,596 INFO [finetune.py:992] (0/2) Epoch 4, batch 8850, loss[loss=0.1627, simple_loss=0.2523, pruned_loss=0.03653, over 12101.00 frames. ], tot_loss[loss=0.1763, simple_loss=0.2639, pruned_loss=0.04433, over 2373196.60 frames. ], batch size: 33, lr: 4.77e-03, grad_scale: 16.0 2023-05-16 00:43:59,576 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=145594.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 00:44:04,789 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.3550, 2.1487, 3.0907, 4.1163, 2.2242, 4.2323, 4.2539, 4.3949], device='cuda:0'), covar=tensor([0.0111, 0.1331, 0.0460, 0.0160, 0.1338, 0.0199, 0.0139, 0.0091], device='cuda:0'), in_proj_covar=tensor([0.0120, 0.0203, 0.0186, 0.0120, 0.0190, 0.0176, 0.0170, 0.0120], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 00:44:05,305 INFO [finetune.py:992] (0/2) Epoch 4, batch 8900, loss[loss=0.1882, simple_loss=0.2821, pruned_loss=0.04719, over 12029.00 frames. ], tot_loss[loss=0.1769, simple_loss=0.2645, pruned_loss=0.04468, over 2367392.10 frames. ], batch size: 42, lr: 4.77e-03, grad_scale: 16.0 2023-05-16 00:44:07,665 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.9822, 4.6211, 4.8185, 4.7629, 4.6626, 4.8732, 4.7359, 2.5903], device='cuda:0'), covar=tensor([0.0097, 0.0071, 0.0076, 0.0071, 0.0052, 0.0083, 0.0080, 0.0747], device='cuda:0'), in_proj_covar=tensor([0.0065, 0.0073, 0.0077, 0.0071, 0.0058, 0.0088, 0.0075, 0.0090], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 00:44:25,119 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.126e+02 2.816e+02 3.227e+02 3.903e+02 1.604e+03, threshold=6.453e+02, percent-clipped=3.0 2023-05-16 00:44:35,347 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=145643.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 00:44:41,703 INFO [finetune.py:992] (0/2) Epoch 4, batch 8950, loss[loss=0.2016, simple_loss=0.2873, pruned_loss=0.05795, over 10640.00 frames. ], tot_loss[loss=0.1772, simple_loss=0.2649, pruned_loss=0.04478, over 2366267.42 frames. ], batch size: 68, lr: 4.77e-03, grad_scale: 16.0 2023-05-16 00:44:41,860 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=145652.0, num_to_drop=1, layers_to_drop={1} 2023-05-16 00:44:44,007 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=145655.0, num_to_drop=1, layers_to_drop={1} 2023-05-16 00:44:49,033 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.1815, 2.4840, 3.7906, 3.1340, 3.6063, 3.1869, 2.6158, 3.6481], device='cuda:0'), covar=tensor([0.0128, 0.0328, 0.0130, 0.0242, 0.0114, 0.0177, 0.0337, 0.0108], device='cuda:0'), in_proj_covar=tensor([0.0164, 0.0197, 0.0175, 0.0178, 0.0200, 0.0153, 0.0188, 0.0170], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 00:45:03,695 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=145683.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 00:45:15,585 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=145700.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 00:45:16,951 INFO [finetune.py:992] (0/2) Epoch 4, batch 9000, loss[loss=0.1804, simple_loss=0.2692, pruned_loss=0.0458, over 12124.00 frames. ], tot_loss[loss=0.177, simple_loss=0.2646, pruned_loss=0.04471, over 2371712.17 frames. ], batch size: 33, lr: 4.77e-03, grad_scale: 16.0 2023-05-16 00:45:16,952 INFO [finetune.py:1017] (0/2) Computing validation loss 2023-05-16 00:45:35,139 INFO [finetune.py:1026] (0/2) Epoch 4, validation: loss=0.3366, simple_loss=0.4058, pruned_loss=0.1337, over 1020973.00 frames. 2023-05-16 00:45:35,140 INFO [finetune.py:1027] (0/2) Maximum memory allocated so far is 12745MB 2023-05-16 00:45:36,734 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=145704.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 00:45:50,211 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=145722.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 00:45:53,873 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=145727.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 00:45:55,124 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.765e+02 2.863e+02 3.398e+02 3.852e+02 7.519e+02, threshold=6.796e+02, percent-clipped=5.0 2023-05-16 00:45:56,659 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=145731.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 00:46:11,476 INFO [finetune.py:992] (0/2) Epoch 4, batch 9050, loss[loss=0.189, simple_loss=0.2838, pruned_loss=0.04707, over 12052.00 frames. ], tot_loss[loss=0.1775, simple_loss=0.265, pruned_loss=0.04499, over 2370832.87 frames. ], batch size: 42, lr: 4.77e-03, grad_scale: 16.0 2023-05-16 00:46:24,186 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=145770.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 00:46:47,008 INFO [finetune.py:992] (0/2) Epoch 4, batch 9100, loss[loss=0.1406, simple_loss=0.2195, pruned_loss=0.03081, over 12341.00 frames. ], tot_loss[loss=0.1773, simple_loss=0.2649, pruned_loss=0.04487, over 2369824.84 frames. ], batch size: 30, lr: 4.77e-03, grad_scale: 16.0 2023-05-16 00:46:53,042 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.5503, 4.6996, 4.1181, 5.0722, 4.6195, 3.0038, 4.4084, 3.0503], device='cuda:0'), covar=tensor([0.0646, 0.0790, 0.1281, 0.0358, 0.0899, 0.1400, 0.0855, 0.3062], device='cuda:0'), in_proj_covar=tensor([0.0290, 0.0364, 0.0341, 0.0258, 0.0350, 0.0257, 0.0328, 0.0351], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 00:47:00,763 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=145821.0, num_to_drop=1, layers_to_drop={0} 2023-05-16 00:47:07,048 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.899e+02 2.816e+02 3.541e+02 4.634e+02 1.137e+03, threshold=7.082e+02, percent-clipped=3.0 2023-05-16 00:47:23,171 INFO [finetune.py:992] (0/2) Epoch 4, batch 9150, loss[loss=0.2623, simple_loss=0.3229, pruned_loss=0.1009, over 8299.00 frames. ], tot_loss[loss=0.1768, simple_loss=0.2643, pruned_loss=0.04465, over 2370779.96 frames. ], batch size: 98, lr: 4.77e-03, grad_scale: 16.0 2023-05-16 00:47:36,044 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.5522, 5.0428, 5.4841, 4.7480, 5.1839, 4.9118, 5.5205, 5.1351], device='cuda:0'), covar=tensor([0.0195, 0.0332, 0.0249, 0.0222, 0.0245, 0.0260, 0.0194, 0.0183], device='cuda:0'), in_proj_covar=tensor([0.0238, 0.0245, 0.0264, 0.0239, 0.0236, 0.0237, 0.0216, 0.0197], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-05-16 00:47:45,448 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=145882.0, num_to_drop=1, layers_to_drop={2} 2023-05-16 00:47:59,147 INFO [finetune.py:992] (0/2) Epoch 4, batch 9200, loss[loss=0.1727, simple_loss=0.2597, pruned_loss=0.04286, over 12031.00 frames. ], tot_loss[loss=0.1768, simple_loss=0.2645, pruned_loss=0.04448, over 2375844.93 frames. ], batch size: 40, lr: 4.77e-03, grad_scale: 16.0 2023-05-16 00:48:18,036 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.866e+02 2.848e+02 3.273e+02 4.077e+02 7.326e+02, threshold=6.546e+02, percent-clipped=2.0 2023-05-16 00:48:27,612 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.31 vs. limit=2.0 2023-05-16 00:48:33,041 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=145950.0, num_to_drop=1, layers_to_drop={0} 2023-05-16 00:48:34,105 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.91 vs. limit=2.0 2023-05-16 00:48:34,319 INFO [finetune.py:992] (0/2) Epoch 4, batch 9250, loss[loss=0.1879, simple_loss=0.2834, pruned_loss=0.04614, over 12123.00 frames. ], tot_loss[loss=0.1773, simple_loss=0.265, pruned_loss=0.0448, over 2374523.54 frames. ], batch size: 38, lr: 4.77e-03, grad_scale: 16.0 2023-05-16 00:48:36,625 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.2859, 4.8576, 5.2551, 4.5122, 4.9291, 4.6877, 5.2687, 4.9818], device='cuda:0'), covar=tensor([0.0225, 0.0318, 0.0231, 0.0254, 0.0285, 0.0274, 0.0212, 0.0255], device='cuda:0'), in_proj_covar=tensor([0.0237, 0.0243, 0.0262, 0.0237, 0.0235, 0.0235, 0.0214, 0.0195], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 00:48:57,428 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.4301, 3.2256, 4.8428, 2.6376, 2.6654, 3.7444, 3.0435, 3.7997], device='cuda:0'), covar=tensor([0.0465, 0.1115, 0.0262, 0.1056, 0.1828, 0.1126, 0.1313, 0.0966], device='cuda:0'), in_proj_covar=tensor([0.0227, 0.0225, 0.0228, 0.0175, 0.0230, 0.0276, 0.0219, 0.0245], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:0') 2023-05-16 00:49:02,999 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.6609, 2.3750, 3.1203, 2.7941, 3.0741, 2.8552, 2.2826, 3.1835], device='cuda:0'), covar=tensor([0.0107, 0.0257, 0.0138, 0.0209, 0.0130, 0.0163, 0.0305, 0.0106], device='cuda:0'), in_proj_covar=tensor([0.0162, 0.0195, 0.0174, 0.0175, 0.0198, 0.0152, 0.0186, 0.0167], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 00:49:08,458 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=145999.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 00:49:09,417 INFO [checkpoint.py:75] (0/2) Saving checkpoint to pruned_transducer_stateless7/exp_giga_finetune/checkpoint-46000.pt 2023-05-16 00:49:14,619 INFO [finetune.py:992] (0/2) Epoch 4, batch 9300, loss[loss=0.1549, simple_loss=0.2475, pruned_loss=0.0311, over 12315.00 frames. ], tot_loss[loss=0.1774, simple_loss=0.2649, pruned_loss=0.04496, over 2375422.00 frames. ], batch size: 34, lr: 4.76e-03, grad_scale: 16.0 2023-05-16 00:49:32,619 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=146027.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 00:49:33,858 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.950e+02 2.807e+02 3.392e+02 4.099e+02 9.430e+02, threshold=6.784e+02, percent-clipped=2.0 2023-05-16 00:49:49,907 INFO [finetune.py:992] (0/2) Epoch 4, batch 9350, loss[loss=0.1626, simple_loss=0.2473, pruned_loss=0.0389, over 12084.00 frames. ], tot_loss[loss=0.1778, simple_loss=0.2654, pruned_loss=0.04513, over 2368943.85 frames. ], batch size: 32, lr: 4.76e-03, grad_scale: 16.0 2023-05-16 00:50:06,106 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=146075.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 00:50:25,258 INFO [finetune.py:992] (0/2) Epoch 4, batch 9400, loss[loss=0.1709, simple_loss=0.2558, pruned_loss=0.04297, over 12109.00 frames. ], tot_loss[loss=0.1778, simple_loss=0.2654, pruned_loss=0.04512, over 2369124.88 frames. ], batch size: 33, lr: 4.76e-03, grad_scale: 16.0 2023-05-16 00:50:25,709 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.95 vs. limit=2.0 2023-05-16 00:50:44,872 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.116e+02 2.827e+02 3.211e+02 4.021e+02 6.482e+02, threshold=6.422e+02, percent-clipped=0.0 2023-05-16 00:50:57,712 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([6.1006, 6.0587, 5.8793, 5.3697, 5.1475, 6.0143, 5.5820, 5.3877], device='cuda:0'), covar=tensor([0.0574, 0.0854, 0.0577, 0.1575, 0.0624, 0.0689, 0.1710, 0.0985], device='cuda:0'), in_proj_covar=tensor([0.0581, 0.0517, 0.0485, 0.0596, 0.0386, 0.0662, 0.0736, 0.0540], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0004, 0.0004, 0.0003], device='cuda:0') 2023-05-16 00:50:59,894 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=146149.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 00:51:01,806 INFO [finetune.py:992] (0/2) Epoch 4, batch 9450, loss[loss=0.1741, simple_loss=0.2579, pruned_loss=0.04512, over 12017.00 frames. ], tot_loss[loss=0.1776, simple_loss=0.265, pruned_loss=0.04507, over 2360286.12 frames. ], batch size: 31, lr: 4.76e-03, grad_scale: 16.0 2023-05-16 00:51:19,442 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=146177.0, num_to_drop=1, layers_to_drop={2} 2023-05-16 00:51:28,271 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.8672, 3.3475, 5.1511, 2.6364, 2.8422, 3.8919, 3.4095, 3.9925], device='cuda:0'), covar=tensor([0.0420, 0.1116, 0.0257, 0.1122, 0.1784, 0.1278, 0.1241, 0.1045], device='cuda:0'), in_proj_covar=tensor([0.0228, 0.0225, 0.0229, 0.0175, 0.0230, 0.0278, 0.0219, 0.0245], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:0') 2023-05-16 00:51:37,508 INFO [finetune.py:992] (0/2) Epoch 4, batch 9500, loss[loss=0.1797, simple_loss=0.2689, pruned_loss=0.04522, over 12340.00 frames. ], tot_loss[loss=0.1763, simple_loss=0.2639, pruned_loss=0.04439, over 2367226.31 frames. ], batch size: 36, lr: 4.76e-03, grad_scale: 16.0 2023-05-16 00:51:43,474 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=146210.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 00:51:56,949 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.030e+02 3.014e+02 3.420e+02 4.256e+02 8.963e+02, threshold=6.839e+02, percent-clipped=4.0 2023-05-16 00:52:12,039 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=146250.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 00:52:13,243 INFO [finetune.py:992] (0/2) Epoch 4, batch 9550, loss[loss=0.1787, simple_loss=0.2732, pruned_loss=0.04211, over 11858.00 frames. ], tot_loss[loss=0.176, simple_loss=0.2637, pruned_loss=0.04411, over 2369175.80 frames. ], batch size: 44, lr: 4.76e-03, grad_scale: 16.0 2023-05-16 00:52:30,378 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.33 vs. limit=2.0 2023-05-16 00:52:39,333 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.3829, 5.2729, 5.3477, 5.3819, 4.9804, 5.0534, 4.8905, 5.3499], device='cuda:0'), covar=tensor([0.0677, 0.0524, 0.0676, 0.0522, 0.1723, 0.1039, 0.0510, 0.0868], device='cuda:0'), in_proj_covar=tensor([0.0492, 0.0641, 0.0551, 0.0591, 0.0779, 0.0707, 0.0517, 0.0465], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0005, 0.0004, 0.0004, 0.0005, 0.0005, 0.0004, 0.0004], device='cuda:0') 2023-05-16 00:52:45,773 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.7955, 3.3263, 5.1067, 2.8675, 2.9530, 3.8623, 3.4046, 3.9728], device='cuda:0'), covar=tensor([0.0408, 0.1232, 0.0352, 0.1092, 0.1905, 0.1426, 0.1211, 0.1043], device='cuda:0'), in_proj_covar=tensor([0.0228, 0.0226, 0.0229, 0.0176, 0.0231, 0.0278, 0.0219, 0.0246], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:0') 2023-05-16 00:52:47,039 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=146298.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 00:52:47,919 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=146299.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 00:52:49,827 INFO [finetune.py:992] (0/2) Epoch 4, batch 9600, loss[loss=0.167, simple_loss=0.2452, pruned_loss=0.04445, over 11819.00 frames. ], tot_loss[loss=0.1758, simple_loss=0.2633, pruned_loss=0.04413, over 2372280.07 frames. ], batch size: 26, lr: 4.76e-03, grad_scale: 16.0 2023-05-16 00:53:09,055 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.946e+02 2.926e+02 3.346e+02 4.034e+02 7.441e+02, threshold=6.692e+02, percent-clipped=2.0 2023-05-16 00:53:22,006 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=146347.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 00:53:23,701 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.5008, 4.7241, 4.1823, 5.0701, 4.5393, 2.8649, 4.2445, 3.0988], device='cuda:0'), covar=tensor([0.0648, 0.0743, 0.1291, 0.0392, 0.1074, 0.1507, 0.0952, 0.3100], device='cuda:0'), in_proj_covar=tensor([0.0290, 0.0365, 0.0342, 0.0258, 0.0350, 0.0256, 0.0327, 0.0349], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 00:53:25,559 INFO [finetune.py:992] (0/2) Epoch 4, batch 9650, loss[loss=0.191, simple_loss=0.2787, pruned_loss=0.05162, over 12043.00 frames. ], tot_loss[loss=0.1768, simple_loss=0.2648, pruned_loss=0.04436, over 2379120.38 frames. ], batch size: 37, lr: 4.76e-03, grad_scale: 16.0 2023-05-16 00:53:32,611 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.2893, 5.1939, 5.2569, 5.3089, 4.8825, 4.9541, 4.8068, 5.2451], device='cuda:0'), covar=tensor([0.0677, 0.0530, 0.0707, 0.0557, 0.1875, 0.1221, 0.0532, 0.0928], device='cuda:0'), in_proj_covar=tensor([0.0490, 0.0639, 0.0550, 0.0592, 0.0778, 0.0706, 0.0515, 0.0464], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0005, 0.0004, 0.0004, 0.0005, 0.0005, 0.0004, 0.0004], device='cuda:0') 2023-05-16 00:53:53,289 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 2023-05-16 00:53:55,886 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.2154, 2.5545, 3.7543, 3.2281, 3.6242, 3.2579, 2.6378, 3.6793], device='cuda:0'), covar=tensor([0.0103, 0.0292, 0.0128, 0.0188, 0.0100, 0.0148, 0.0295, 0.0089], device='cuda:0'), in_proj_covar=tensor([0.0161, 0.0193, 0.0172, 0.0173, 0.0195, 0.0150, 0.0183, 0.0165], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 00:54:00,913 INFO [finetune.py:992] (0/2) Epoch 4, batch 9700, loss[loss=0.195, simple_loss=0.2842, pruned_loss=0.05294, over 10664.00 frames. ], tot_loss[loss=0.1779, simple_loss=0.266, pruned_loss=0.04492, over 2368955.95 frames. ], batch size: 68, lr: 4.76e-03, grad_scale: 16.0 2023-05-16 00:54:09,524 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.3450, 3.2754, 3.1522, 3.0199, 2.8216, 2.6030, 3.3057, 2.0677], device='cuda:0'), covar=tensor([0.0330, 0.0156, 0.0157, 0.0151, 0.0335, 0.0309, 0.0135, 0.0440], device='cuda:0'), in_proj_covar=tensor([0.0181, 0.0154, 0.0149, 0.0175, 0.0195, 0.0189, 0.0153, 0.0193], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], device='cuda:0') 2023-05-16 00:54:11,266 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.35 vs. limit=2.0 2023-05-16 00:54:20,667 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.186e+02 3.038e+02 3.461e+02 3.924e+02 8.082e+02, threshold=6.922e+02, percent-clipped=2.0 2023-05-16 00:54:37,536 INFO [finetune.py:992] (0/2) Epoch 4, batch 9750, loss[loss=0.1578, simple_loss=0.24, pruned_loss=0.03777, over 12339.00 frames. ], tot_loss[loss=0.1776, simple_loss=0.2656, pruned_loss=0.04474, over 2370955.56 frames. ], batch size: 30, lr: 4.76e-03, grad_scale: 16.0 2023-05-16 00:54:55,398 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=146477.0, num_to_drop=1, layers_to_drop={1} 2023-05-16 00:55:01,720 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.4033, 5.0416, 5.4007, 4.6756, 5.0334, 4.8261, 5.3542, 5.0017], device='cuda:0'), covar=tensor([0.0214, 0.0242, 0.0208, 0.0235, 0.0261, 0.0260, 0.0226, 0.0219], device='cuda:0'), in_proj_covar=tensor([0.0238, 0.0243, 0.0262, 0.0236, 0.0234, 0.0235, 0.0214, 0.0196], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 00:55:12,874 INFO [finetune.py:992] (0/2) Epoch 4, batch 9800, loss[loss=0.1697, simple_loss=0.2553, pruned_loss=0.04204, over 12118.00 frames. ], tot_loss[loss=0.1764, simple_loss=0.2644, pruned_loss=0.04424, over 2371419.45 frames. ], batch size: 30, lr: 4.76e-03, grad_scale: 16.0 2023-05-16 00:55:15,107 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=146505.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 00:55:16,054 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.57 vs. limit=2.0 2023-05-16 00:55:29,388 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=146525.0, num_to_drop=1, layers_to_drop={0} 2023-05-16 00:55:32,075 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.067e+02 2.736e+02 3.214e+02 3.887e+02 6.730e+02, threshold=6.427e+02, percent-clipped=0.0 2023-05-16 00:55:41,541 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.3521, 4.9376, 5.3589, 4.6028, 4.9926, 4.7643, 5.3452, 4.9536], device='cuda:0'), covar=tensor([0.0225, 0.0288, 0.0215, 0.0264, 0.0287, 0.0288, 0.0226, 0.0262], device='cuda:0'), in_proj_covar=tensor([0.0238, 0.0243, 0.0262, 0.0236, 0.0233, 0.0235, 0.0213, 0.0196], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 00:55:49,188 INFO [finetune.py:992] (0/2) Epoch 4, batch 9850, loss[loss=0.1648, simple_loss=0.2537, pruned_loss=0.03791, over 12310.00 frames. ], tot_loss[loss=0.1763, simple_loss=0.2642, pruned_loss=0.04423, over 2376268.74 frames. ], batch size: 34, lr: 4.76e-03, grad_scale: 16.0 2023-05-16 00:56:09,326 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.0616, 4.7204, 4.9049, 4.9258, 4.8260, 4.9384, 4.8745, 2.6609], device='cuda:0'), covar=tensor([0.0076, 0.0066, 0.0082, 0.0059, 0.0046, 0.0093, 0.0079, 0.0655], device='cuda:0'), in_proj_covar=tensor([0.0064, 0.0072, 0.0075, 0.0069, 0.0057, 0.0087, 0.0074, 0.0088], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 00:56:17,813 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=146591.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 00:56:23,451 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.4769, 4.8210, 2.9168, 2.9431, 4.0776, 2.6038, 4.1427, 3.5272], device='cuda:0'), covar=tensor([0.0571, 0.0521, 0.1091, 0.1229, 0.0294, 0.1153, 0.0419, 0.0647], device='cuda:0'), in_proj_covar=tensor([0.0184, 0.0249, 0.0173, 0.0195, 0.0138, 0.0177, 0.0192, 0.0172], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 00:56:25,565 INFO [finetune.py:992] (0/2) Epoch 4, batch 9900, loss[loss=0.2037, simple_loss=0.2919, pruned_loss=0.05773, over 12158.00 frames. ], tot_loss[loss=0.1762, simple_loss=0.2639, pruned_loss=0.0443, over 2379244.62 frames. ], batch size: 36, lr: 4.76e-03, grad_scale: 16.0 2023-05-16 00:56:39,167 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.3674, 5.2242, 5.3189, 5.3707, 4.9460, 4.9800, 4.8249, 5.3151], device='cuda:0'), covar=tensor([0.0641, 0.0523, 0.0639, 0.0482, 0.1708, 0.1283, 0.0505, 0.0831], device='cuda:0'), in_proj_covar=tensor([0.0491, 0.0638, 0.0547, 0.0589, 0.0777, 0.0703, 0.0513, 0.0464], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0005, 0.0004, 0.0004, 0.0005, 0.0005, 0.0004, 0.0004], device='cuda:0') 2023-05-16 00:56:42,709 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=146626.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 00:56:44,570 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.925e+02 2.744e+02 3.233e+02 3.737e+02 7.432e+02, threshold=6.466e+02, percent-clipped=1.0 2023-05-16 00:56:56,891 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.1626, 2.3791, 3.8000, 3.1237, 3.5271, 3.2605, 2.4988, 3.6461], device='cuda:0'), covar=tensor([0.0092, 0.0327, 0.0089, 0.0199, 0.0107, 0.0119, 0.0296, 0.0101], device='cuda:0'), in_proj_covar=tensor([0.0161, 0.0194, 0.0172, 0.0173, 0.0195, 0.0150, 0.0183, 0.0166], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 00:57:00,763 INFO [finetune.py:992] (0/2) Epoch 4, batch 9950, loss[loss=0.2227, simple_loss=0.3019, pruned_loss=0.07171, over 11408.00 frames. ], tot_loss[loss=0.1764, simple_loss=0.2642, pruned_loss=0.04435, over 2379956.95 frames. ], batch size: 55, lr: 4.76e-03, grad_scale: 16.0 2023-05-16 00:57:01,004 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=146652.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 00:57:25,752 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=146687.0, num_to_drop=1, layers_to_drop={1} 2023-05-16 00:57:36,726 INFO [finetune.py:992] (0/2) Epoch 4, batch 10000, loss[loss=0.1825, simple_loss=0.2724, pruned_loss=0.04628, over 12078.00 frames. ], tot_loss[loss=0.1769, simple_loss=0.265, pruned_loss=0.04446, over 2387522.40 frames. ], batch size: 42, lr: 4.76e-03, grad_scale: 16.0 2023-05-16 00:57:55,919 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.009e+02 2.778e+02 3.525e+02 4.163e+02 5.985e+02, threshold=7.050e+02, percent-clipped=0.0 2023-05-16 00:58:11,742 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=2.95 vs. limit=5.0 2023-05-16 00:58:12,760 INFO [finetune.py:992] (0/2) Epoch 4, batch 10050, loss[loss=0.1982, simple_loss=0.2841, pruned_loss=0.05618, over 12039.00 frames. ], tot_loss[loss=0.1769, simple_loss=0.265, pruned_loss=0.04445, over 2383044.09 frames. ], batch size: 37, lr: 4.76e-03, grad_scale: 16.0 2023-05-16 00:58:44,774 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=2.84 vs. limit=5.0 2023-05-16 00:58:48,162 INFO [finetune.py:992] (0/2) Epoch 4, batch 10100, loss[loss=0.1711, simple_loss=0.2618, pruned_loss=0.04018, over 12185.00 frames. ], tot_loss[loss=0.1771, simple_loss=0.2651, pruned_loss=0.04456, over 2379202.88 frames. ], batch size: 35, lr: 4.76e-03, grad_scale: 16.0 2023-05-16 00:58:50,478 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=146805.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 00:59:07,476 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.801e+02 2.944e+02 3.443e+02 4.252e+02 7.137e+02, threshold=6.886e+02, percent-clipped=1.0 2023-05-16 00:59:14,888 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.1440, 5.0882, 4.9045, 4.9723, 4.5647, 5.1404, 5.1197, 5.3387], device='cuda:0'), covar=tensor([0.0269, 0.0124, 0.0177, 0.0316, 0.0703, 0.0207, 0.0128, 0.0146], device='cuda:0'), in_proj_covar=tensor([0.0183, 0.0184, 0.0183, 0.0237, 0.0233, 0.0203, 0.0169, 0.0216], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0005, 0.0005, 0.0005, 0.0004, 0.0005], device='cuda:0') 2023-05-16 00:59:19,115 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=146844.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 00:59:22,791 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=4.71 vs. limit=5.0 2023-05-16 00:59:24,495 INFO [finetune.py:992] (0/2) Epoch 4, batch 10150, loss[loss=0.1568, simple_loss=0.2377, pruned_loss=0.03794, over 11998.00 frames. ], tot_loss[loss=0.176, simple_loss=0.2638, pruned_loss=0.04411, over 2382855.95 frames. ], batch size: 28, lr: 4.76e-03, grad_scale: 16.0 2023-05-16 00:59:25,253 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=146853.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 00:59:40,502 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=4.62 vs. limit=5.0 2023-05-16 00:59:42,322 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.5888, 2.5055, 3.3882, 4.3970, 2.6759, 4.4885, 4.5088, 4.6509], device='cuda:0'), covar=tensor([0.0119, 0.1141, 0.0368, 0.0147, 0.1126, 0.0190, 0.0163, 0.0072], device='cuda:0'), in_proj_covar=tensor([0.0120, 0.0204, 0.0187, 0.0119, 0.0191, 0.0176, 0.0171, 0.0122], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 01:00:00,626 INFO [finetune.py:992] (0/2) Epoch 4, batch 10200, loss[loss=0.1888, simple_loss=0.2966, pruned_loss=0.04053, over 12154.00 frames. ], tot_loss[loss=0.1746, simple_loss=0.2623, pruned_loss=0.04342, over 2385243.78 frames. ], batch size: 34, lr: 4.76e-03, grad_scale: 16.0 2023-05-16 01:00:03,026 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=146905.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 01:00:19,678 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.822e+02 2.856e+02 3.445e+02 4.302e+02 8.868e+02, threshold=6.890e+02, percent-clipped=4.0 2023-05-16 01:00:23,443 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.3206, 3.4529, 3.2341, 3.0904, 2.7739, 2.6800, 3.3731, 2.2608], device='cuda:0'), covar=tensor([0.0340, 0.0102, 0.0145, 0.0143, 0.0312, 0.0282, 0.0108, 0.0383], device='cuda:0'), in_proj_covar=tensor([0.0179, 0.0152, 0.0147, 0.0172, 0.0191, 0.0184, 0.0152, 0.0189], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], device='cuda:0') 2023-05-16 01:00:32,536 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=146947.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 01:00:33,985 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.0656, 4.6136, 4.1987, 4.3229, 4.7057, 4.1096, 4.2945, 4.1365], device='cuda:0'), covar=tensor([0.1345, 0.1117, 0.1351, 0.1883, 0.1084, 0.2094, 0.1639, 0.1493], device='cuda:0'), in_proj_covar=tensor([0.0330, 0.0456, 0.0356, 0.0407, 0.0439, 0.0412, 0.0369, 0.0359], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 01:00:36,008 INFO [finetune.py:992] (0/2) Epoch 4, batch 10250, loss[loss=0.1723, simple_loss=0.2495, pruned_loss=0.04757, over 12303.00 frames. ], tot_loss[loss=0.1755, simple_loss=0.263, pruned_loss=0.04403, over 2371651.47 frames. ], batch size: 28, lr: 4.76e-03, grad_scale: 16.0 2023-05-16 01:00:53,373 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=2.81 vs. limit=5.0 2023-05-16 01:00:57,394 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=146982.0, num_to_drop=1, layers_to_drop={3} 2023-05-16 01:01:12,375 INFO [finetune.py:992] (0/2) Epoch 4, batch 10300, loss[loss=0.1705, simple_loss=0.2677, pruned_loss=0.03661, over 12266.00 frames. ], tot_loss[loss=0.1757, simple_loss=0.2631, pruned_loss=0.04418, over 2381062.40 frames. ], batch size: 37, lr: 4.76e-03, grad_scale: 16.0 2023-05-16 01:01:29,443 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.1547, 5.0527, 4.9303, 5.0190, 4.6139, 5.1455, 5.1050, 5.3677], device='cuda:0'), covar=tensor([0.0161, 0.0116, 0.0173, 0.0251, 0.0650, 0.0223, 0.0135, 0.0125], device='cuda:0'), in_proj_covar=tensor([0.0180, 0.0182, 0.0179, 0.0233, 0.0229, 0.0199, 0.0166, 0.0213], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0005, 0.0005, 0.0005, 0.0004, 0.0005], device='cuda:0') 2023-05-16 01:01:32,081 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.042e+02 2.716e+02 3.181e+02 3.895e+02 6.567e+02, threshold=6.362e+02, percent-clipped=0.0 2023-05-16 01:01:47,376 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.60 vs. limit=2.0 2023-05-16 01:01:48,359 INFO [finetune.py:992] (0/2) Epoch 4, batch 10350, loss[loss=0.1419, simple_loss=0.2207, pruned_loss=0.03158, over 11808.00 frames. ], tot_loss[loss=0.1757, simple_loss=0.2631, pruned_loss=0.04412, over 2379221.80 frames. ], batch size: 26, lr: 4.76e-03, grad_scale: 16.0 2023-05-16 01:02:22,597 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.1584, 3.8997, 4.1736, 4.3769, 3.0838, 3.9536, 2.7995, 4.0507], device='cuda:0'), covar=tensor([0.1571, 0.0646, 0.0746, 0.0552, 0.1014, 0.0540, 0.1531, 0.1316], device='cuda:0'), in_proj_covar=tensor([0.0221, 0.0254, 0.0285, 0.0339, 0.0232, 0.0233, 0.0249, 0.0373], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 01:02:23,871 INFO [finetune.py:992] (0/2) Epoch 4, batch 10400, loss[loss=0.1744, simple_loss=0.2663, pruned_loss=0.04127, over 12122.00 frames. ], tot_loss[loss=0.1755, simple_loss=0.2629, pruned_loss=0.04402, over 2372263.30 frames. ], batch size: 45, lr: 4.75e-03, grad_scale: 16.0 2023-05-16 01:02:43,171 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.124e+02 2.939e+02 3.483e+02 4.123e+02 8.452e+02, threshold=6.965e+02, percent-clipped=4.0 2023-05-16 01:03:00,258 INFO [finetune.py:992] (0/2) Epoch 4, batch 10450, loss[loss=0.1825, simple_loss=0.2704, pruned_loss=0.04734, over 12274.00 frames. ], tot_loss[loss=0.1753, simple_loss=0.2629, pruned_loss=0.04384, over 2378551.27 frames. ], batch size: 37, lr: 4.75e-03, grad_scale: 16.0 2023-05-16 01:03:14,552 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.3543, 2.1395, 3.2451, 4.1305, 1.9903, 4.2347, 4.3403, 4.4481], device='cuda:0'), covar=tensor([0.0176, 0.1328, 0.0425, 0.0168, 0.1414, 0.0275, 0.0143, 0.0086], device='cuda:0'), in_proj_covar=tensor([0.0118, 0.0201, 0.0185, 0.0118, 0.0188, 0.0174, 0.0168, 0.0121], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 01:03:35,100 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=147200.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 01:03:36,374 INFO [finetune.py:992] (0/2) Epoch 4, batch 10500, loss[loss=0.1543, simple_loss=0.2357, pruned_loss=0.03648, over 12031.00 frames. ], tot_loss[loss=0.1762, simple_loss=0.2639, pruned_loss=0.04424, over 2377239.76 frames. ], batch size: 28, lr: 4.75e-03, grad_scale: 16.0 2023-05-16 01:03:46,967 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.9400, 4.9804, 4.7812, 4.8520, 4.4369, 5.0077, 4.9699, 5.2516], device='cuda:0'), covar=tensor([0.0239, 0.0137, 0.0200, 0.0301, 0.0774, 0.0327, 0.0151, 0.0163], device='cuda:0'), in_proj_covar=tensor([0.0180, 0.0182, 0.0180, 0.0233, 0.0229, 0.0199, 0.0166, 0.0214], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0005, 0.0005, 0.0005, 0.0004, 0.0005], device='cuda:0') 2023-05-16 01:03:55,489 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.857e+02 3.015e+02 3.764e+02 4.290e+02 1.266e+03, threshold=7.528e+02, percent-clipped=2.0 2023-05-16 01:04:04,304 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=147241.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 01:04:08,491 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=147247.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 01:04:08,513 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.3938, 2.3775, 3.4909, 4.4129, 3.7391, 4.2538, 3.8153, 3.0705], device='cuda:0'), covar=tensor([0.0030, 0.0366, 0.0161, 0.0031, 0.0118, 0.0072, 0.0102, 0.0311], device='cuda:0'), in_proj_covar=tensor([0.0083, 0.0122, 0.0104, 0.0075, 0.0101, 0.0112, 0.0090, 0.0137], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 01:04:11,828 INFO [finetune.py:992] (0/2) Epoch 4, batch 10550, loss[loss=0.2245, simple_loss=0.3073, pruned_loss=0.07087, over 12036.00 frames. ], tot_loss[loss=0.1761, simple_loss=0.2637, pruned_loss=0.04429, over 2374076.83 frames. ], batch size: 40, lr: 4.75e-03, grad_scale: 16.0 2023-05-16 01:04:13,628 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=2.49 vs. limit=5.0 2023-05-16 01:04:15,145 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=2.99 vs. limit=5.0 2023-05-16 01:04:33,452 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=147282.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 01:04:35,900 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.36 vs. limit=2.0 2023-05-16 01:04:43,294 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=147295.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 01:04:48,322 INFO [finetune.py:992] (0/2) Epoch 4, batch 10600, loss[loss=0.2525, simple_loss=0.3313, pruned_loss=0.08689, over 7728.00 frames. ], tot_loss[loss=0.1757, simple_loss=0.2633, pruned_loss=0.04399, over 2377078.15 frames. ], batch size: 99, lr: 4.75e-03, grad_scale: 16.0 2023-05-16 01:04:48,505 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=147302.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 01:04:50,459 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.40 vs. limit=2.0 2023-05-16 01:04:54,252 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=147310.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 01:05:08,437 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.912e+02 2.841e+02 3.295e+02 4.134e+02 6.793e+02, threshold=6.590e+02, percent-clipped=0.0 2023-05-16 01:05:09,209 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=147330.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 01:05:24,645 INFO [finetune.py:992] (0/2) Epoch 4, batch 10650, loss[loss=0.18, simple_loss=0.2694, pruned_loss=0.04528, over 12310.00 frames. ], tot_loss[loss=0.1763, simple_loss=0.2641, pruned_loss=0.04423, over 2368120.91 frames. ], batch size: 34, lr: 4.75e-03, grad_scale: 16.0 2023-05-16 01:05:26,402 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.0132, 2.3430, 3.4651, 3.0637, 3.3385, 3.1582, 2.5449, 3.3937], device='cuda:0'), covar=tensor([0.0094, 0.0292, 0.0125, 0.0185, 0.0122, 0.0128, 0.0282, 0.0114], device='cuda:0'), in_proj_covar=tensor([0.0161, 0.0193, 0.0173, 0.0174, 0.0196, 0.0151, 0.0185, 0.0166], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 01:05:38,501 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=147371.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 01:05:54,281 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2023-05-16 01:06:00,679 INFO [finetune.py:992] (0/2) Epoch 4, batch 10700, loss[loss=0.1718, simple_loss=0.2641, pruned_loss=0.03978, over 11320.00 frames. ], tot_loss[loss=0.176, simple_loss=0.264, pruned_loss=0.04405, over 2371780.10 frames. ], batch size: 55, lr: 4.75e-03, grad_scale: 32.0 2023-05-16 01:06:04,527 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=147407.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 01:06:20,461 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.887e+02 2.905e+02 3.370e+02 4.057e+02 8.796e+02, threshold=6.739e+02, percent-clipped=3.0 2023-05-16 01:06:24,196 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=147434.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 01:06:36,933 INFO [finetune.py:992] (0/2) Epoch 4, batch 10750, loss[loss=0.1736, simple_loss=0.2622, pruned_loss=0.04255, over 12096.00 frames. ], tot_loss[loss=0.1763, simple_loss=0.2641, pruned_loss=0.04423, over 2368581.72 frames. ], batch size: 32, lr: 4.75e-03, grad_scale: 32.0 2023-05-16 01:06:48,857 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=147468.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 01:07:08,025 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=147495.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 01:07:11,584 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=147500.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 01:07:12,877 INFO [finetune.py:992] (0/2) Epoch 4, batch 10800, loss[loss=0.1665, simple_loss=0.2636, pruned_loss=0.03465, over 12195.00 frames. ], tot_loss[loss=0.1761, simple_loss=0.2641, pruned_loss=0.04406, over 2371896.83 frames. ], batch size: 35, lr: 4.75e-03, grad_scale: 16.0 2023-05-16 01:07:32,863 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.218e+02 2.861e+02 3.389e+02 3.924e+02 8.164e+02, threshold=6.777e+02, percent-clipped=1.0 2023-05-16 01:07:45,601 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=147548.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 01:07:48,403 INFO [finetune.py:992] (0/2) Epoch 4, batch 10850, loss[loss=0.1806, simple_loss=0.2695, pruned_loss=0.04586, over 11179.00 frames. ], tot_loss[loss=0.1771, simple_loss=0.265, pruned_loss=0.0446, over 2363232.36 frames. ], batch size: 55, lr: 4.75e-03, grad_scale: 16.0 2023-05-16 01:08:22,347 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=147597.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 01:08:26,818 INFO [finetune.py:992] (0/2) Epoch 4, batch 10900, loss[loss=0.1728, simple_loss=0.2586, pruned_loss=0.04344, over 12342.00 frames. ], tot_loss[loss=0.1773, simple_loss=0.265, pruned_loss=0.04485, over 2364044.36 frames. ], batch size: 36, lr: 4.75e-03, grad_scale: 16.0 2023-05-16 01:08:46,843 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.022e+02 2.788e+02 3.418e+02 4.406e+02 8.127e+02, threshold=6.835e+02, percent-clipped=5.0 2023-05-16 01:09:02,426 INFO [finetune.py:992] (0/2) Epoch 4, batch 10950, loss[loss=0.244, simple_loss=0.3333, pruned_loss=0.07736, over 11043.00 frames. ], tot_loss[loss=0.1784, simple_loss=0.2659, pruned_loss=0.04546, over 2365816.43 frames. ], batch size: 55, lr: 4.75e-03, grad_scale: 16.0 2023-05-16 01:09:07,623 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=147659.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 01:09:12,560 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=147666.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 01:09:25,273 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2023-05-16 01:09:33,575 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.4307, 3.4725, 3.3161, 3.1688, 2.9087, 2.6870, 3.5612, 2.1514], device='cuda:0'), covar=tensor([0.0341, 0.0164, 0.0149, 0.0177, 0.0343, 0.0341, 0.0115, 0.0454], device='cuda:0'), in_proj_covar=tensor([0.0183, 0.0156, 0.0150, 0.0177, 0.0195, 0.0190, 0.0155, 0.0193], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], device='cuda:0') 2023-05-16 01:09:38,152 INFO [finetune.py:992] (0/2) Epoch 4, batch 11000, loss[loss=0.2239, simple_loss=0.2948, pruned_loss=0.07647, over 8239.00 frames. ], tot_loss[loss=0.1807, simple_loss=0.2678, pruned_loss=0.04682, over 2343437.51 frames. ], batch size: 98, lr: 4.75e-03, grad_scale: 8.0 2023-05-16 01:09:44,836 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([6.0633, 6.0432, 5.8260, 5.4432, 5.2589, 5.9739, 5.5471, 5.4367], device='cuda:0'), covar=tensor([0.0612, 0.0873, 0.0589, 0.1332, 0.0563, 0.0606, 0.1323, 0.0807], device='cuda:0'), in_proj_covar=tensor([0.0574, 0.0515, 0.0481, 0.0592, 0.0381, 0.0660, 0.0734, 0.0538], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0004, 0.0004, 0.0003], device='cuda:0') 2023-05-16 01:09:51,093 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=147720.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 01:09:59,216 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.248e+02 3.094e+02 3.686e+02 4.339e+02 1.027e+03, threshold=7.372e+02, percent-clipped=4.0 2023-05-16 01:10:00,288 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.1900, 2.4676, 3.6910, 3.1442, 3.5538, 3.1552, 2.6154, 3.6785], device='cuda:0'), covar=tensor([0.0118, 0.0349, 0.0129, 0.0225, 0.0149, 0.0185, 0.0334, 0.0099], device='cuda:0'), in_proj_covar=tensor([0.0161, 0.0193, 0.0172, 0.0174, 0.0195, 0.0150, 0.0183, 0.0165], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 01:10:14,136 INFO [finetune.py:992] (0/2) Epoch 4, batch 11050, loss[loss=0.186, simple_loss=0.2788, pruned_loss=0.04661, over 12340.00 frames. ], tot_loss[loss=0.1844, simple_loss=0.2709, pruned_loss=0.04896, over 2314630.30 frames. ], batch size: 36, lr: 4.75e-03, grad_scale: 8.0 2023-05-16 01:10:21,780 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=147763.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 01:10:39,527 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.35 vs. limit=2.0 2023-05-16 01:10:40,574 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=147790.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 01:10:43,366 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.3859, 4.9510, 5.3459, 4.6286, 5.0431, 4.7751, 5.3557, 4.9879], device='cuda:0'), covar=tensor([0.0256, 0.0338, 0.0260, 0.0261, 0.0311, 0.0285, 0.0210, 0.0245], device='cuda:0'), in_proj_covar=tensor([0.0237, 0.0242, 0.0261, 0.0234, 0.0232, 0.0233, 0.0210, 0.0195], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 01:10:49,709 INFO [finetune.py:992] (0/2) Epoch 4, batch 11100, loss[loss=0.1862, simple_loss=0.2695, pruned_loss=0.05146, over 12143.00 frames. ], tot_loss[loss=0.1869, simple_loss=0.2729, pruned_loss=0.05047, over 2277833.21 frames. ], batch size: 30, lr: 4.75e-03, grad_scale: 8.0 2023-05-16 01:10:50,235 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.43 vs. limit=2.0 2023-05-16 01:10:58,821 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=147815.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 01:11:04,742 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2023-05-16 01:11:10,611 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.150e+02 3.432e+02 4.091e+02 4.867e+02 8.600e+02, threshold=8.181e+02, percent-clipped=2.0 2023-05-16 01:11:25,141 INFO [finetune.py:992] (0/2) Epoch 4, batch 11150, loss[loss=0.1831, simple_loss=0.2779, pruned_loss=0.04411, over 12346.00 frames. ], tot_loss[loss=0.1946, simple_loss=0.2796, pruned_loss=0.05484, over 2219653.60 frames. ], batch size: 36, lr: 4.75e-03, grad_scale: 8.0 2023-05-16 01:11:28,839 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=147857.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 01:11:41,944 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=147876.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 01:11:57,299 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=147897.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 01:12:00,317 INFO [finetune.py:992] (0/2) Epoch 4, batch 11200, loss[loss=0.3318, simple_loss=0.385, pruned_loss=0.1393, over 7156.00 frames. ], tot_loss[loss=0.2009, simple_loss=0.2851, pruned_loss=0.05836, over 2185536.02 frames. ], batch size: 98, lr: 4.75e-03, grad_scale: 8.0 2023-05-16 01:12:07,385 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=147911.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 01:12:12,153 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=147918.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 01:12:20,930 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.360e+02 3.735e+02 4.518e+02 5.395e+02 1.112e+03, threshold=9.037e+02, percent-clipped=6.0 2023-05-16 01:12:30,868 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=147945.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 01:12:35,694 INFO [finetune.py:992] (0/2) Epoch 4, batch 11250, loss[loss=0.2794, simple_loss=0.3419, pruned_loss=0.1085, over 7096.00 frames. ], tot_loss[loss=0.2084, simple_loss=0.2912, pruned_loss=0.06277, over 2124422.39 frames. ], batch size: 98, lr: 4.75e-03, grad_scale: 8.0 2023-05-16 01:12:46,201 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=147966.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 01:12:50,096 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.9312, 5.6430, 5.2318, 5.2731, 5.7387, 5.1110, 5.2424, 5.2689], device='cuda:0'), covar=tensor([0.1179, 0.0904, 0.0974, 0.1626, 0.0802, 0.2110, 0.1575, 0.1073], device='cuda:0'), in_proj_covar=tensor([0.0325, 0.0448, 0.0351, 0.0398, 0.0431, 0.0405, 0.0365, 0.0351], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 01:12:50,203 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=147972.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 01:13:09,690 INFO [checkpoint.py:75] (0/2) Saving checkpoint to pruned_transducer_stateless7/exp_giga_finetune/checkpoint-48000.pt 2023-05-16 01:13:14,697 INFO [finetune.py:992] (0/2) Epoch 4, batch 11300, loss[loss=0.1651, simple_loss=0.2614, pruned_loss=0.03435, over 12154.00 frames. ], tot_loss[loss=0.2168, simple_loss=0.2986, pruned_loss=0.06749, over 2073389.67 frames. ], batch size: 36, lr: 4.75e-03, grad_scale: 8.0 2023-05-16 01:13:18,795 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=4.75 vs. limit=5.0 2023-05-16 01:13:23,149 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=148014.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 01:13:23,929 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=148015.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 01:13:35,232 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.605e+02 3.606e+02 4.302e+02 4.951e+02 7.972e+02, threshold=8.603e+02, percent-clipped=0.0 2023-05-16 01:13:50,025 INFO [finetune.py:992] (0/2) Epoch 4, batch 11350, loss[loss=0.2353, simple_loss=0.3126, pruned_loss=0.079, over 11552.00 frames. ], tot_loss[loss=0.2231, simple_loss=0.3039, pruned_loss=0.07119, over 2017082.04 frames. ], batch size: 48, lr: 4.75e-03, grad_scale: 8.0 2023-05-16 01:13:57,621 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=148063.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 01:14:06,534 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.0467, 2.0927, 2.6348, 3.0522, 2.2088, 3.0796, 3.0975, 3.1491], device='cuda:0'), covar=tensor([0.0147, 0.0990, 0.0386, 0.0144, 0.0949, 0.0273, 0.0251, 0.0116], device='cuda:0'), in_proj_covar=tensor([0.0116, 0.0199, 0.0183, 0.0115, 0.0187, 0.0171, 0.0163, 0.0119], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 01:14:16,774 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=148090.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 01:14:24,637 INFO [finetune.py:992] (0/2) Epoch 4, batch 11400, loss[loss=0.2801, simple_loss=0.345, pruned_loss=0.1076, over 6902.00 frames. ], tot_loss[loss=0.2274, simple_loss=0.3072, pruned_loss=0.07384, over 1954470.69 frames. ], batch size: 99, lr: 4.75e-03, grad_scale: 8.0 2023-05-16 01:14:26,560 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=4.66 vs. limit=5.0 2023-05-16 01:14:31,564 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=148111.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 01:14:44,859 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.328e+02 3.768e+02 4.287e+02 4.920e+02 1.036e+03, threshold=8.574e+02, percent-clipped=2.0 2023-05-16 01:14:45,080 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.7751, 2.0728, 2.9336, 3.7515, 2.1892, 3.8169, 3.7964, 3.8496], device='cuda:0'), covar=tensor([0.0130, 0.1307, 0.0426, 0.0122, 0.1250, 0.0198, 0.0190, 0.0102], device='cuda:0'), in_proj_covar=tensor([0.0115, 0.0198, 0.0182, 0.0114, 0.0185, 0.0170, 0.0161, 0.0118], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 01:14:49,562 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=148138.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 01:14:59,573 INFO [finetune.py:992] (0/2) Epoch 4, batch 11450, loss[loss=0.2759, simple_loss=0.351, pruned_loss=0.1004, over 10421.00 frames. ], tot_loss[loss=0.2339, simple_loss=0.3125, pruned_loss=0.07766, over 1907891.73 frames. ], batch size: 68, lr: 4.74e-03, grad_scale: 8.0 2023-05-16 01:15:12,450 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=148171.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 01:15:22,947 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2023-05-16 01:15:34,158 INFO [finetune.py:992] (0/2) Epoch 4, batch 11500, loss[loss=0.2933, simple_loss=0.3492, pruned_loss=0.1187, over 6497.00 frames. ], tot_loss[loss=0.2381, simple_loss=0.3158, pruned_loss=0.08022, over 1870292.93 frames. ], batch size: 99, lr: 4.74e-03, grad_scale: 8.0 2023-05-16 01:15:42,489 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=148213.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 01:15:46,197 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.33 vs. limit=2.0 2023-05-16 01:15:49,887 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.9563, 2.1590, 3.1855, 2.7510, 3.1207, 2.9188, 2.1561, 3.1821], device='cuda:0'), covar=tensor([0.0097, 0.0345, 0.0079, 0.0215, 0.0104, 0.0143, 0.0356, 0.0078], device='cuda:0'), in_proj_covar=tensor([0.0151, 0.0184, 0.0160, 0.0163, 0.0183, 0.0140, 0.0174, 0.0155], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-16 01:15:54,174 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.9691, 2.1243, 2.5757, 2.9829, 2.2243, 2.9874, 3.0335, 3.0901], device='cuda:0'), covar=tensor([0.0158, 0.0924, 0.0382, 0.0155, 0.0880, 0.0276, 0.0232, 0.0122], device='cuda:0'), in_proj_covar=tensor([0.0113, 0.0195, 0.0179, 0.0112, 0.0183, 0.0167, 0.0159, 0.0116], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 01:15:54,605 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.676e+02 3.747e+02 4.510e+02 5.605e+02 1.553e+03, threshold=9.020e+02, percent-clipped=3.0 2023-05-16 01:16:02,015 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([1.9794, 2.2128, 2.3017, 2.2309, 2.0531, 1.9082, 2.0662, 1.6469], device='cuda:0'), covar=tensor([0.0262, 0.0128, 0.0114, 0.0165, 0.0204, 0.0176, 0.0122, 0.0325], device='cuda:0'), in_proj_covar=tensor([0.0178, 0.0149, 0.0145, 0.0172, 0.0189, 0.0185, 0.0149, 0.0190], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], device='cuda:0') 2023-05-16 01:16:09,820 INFO [finetune.py:992] (0/2) Epoch 4, batch 11550, loss[loss=0.2023, simple_loss=0.2866, pruned_loss=0.05901, over 10263.00 frames. ], tot_loss[loss=0.2419, simple_loss=0.3184, pruned_loss=0.08272, over 1832777.55 frames. ], batch size: 69, lr: 4.74e-03, grad_scale: 8.0 2023-05-16 01:16:11,960 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.4354, 3.0778, 3.2062, 3.3771, 2.8234, 3.1106, 2.5882, 3.0018], device='cuda:0'), covar=tensor([0.1512, 0.0840, 0.1004, 0.0549, 0.0950, 0.0791, 0.1640, 0.0905], device='cuda:0'), in_proj_covar=tensor([0.0226, 0.0262, 0.0291, 0.0344, 0.0236, 0.0238, 0.0257, 0.0378], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 01:16:19,911 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=148267.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 01:16:33,947 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=148287.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 01:16:35,217 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.7436, 3.1974, 2.3571, 2.1862, 2.8015, 2.2275, 3.0429, 2.5456], device='cuda:0'), covar=tensor([0.0517, 0.0449, 0.0870, 0.1409, 0.0244, 0.1166, 0.0399, 0.0786], device='cuda:0'), in_proj_covar=tensor([0.0177, 0.0233, 0.0167, 0.0193, 0.0132, 0.0174, 0.0183, 0.0167], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 01:16:43,226 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.7239, 3.6601, 3.6322, 3.7215, 3.5063, 3.7846, 3.8105, 3.8685], device='cuda:0'), covar=tensor([0.0179, 0.0149, 0.0159, 0.0273, 0.0450, 0.0240, 0.0145, 0.0194], device='cuda:0'), in_proj_covar=tensor([0.0161, 0.0162, 0.0161, 0.0207, 0.0205, 0.0176, 0.0149, 0.0194], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:0') 2023-05-16 01:16:44,358 INFO [finetune.py:992] (0/2) Epoch 4, batch 11600, loss[loss=0.2209, simple_loss=0.3067, pruned_loss=0.06756, over 10536.00 frames. ], tot_loss[loss=0.2454, simple_loss=0.3207, pruned_loss=0.08502, over 1805205.93 frames. ], batch size: 69, lr: 4.74e-03, grad_scale: 8.0 2023-05-16 01:16:53,076 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=148315.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 01:16:56,479 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.0609, 1.8800, 2.2511, 2.0226, 2.1661, 2.2826, 1.7802, 2.2178], device='cuda:0'), covar=tensor([0.0082, 0.0257, 0.0113, 0.0179, 0.0121, 0.0121, 0.0251, 0.0098], device='cuda:0'), in_proj_covar=tensor([0.0150, 0.0182, 0.0158, 0.0161, 0.0181, 0.0139, 0.0173, 0.0153], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-16 01:16:58,756 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=4.28 vs. limit=5.0 2023-05-16 01:16:59,913 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.5446, 4.2877, 4.3967, 4.4188, 4.2331, 4.5176, 4.4156, 2.4433], device='cuda:0'), covar=tensor([0.0125, 0.0081, 0.0113, 0.0081, 0.0075, 0.0122, 0.0095, 0.0983], device='cuda:0'), in_proj_covar=tensor([0.0061, 0.0069, 0.0072, 0.0066, 0.0054, 0.0083, 0.0071, 0.0087], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-16 01:17:04,080 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.748e+02 3.470e+02 3.977e+02 4.660e+02 1.039e+03, threshold=7.954e+02, percent-clipped=1.0 2023-05-16 01:17:17,662 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=148348.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 01:17:19,602 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.92 vs. limit=2.0 2023-05-16 01:17:20,374 INFO [finetune.py:992] (0/2) Epoch 4, batch 11650, loss[loss=0.2499, simple_loss=0.3139, pruned_loss=0.0929, over 6659.00 frames. ], tot_loss[loss=0.245, simple_loss=0.3199, pruned_loss=0.08504, over 1784156.83 frames. ], batch size: 100, lr: 4.74e-03, grad_scale: 8.0 2023-05-16 01:17:28,885 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=148363.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 01:17:51,650 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.7333, 2.3020, 2.9867, 3.7022, 2.2030, 3.7629, 3.7800, 3.8403], device='cuda:0'), covar=tensor([0.0130, 0.1109, 0.0375, 0.0134, 0.1245, 0.0178, 0.0186, 0.0080], device='cuda:0'), in_proj_covar=tensor([0.0112, 0.0192, 0.0176, 0.0110, 0.0179, 0.0163, 0.0155, 0.0114], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 01:17:55,683 INFO [finetune.py:992] (0/2) Epoch 4, batch 11700, loss[loss=0.2239, simple_loss=0.3006, pruned_loss=0.07356, over 10104.00 frames. ], tot_loss[loss=0.2455, simple_loss=0.3196, pruned_loss=0.08568, over 1755632.70 frames. ], batch size: 68, lr: 4.74e-03, grad_scale: 8.0 2023-05-16 01:17:58,056 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 2023-05-16 01:18:14,542 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=148429.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 01:18:15,797 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.380e+02 3.656e+02 4.264e+02 4.955e+02 6.990e+02, threshold=8.528e+02, percent-clipped=1.0 2023-05-16 01:18:29,859 INFO [finetune.py:992] (0/2) Epoch 4, batch 11750, loss[loss=0.2694, simple_loss=0.3328, pruned_loss=0.103, over 6600.00 frames. ], tot_loss[loss=0.2468, simple_loss=0.3199, pruned_loss=0.08679, over 1724609.93 frames. ], batch size: 99, lr: 4.74e-03, grad_scale: 8.0 2023-05-16 01:18:43,465 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=148471.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 01:18:56,982 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=148490.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 01:18:58,369 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.5776, 3.2516, 3.4868, 3.5657, 3.5007, 3.6012, 3.4758, 2.7225], device='cuda:0'), covar=tensor([0.0092, 0.0096, 0.0117, 0.0082, 0.0070, 0.0111, 0.0101, 0.0624], device='cuda:0'), in_proj_covar=tensor([0.0060, 0.0067, 0.0071, 0.0065, 0.0053, 0.0081, 0.0069, 0.0086], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-16 01:19:02,408 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=148498.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 01:19:04,982 INFO [finetune.py:992] (0/2) Epoch 4, batch 11800, loss[loss=0.2692, simple_loss=0.3296, pruned_loss=0.1044, over 6941.00 frames. ], tot_loss[loss=0.2493, simple_loss=0.322, pruned_loss=0.08827, over 1715028.79 frames. ], batch size: 104, lr: 4.74e-03, grad_scale: 8.0 2023-05-16 01:19:12,732 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=148513.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 01:19:17,496 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=148519.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 01:19:25,325 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.324e+02 3.595e+02 4.321e+02 4.889e+02 1.131e+03, threshold=8.643e+02, percent-clipped=3.0 2023-05-16 01:19:40,055 INFO [finetune.py:992] (0/2) Epoch 4, batch 11850, loss[loss=0.2245, simple_loss=0.3137, pruned_loss=0.06763, over 10401.00 frames. ], tot_loss[loss=0.2513, simple_loss=0.3241, pruned_loss=0.08923, over 1693138.84 frames. ], batch size: 69, lr: 4.74e-03, grad_scale: 8.0 2023-05-16 01:19:44,974 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=148559.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 01:19:46,253 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=148561.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 01:19:50,428 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=148567.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 01:20:14,989 INFO [finetune.py:992] (0/2) Epoch 4, batch 11900, loss[loss=0.2948, simple_loss=0.346, pruned_loss=0.1218, over 6818.00 frames. ], tot_loss[loss=0.2509, simple_loss=0.3243, pruned_loss=0.08876, over 1672327.27 frames. ], batch size: 100, lr: 4.74e-03, grad_scale: 8.0 2023-05-16 01:20:23,806 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=148615.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 01:20:34,368 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.445e+02 3.443e+02 3.857e+02 4.683e+02 7.167e+02, threshold=7.714e+02, percent-clipped=0.0 2023-05-16 01:20:43,191 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=148643.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 01:20:48,528 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=148651.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 01:20:48,967 INFO [finetune.py:992] (0/2) Epoch 4, batch 11950, loss[loss=0.2251, simple_loss=0.2985, pruned_loss=0.07585, over 7165.00 frames. ], tot_loss[loss=0.2462, simple_loss=0.3208, pruned_loss=0.08577, over 1671201.72 frames. ], batch size: 98, lr: 4.74e-03, grad_scale: 8.0 2023-05-16 01:21:24,250 INFO [finetune.py:992] (0/2) Epoch 4, batch 12000, loss[loss=0.2211, simple_loss=0.3026, pruned_loss=0.06986, over 10513.00 frames. ], tot_loss[loss=0.2392, simple_loss=0.3156, pruned_loss=0.08134, over 1678779.20 frames. ], batch size: 70, lr: 4.74e-03, grad_scale: 8.0 2023-05-16 01:21:24,251 INFO [finetune.py:1017] (0/2) Computing validation loss 2023-05-16 01:21:37,899 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.6737, 5.6218, 5.5807, 5.0407, 5.0270, 5.6304, 5.2768, 5.2520], device='cuda:0'), covar=tensor([0.0492, 0.0866, 0.0480, 0.1309, 0.0421, 0.0510, 0.1068, 0.0716], device='cuda:0'), in_proj_covar=tensor([0.0535, 0.0479, 0.0448, 0.0549, 0.0355, 0.0607, 0.0663, 0.0494], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 01:21:38,333 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.1295, 4.3193, 4.0821, 4.9679, 4.4657, 2.7398, 4.2294, 2.9728], device='cuda:0'), covar=tensor([0.0886, 0.0998, 0.1196, 0.0287, 0.1339, 0.1959, 0.1106, 0.3930], device='cuda:0'), in_proj_covar=tensor([0.0274, 0.0342, 0.0321, 0.0235, 0.0330, 0.0247, 0.0310, 0.0334], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-16 01:21:38,866 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.1103, 3.2269, 4.6724, 2.3575, 2.6030, 3.5786, 2.9078, 3.6196], device='cuda:0'), covar=tensor([0.0556, 0.1219, 0.0163, 0.1464, 0.2181, 0.1408, 0.1705, 0.1131], device='cuda:0'), in_proj_covar=tensor([0.0213, 0.0216, 0.0210, 0.0168, 0.0219, 0.0260, 0.0209, 0.0227], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:0') 2023-05-16 01:21:43,114 INFO [finetune.py:1026] (0/2) Epoch 4, validation: loss=0.2922, simple_loss=0.3681, pruned_loss=0.1082, over 1020973.00 frames. 2023-05-16 01:21:43,115 INFO [finetune.py:1027] (0/2) Maximum memory allocated so far is 12745MB 2023-05-16 01:21:49,953 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=148712.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 01:21:52,596 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=148716.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 01:22:02,471 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.183e+02 3.126e+02 3.672e+02 4.633e+02 1.109e+03, threshold=7.344e+02, percent-clipped=3.0 2023-05-16 01:22:17,273 INFO [finetune.py:992] (0/2) Epoch 4, batch 12050, loss[loss=0.2536, simple_loss=0.3242, pruned_loss=0.0915, over 7037.00 frames. ], tot_loss[loss=0.2326, simple_loss=0.3105, pruned_loss=0.07732, over 1680044.06 frames. ], batch size: 102, lr: 4.74e-03, grad_scale: 8.0 2023-05-16 01:22:25,532 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.4455, 3.1772, 3.2211, 3.4583, 2.8096, 3.0736, 2.6689, 2.9201], device='cuda:0'), covar=tensor([0.1516, 0.0817, 0.1010, 0.0763, 0.0979, 0.0755, 0.1614, 0.1127], device='cuda:0'), in_proj_covar=tensor([0.0223, 0.0257, 0.0284, 0.0336, 0.0231, 0.0234, 0.0253, 0.0368], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 01:22:34,163 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=148777.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 01:22:39,031 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=148785.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 01:22:50,065 INFO [finetune.py:992] (0/2) Epoch 4, batch 12100, loss[loss=0.2322, simple_loss=0.3081, pruned_loss=0.07815, over 6598.00 frames. ], tot_loss[loss=0.2305, simple_loss=0.3089, pruned_loss=0.07606, over 1678161.57 frames. ], batch size: 97, lr: 4.74e-03, grad_scale: 8.0 2023-05-16 01:22:52,113 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.1445, 5.1339, 4.9830, 5.0777, 4.6788, 5.1413, 5.2016, 5.2832], device='cuda:0'), covar=tensor([0.0185, 0.0106, 0.0144, 0.0212, 0.0508, 0.0247, 0.0144, 0.0159], device='cuda:0'), in_proj_covar=tensor([0.0152, 0.0152, 0.0152, 0.0193, 0.0192, 0.0167, 0.0141, 0.0185], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:0') 2023-05-16 01:23:08,411 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.936e+02 3.069e+02 3.619e+02 4.456e+02 7.686e+02, threshold=7.237e+02, percent-clipped=2.0 2023-05-16 01:23:18,861 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.7096, 2.8191, 4.2114, 4.4961, 3.0132, 2.6639, 2.9400, 1.9157], device='cuda:0'), covar=tensor([0.1341, 0.2597, 0.0433, 0.0312, 0.1024, 0.2062, 0.2346, 0.4212], device='cuda:0'), in_proj_covar=tensor([0.0262, 0.0353, 0.0250, 0.0269, 0.0239, 0.0268, 0.0338, 0.0342], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-16 01:23:21,827 INFO [finetune.py:992] (0/2) Epoch 4, batch 12150, loss[loss=0.2465, simple_loss=0.3266, pruned_loss=0.08315, over 10062.00 frames. ], tot_loss[loss=0.2318, simple_loss=0.31, pruned_loss=0.07681, over 1678516.94 frames. ], batch size: 68, lr: 4.74e-03, grad_scale: 8.0 2023-05-16 01:23:23,190 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=148854.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 01:23:36,630 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.1743, 3.8715, 3.9106, 4.3433, 2.9386, 3.7664, 2.4796, 3.6495], device='cuda:0'), covar=tensor([0.2186, 0.0931, 0.1063, 0.0602, 0.1434, 0.0800, 0.2475, 0.1282], device='cuda:0'), in_proj_covar=tensor([0.0220, 0.0252, 0.0279, 0.0329, 0.0227, 0.0230, 0.0247, 0.0360], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 01:23:53,743 INFO [finetune.py:992] (0/2) Epoch 4, batch 12200, loss[loss=0.2524, simple_loss=0.3231, pruned_loss=0.09085, over 6647.00 frames. ], tot_loss[loss=0.2336, simple_loss=0.311, pruned_loss=0.07809, over 1649793.20 frames. ], batch size: 100, lr: 4.74e-03, grad_scale: 8.0 2023-05-16 01:24:11,579 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.560e+02 3.330e+02 3.899e+02 4.842e+02 8.998e+02, threshold=7.799e+02, percent-clipped=1.0 2023-05-16 01:24:15,070 INFO [checkpoint.py:75] (0/2) Saving checkpoint to pruned_transducer_stateless7/exp_giga_finetune/epoch-4.pt 2023-05-16 01:24:39,470 INFO [finetune.py:992] (0/2) Epoch 5, batch 0, loss[loss=0.2154, simple_loss=0.3035, pruned_loss=0.06363, over 12130.00 frames. ], tot_loss[loss=0.2154, simple_loss=0.3035, pruned_loss=0.06363, over 12130.00 frames. ], batch size: 39, lr: 4.74e-03, grad_scale: 8.0 2023-05-16 01:24:39,471 INFO [finetune.py:1017] (0/2) Computing validation loss 2023-05-16 01:24:57,513 INFO [finetune.py:1026] (0/2) Epoch 5, validation: loss=0.2918, simple_loss=0.3673, pruned_loss=0.1081, over 1020973.00 frames. 2023-05-16 01:24:57,514 INFO [finetune.py:1027] (0/2) Maximum memory allocated so far is 12745MB 2023-05-16 01:25:03,316 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=148943.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 01:25:33,858 INFO [finetune.py:992] (0/2) Epoch 5, batch 50, loss[loss=0.1917, simple_loss=0.2823, pruned_loss=0.05052, over 12270.00 frames. ], tot_loss[loss=0.1875, simple_loss=0.2744, pruned_loss=0.05029, over 541158.12 frames. ], batch size: 37, lr: 4.74e-03, grad_scale: 8.0 2023-05-16 01:25:38,884 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=148991.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 01:25:50,654 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=149007.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 01:26:07,644 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.225e+02 3.099e+02 3.518e+02 4.160e+02 1.565e+03, threshold=7.036e+02, percent-clipped=2.0 2023-05-16 01:26:10,541 INFO [finetune.py:992] (0/2) Epoch 5, batch 100, loss[loss=0.1676, simple_loss=0.2568, pruned_loss=0.03917, over 12417.00 frames. ], tot_loss[loss=0.1861, simple_loss=0.2739, pruned_loss=0.04913, over 945745.58 frames. ], batch size: 32, lr: 4.74e-03, grad_scale: 8.0 2023-05-16 01:26:25,646 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.2185, 4.8337, 5.1566, 4.5239, 4.8353, 4.5856, 5.2122, 4.9145], device='cuda:0'), covar=tensor([0.0250, 0.0318, 0.0255, 0.0257, 0.0342, 0.0315, 0.0246, 0.0245], device='cuda:0'), in_proj_covar=tensor([0.0214, 0.0219, 0.0235, 0.0214, 0.0212, 0.0210, 0.0191, 0.0177], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 01:26:37,085 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=149072.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 01:26:46,271 INFO [finetune.py:992] (0/2) Epoch 5, batch 150, loss[loss=0.1512, simple_loss=0.2405, pruned_loss=0.03094, over 12187.00 frames. ], tot_loss[loss=0.1835, simple_loss=0.2712, pruned_loss=0.04791, over 1267314.84 frames. ], batch size: 31, lr: 4.74e-03, grad_scale: 8.0 2023-05-16 01:26:46,420 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=149085.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 01:27:19,632 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.062e+02 3.007e+02 3.400e+02 3.916e+02 6.170e+02, threshold=6.801e+02, percent-clipped=0.0 2023-05-16 01:27:21,186 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=149133.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 01:27:22,581 INFO [finetune.py:992] (0/2) Epoch 5, batch 200, loss[loss=0.1723, simple_loss=0.2664, pruned_loss=0.03905, over 12361.00 frames. ], tot_loss[loss=0.1821, simple_loss=0.2692, pruned_loss=0.04753, over 1507619.68 frames. ], batch size: 35, lr: 4.73e-03, grad_scale: 8.0 2023-05-16 01:27:37,039 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=149154.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 01:27:58,882 INFO [finetune.py:992] (0/2) Epoch 5, batch 250, loss[loss=0.161, simple_loss=0.2386, pruned_loss=0.04168, over 12353.00 frames. ], tot_loss[loss=0.1804, simple_loss=0.2677, pruned_loss=0.04651, over 1700996.51 frames. ], batch size: 30, lr: 4.73e-03, grad_scale: 8.0 2023-05-16 01:28:11,297 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=149202.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 01:28:31,782 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.987e+02 2.796e+02 3.378e+02 4.055e+02 7.311e+02, threshold=6.757e+02, percent-clipped=2.0 2023-05-16 01:28:34,668 INFO [finetune.py:992] (0/2) Epoch 5, batch 300, loss[loss=0.182, simple_loss=0.2699, pruned_loss=0.04705, over 11254.00 frames. ], tot_loss[loss=0.18, simple_loss=0.2672, pruned_loss=0.04639, over 1862688.60 frames. ], batch size: 55, lr: 4.73e-03, grad_scale: 8.0 2023-05-16 01:28:36,427 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2023-05-16 01:29:10,881 INFO [finetune.py:992] (0/2) Epoch 5, batch 350, loss[loss=0.2169, simple_loss=0.2972, pruned_loss=0.06829, over 8344.00 frames. ], tot_loss[loss=0.1802, simple_loss=0.268, pruned_loss=0.04621, over 1970888.70 frames. ], batch size: 98, lr: 4.73e-03, grad_scale: 8.0 2023-05-16 01:29:27,555 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=149307.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 01:29:42,557 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.2490, 4.5068, 2.7103, 2.5483, 3.9695, 2.5184, 3.9656, 3.1581], device='cuda:0'), covar=tensor([0.0643, 0.0613, 0.1174, 0.1599, 0.0279, 0.1424, 0.0489, 0.0854], device='cuda:0'), in_proj_covar=tensor([0.0178, 0.0234, 0.0167, 0.0195, 0.0131, 0.0176, 0.0182, 0.0168], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 01:29:44,552 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.132e+02 2.873e+02 3.199e+02 3.807e+02 8.710e+02, threshold=6.398e+02, percent-clipped=1.0 2023-05-16 01:29:47,355 INFO [finetune.py:992] (0/2) Epoch 5, batch 400, loss[loss=0.1431, simple_loss=0.2294, pruned_loss=0.02841, over 12158.00 frames. ], tot_loss[loss=0.1795, simple_loss=0.2675, pruned_loss=0.04574, over 2058829.70 frames. ], batch size: 29, lr: 4.73e-03, grad_scale: 8.0 2023-05-16 01:29:52,081 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=4.11 vs. limit=5.0 2023-05-16 01:30:01,900 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=149355.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 01:30:14,263 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=149372.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 01:30:23,334 INFO [finetune.py:992] (0/2) Epoch 5, batch 450, loss[loss=0.1783, simple_loss=0.2647, pruned_loss=0.046, over 12159.00 frames. ], tot_loss[loss=0.1791, simple_loss=0.2671, pruned_loss=0.04556, over 2130038.51 frames. ], batch size: 34, lr: 4.73e-03, grad_scale: 8.0 2023-05-16 01:30:43,610 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.5144, 4.4734, 4.3579, 4.7380, 3.3935, 4.2590, 3.1357, 4.3311], device='cuda:0'), covar=tensor([0.1541, 0.0575, 0.0881, 0.0617, 0.0957, 0.0518, 0.1513, 0.1327], device='cuda:0'), in_proj_covar=tensor([0.0223, 0.0257, 0.0287, 0.0339, 0.0232, 0.0236, 0.0252, 0.0371], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 01:30:49,046 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=149420.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 01:30:56,823 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.904e+02 2.946e+02 3.604e+02 4.194e+02 7.984e+02, threshold=7.208e+02, percent-clipped=3.0 2023-05-16 01:30:58,569 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=2.29 vs. limit=5.0 2023-05-16 01:30:59,614 INFO [finetune.py:992] (0/2) Epoch 5, batch 500, loss[loss=0.2271, simple_loss=0.3056, pruned_loss=0.07426, over 8156.00 frames. ], tot_loss[loss=0.1793, simple_loss=0.2673, pruned_loss=0.0457, over 2182997.27 frames. ], batch size: 100, lr: 4.73e-03, grad_scale: 8.0 2023-05-16 01:31:07,542 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.0345, 4.0003, 3.9906, 4.0610, 3.7915, 3.7998, 3.7273, 3.9840], device='cuda:0'), covar=tensor([0.0839, 0.0689, 0.1093, 0.0711, 0.1842, 0.1467, 0.0594, 0.0980], device='cuda:0'), in_proj_covar=tensor([0.0465, 0.0598, 0.0523, 0.0547, 0.0732, 0.0668, 0.0489, 0.0441], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0005, 0.0005, 0.0003, 0.0003], device='cuda:0') 2023-05-16 01:31:17,512 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.73 vs. limit=2.0 2023-05-16 01:31:23,493 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.41 vs. limit=2.0 2023-05-16 01:31:35,610 INFO [finetune.py:992] (0/2) Epoch 5, batch 550, loss[loss=0.2231, simple_loss=0.3119, pruned_loss=0.06717, over 12127.00 frames. ], tot_loss[loss=0.1779, simple_loss=0.2662, pruned_loss=0.04476, over 2229513.14 frames. ], batch size: 39, lr: 4.73e-03, grad_scale: 8.0 2023-05-16 01:31:37,349 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.8987, 3.6708, 5.2880, 2.6378, 2.7545, 3.8703, 3.3486, 3.9429], device='cuda:0'), covar=tensor([0.0375, 0.0945, 0.0197, 0.1217, 0.1993, 0.1384, 0.1296, 0.0960], device='cuda:0'), in_proj_covar=tensor([0.0223, 0.0225, 0.0223, 0.0176, 0.0230, 0.0272, 0.0218, 0.0239], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:0') 2023-05-16 01:31:56,707 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=149514.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 01:31:57,434 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.4420, 4.1675, 4.1896, 4.5580, 3.4984, 3.9568, 2.6967, 4.1462], device='cuda:0'), covar=tensor([0.1597, 0.0644, 0.0865, 0.0722, 0.0954, 0.0631, 0.1860, 0.1299], device='cuda:0'), in_proj_covar=tensor([0.0225, 0.0259, 0.0291, 0.0343, 0.0235, 0.0238, 0.0255, 0.0375], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 01:32:08,207 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.866e+02 2.934e+02 3.488e+02 4.250e+02 1.156e+03, threshold=6.977e+02, percent-clipped=3.0 2023-05-16 01:32:10,981 INFO [finetune.py:992] (0/2) Epoch 5, batch 600, loss[loss=0.182, simple_loss=0.2743, pruned_loss=0.04482, over 12346.00 frames. ], tot_loss[loss=0.1778, simple_loss=0.2658, pruned_loss=0.04488, over 2259022.40 frames. ], batch size: 36, lr: 4.73e-03, grad_scale: 8.0 2023-05-16 01:32:33,160 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=149565.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 01:32:39,739 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 2023-05-16 01:32:40,330 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=149575.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 01:32:47,896 INFO [finetune.py:992] (0/2) Epoch 5, batch 650, loss[loss=0.1744, simple_loss=0.2642, pruned_loss=0.04237, over 12181.00 frames. ], tot_loss[loss=0.1772, simple_loss=0.2651, pruned_loss=0.04461, over 2281786.64 frames. ], batch size: 31, lr: 4.73e-03, grad_scale: 8.0 2023-05-16 01:33:13,558 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.4965, 4.8757, 2.9676, 2.8803, 4.2814, 2.7998, 4.2106, 3.4539], device='cuda:0'), covar=tensor([0.0634, 0.0511, 0.1166, 0.1537, 0.0221, 0.1278, 0.0462, 0.0783], device='cuda:0'), in_proj_covar=tensor([0.0179, 0.0237, 0.0169, 0.0195, 0.0133, 0.0177, 0.0185, 0.0169], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 01:33:17,875 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=149626.0, num_to_drop=1, layers_to_drop={3} 2023-05-16 01:33:21,111 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.180e+02 2.895e+02 3.340e+02 3.916e+02 6.633e+02, threshold=6.679e+02, percent-clipped=0.0 2023-05-16 01:33:24,015 INFO [finetune.py:992] (0/2) Epoch 5, batch 700, loss[loss=0.2063, simple_loss=0.296, pruned_loss=0.05827, over 10748.00 frames. ], tot_loss[loss=0.1768, simple_loss=0.2652, pruned_loss=0.04415, over 2301266.79 frames. ], batch size: 70, lr: 4.73e-03, grad_scale: 8.0 2023-05-16 01:33:59,614 INFO [finetune.py:992] (0/2) Epoch 5, batch 750, loss[loss=0.1936, simple_loss=0.2799, pruned_loss=0.05364, over 11331.00 frames. ], tot_loss[loss=0.1762, simple_loss=0.264, pruned_loss=0.04422, over 2319708.52 frames. ], batch size: 55, lr: 4.73e-03, grad_scale: 8.0 2023-05-16 01:34:33,104 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.109e+02 2.855e+02 3.448e+02 4.582e+02 2.974e+03, threshold=6.895e+02, percent-clipped=9.0 2023-05-16 01:34:36,613 INFO [finetune.py:992] (0/2) Epoch 5, batch 800, loss[loss=0.1816, simple_loss=0.2707, pruned_loss=0.04626, over 12116.00 frames. ], tot_loss[loss=0.1763, simple_loss=0.2644, pruned_loss=0.04412, over 2336201.01 frames. ], batch size: 39, lr: 4.73e-03, grad_scale: 16.0 2023-05-16 01:34:44,576 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=149746.0, num_to_drop=1, layers_to_drop={1} 2023-05-16 01:35:12,263 INFO [finetune.py:992] (0/2) Epoch 5, batch 850, loss[loss=0.1919, simple_loss=0.2802, pruned_loss=0.0518, over 12142.00 frames. ], tot_loss[loss=0.1755, simple_loss=0.2638, pruned_loss=0.0436, over 2354385.31 frames. ], batch size: 34, lr: 4.73e-03, grad_scale: 16.0 2023-05-16 01:35:16,327 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.44 vs. limit=2.0 2023-05-16 01:35:24,785 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.0746, 4.6920, 4.9537, 5.0121, 4.8870, 4.9913, 4.8040, 2.8733], device='cuda:0'), covar=tensor([0.0111, 0.0054, 0.0069, 0.0048, 0.0038, 0.0089, 0.0064, 0.0620], device='cuda:0'), in_proj_covar=tensor([0.0062, 0.0068, 0.0073, 0.0066, 0.0054, 0.0083, 0.0070, 0.0088], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-16 01:35:28,164 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=149807.0, num_to_drop=1, layers_to_drop={0} 2023-05-16 01:35:37,500 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.5461, 4.8307, 4.2082, 5.2492, 4.6170, 3.0624, 4.4689, 3.2075], device='cuda:0'), covar=tensor([0.0639, 0.0670, 0.1317, 0.0297, 0.1040, 0.1460, 0.0844, 0.2943], device='cuda:0'), in_proj_covar=tensor([0.0287, 0.0360, 0.0339, 0.0248, 0.0347, 0.0257, 0.0325, 0.0349], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 01:35:42,889 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=4.32 vs. limit=5.0 2023-05-16 01:35:45,308 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.919e+02 2.781e+02 3.277e+02 3.827e+02 7.458e+02, threshold=6.553e+02, percent-clipped=1.0 2023-05-16 01:35:46,973 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.3160, 4.6387, 2.7895, 2.5527, 4.0622, 2.7161, 4.0133, 3.1286], device='cuda:0'), covar=tensor([0.0654, 0.0602, 0.1141, 0.1675, 0.0274, 0.1334, 0.0407, 0.0882], device='cuda:0'), in_proj_covar=tensor([0.0181, 0.0239, 0.0169, 0.0196, 0.0134, 0.0177, 0.0186, 0.0171], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 01:35:48,169 INFO [finetune.py:992] (0/2) Epoch 5, batch 900, loss[loss=0.1472, simple_loss=0.2418, pruned_loss=0.0263, over 12098.00 frames. ], tot_loss[loss=0.1753, simple_loss=0.2637, pruned_loss=0.04342, over 2367404.74 frames. ], batch size: 32, lr: 4.73e-03, grad_scale: 16.0 2023-05-16 01:36:13,613 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=149870.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 01:36:24,049 INFO [finetune.py:992] (0/2) Epoch 5, batch 950, loss[loss=0.1783, simple_loss=0.2675, pruned_loss=0.04458, over 12098.00 frames. ], tot_loss[loss=0.1758, simple_loss=0.264, pruned_loss=0.04379, over 2367545.72 frames. ], batch size: 33, lr: 4.73e-03, grad_scale: 16.0 2023-05-16 01:36:24,238 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.8133, 4.5261, 4.7495, 4.7421, 4.5752, 4.7646, 4.5695, 2.6146], device='cuda:0'), covar=tensor([0.0134, 0.0060, 0.0079, 0.0057, 0.0057, 0.0100, 0.0074, 0.0688], device='cuda:0'), in_proj_covar=tensor([0.0062, 0.0069, 0.0073, 0.0066, 0.0054, 0.0084, 0.0070, 0.0088], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-16 01:36:50,163 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=149921.0, num_to_drop=1, layers_to_drop={2} 2023-05-16 01:36:52,266 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.3370, 4.8454, 5.3041, 4.6035, 4.9201, 4.6651, 5.2904, 5.0053], device='cuda:0'), covar=tensor([0.0238, 0.0367, 0.0247, 0.0259, 0.0334, 0.0309, 0.0256, 0.0257], device='cuda:0'), in_proj_covar=tensor([0.0230, 0.0236, 0.0255, 0.0230, 0.0229, 0.0229, 0.0207, 0.0192], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 01:36:57,092 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.927e+02 2.907e+02 3.388e+02 4.028e+02 1.015e+03, threshold=6.776e+02, percent-clipped=5.0 2023-05-16 01:36:59,949 INFO [finetune.py:992] (0/2) Epoch 5, batch 1000, loss[loss=0.181, simple_loss=0.2639, pruned_loss=0.04902, over 12121.00 frames. ], tot_loss[loss=0.1758, simple_loss=0.264, pruned_loss=0.04381, over 2376144.83 frames. ], batch size: 33, lr: 4.73e-03, grad_scale: 16.0 2023-05-16 01:37:02,381 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.9715, 4.2786, 3.7742, 4.7645, 4.2436, 2.7025, 4.0746, 2.8805], device='cuda:0'), covar=tensor([0.0880, 0.0953, 0.1600, 0.0367, 0.1171, 0.1635, 0.0990, 0.3281], device='cuda:0'), in_proj_covar=tensor([0.0290, 0.0366, 0.0343, 0.0253, 0.0352, 0.0261, 0.0328, 0.0354], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 01:37:35,499 INFO [finetune.py:992] (0/2) Epoch 5, batch 1050, loss[loss=0.1809, simple_loss=0.267, pruned_loss=0.04739, over 12176.00 frames. ], tot_loss[loss=0.1756, simple_loss=0.2638, pruned_loss=0.04375, over 2384708.88 frames. ], batch size: 31, lr: 4.73e-03, grad_scale: 16.0 2023-05-16 01:37:43,764 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.0910, 2.4502, 3.7694, 3.0664, 3.5031, 3.2140, 2.5243, 3.6171], device='cuda:0'), covar=tensor([0.0130, 0.0334, 0.0117, 0.0225, 0.0120, 0.0147, 0.0316, 0.0100], device='cuda:0'), in_proj_covar=tensor([0.0157, 0.0192, 0.0166, 0.0169, 0.0192, 0.0146, 0.0182, 0.0161], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 01:37:46,782 INFO [checkpoint.py:75] (0/2) Saving checkpoint to pruned_transducer_stateless7/exp_giga_finetune/checkpoint-50000.pt 2023-05-16 01:38:12,044 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.860e+02 2.704e+02 3.135e+02 3.957e+02 6.595e+02, threshold=6.270e+02, percent-clipped=0.0 2023-05-16 01:38:14,898 INFO [finetune.py:992] (0/2) Epoch 5, batch 1100, loss[loss=0.1339, simple_loss=0.2148, pruned_loss=0.02647, over 12001.00 frames. ], tot_loss[loss=0.1746, simple_loss=0.2627, pruned_loss=0.04329, over 2393437.81 frames. ], batch size: 28, lr: 4.73e-03, grad_scale: 16.0 2023-05-16 01:38:47,023 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=150079.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 01:38:51,135 INFO [finetune.py:992] (0/2) Epoch 5, batch 1150, loss[loss=0.1799, simple_loss=0.2739, pruned_loss=0.04289, over 12155.00 frames. ], tot_loss[loss=0.175, simple_loss=0.2628, pruned_loss=0.04364, over 2387408.37 frames. ], batch size: 39, lr: 4.73e-03, grad_scale: 16.0 2023-05-16 01:38:52,692 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.9523, 5.9215, 5.6634, 5.1905, 5.1027, 5.8358, 5.4228, 5.2156], device='cuda:0'), covar=tensor([0.0793, 0.0850, 0.0750, 0.1510, 0.0675, 0.0692, 0.1414, 0.1075], device='cuda:0'), in_proj_covar=tensor([0.0570, 0.0504, 0.0475, 0.0587, 0.0378, 0.0648, 0.0710, 0.0528], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0004, 0.0004, 0.0003], device='cuda:0') 2023-05-16 01:39:03,256 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=150102.0, num_to_drop=1, layers_to_drop={2} 2023-05-16 01:39:14,014 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.4236, 5.2770, 5.3246, 5.3926, 4.9820, 5.0348, 4.8134, 5.4032], device='cuda:0'), covar=tensor([0.0664, 0.0531, 0.0830, 0.0631, 0.1958, 0.1314, 0.0602, 0.0815], device='cuda:0'), in_proj_covar=tensor([0.0481, 0.0623, 0.0537, 0.0563, 0.0760, 0.0684, 0.0505, 0.0457], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0005, 0.0005, 0.0004, 0.0004], device='cuda:0') 2023-05-16 01:39:23,520 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.969e+02 2.882e+02 3.382e+02 3.965e+02 8.013e+02, threshold=6.765e+02, percent-clipped=2.0 2023-05-16 01:39:26,418 INFO [finetune.py:992] (0/2) Epoch 5, batch 1200, loss[loss=0.1896, simple_loss=0.2838, pruned_loss=0.04772, over 12113.00 frames. ], tot_loss[loss=0.1753, simple_loss=0.263, pruned_loss=0.04373, over 2387915.22 frames. ], batch size: 39, lr: 4.73e-03, grad_scale: 16.0 2023-05-16 01:39:30,142 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=150140.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 01:39:47,279 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 2023-05-16 01:39:51,345 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=150170.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 01:40:02,423 INFO [finetune.py:992] (0/2) Epoch 5, batch 1250, loss[loss=0.1715, simple_loss=0.2548, pruned_loss=0.04414, over 12312.00 frames. ], tot_loss[loss=0.1749, simple_loss=0.2625, pruned_loss=0.04365, over 2391556.70 frames. ], batch size: 34, lr: 4.73e-03, grad_scale: 16.0 2023-05-16 01:40:27,189 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=150218.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 01:40:29,435 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=150221.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 01:40:36,479 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.759e+02 2.917e+02 3.432e+02 3.928e+02 1.096e+03, threshold=6.863e+02, percent-clipped=3.0 2023-05-16 01:40:39,358 INFO [finetune.py:992] (0/2) Epoch 5, batch 1300, loss[loss=0.2019, simple_loss=0.291, pruned_loss=0.05638, over 10334.00 frames. ], tot_loss[loss=0.1746, simple_loss=0.2625, pruned_loss=0.04334, over 2386549.16 frames. ], batch size: 68, lr: 4.72e-03, grad_scale: 16.0 2023-05-16 01:40:39,693 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=3.13 vs. limit=5.0 2023-05-16 01:40:46,383 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.40 vs. limit=2.0 2023-05-16 01:40:58,115 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.4706, 5.3630, 5.4117, 5.4798, 5.0445, 5.0679, 4.9099, 5.4427], device='cuda:0'), covar=tensor([0.0656, 0.0476, 0.0591, 0.0550, 0.1792, 0.1251, 0.0520, 0.0743], device='cuda:0'), in_proj_covar=tensor([0.0486, 0.0626, 0.0542, 0.0572, 0.0768, 0.0694, 0.0511, 0.0463], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0005, 0.0004, 0.0004, 0.0005, 0.0005, 0.0004, 0.0004], device='cuda:0') 2023-05-16 01:41:03,803 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=150269.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 01:41:06,090 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.9595, 2.3086, 3.5758, 2.9336, 3.3678, 3.1010, 2.2937, 3.5001], device='cuda:0'), covar=tensor([0.0122, 0.0312, 0.0147, 0.0210, 0.0122, 0.0160, 0.0379, 0.0100], device='cuda:0'), in_proj_covar=tensor([0.0159, 0.0194, 0.0168, 0.0172, 0.0194, 0.0147, 0.0185, 0.0164], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 01:41:07,525 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=150274.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 01:41:15,265 INFO [finetune.py:992] (0/2) Epoch 5, batch 1350, loss[loss=0.1835, simple_loss=0.2761, pruned_loss=0.04548, over 12139.00 frames. ], tot_loss[loss=0.1745, simple_loss=0.2624, pruned_loss=0.04336, over 2376800.70 frames. ], batch size: 39, lr: 4.72e-03, grad_scale: 16.0 2023-05-16 01:41:39,607 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.44 vs. limit=2.0 2023-05-16 01:41:48,909 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.899e+02 2.823e+02 3.330e+02 3.848e+02 6.429e+02, threshold=6.661e+02, percent-clipped=0.0 2023-05-16 01:41:51,802 INFO [finetune.py:992] (0/2) Epoch 5, batch 1400, loss[loss=0.1639, simple_loss=0.2478, pruned_loss=0.03999, over 11871.00 frames. ], tot_loss[loss=0.1742, simple_loss=0.2623, pruned_loss=0.04304, over 2379638.76 frames. ], batch size: 26, lr: 4.72e-03, grad_scale: 16.0 2023-05-16 01:41:52,031 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=150335.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 01:41:52,895 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=3.30 vs. limit=5.0 2023-05-16 01:42:19,706 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.2641, 6.1163, 5.7628, 5.6785, 6.2047, 5.5119, 5.7360, 5.6690], device='cuda:0'), covar=tensor([0.1336, 0.0949, 0.0867, 0.1754, 0.0965, 0.2111, 0.1540, 0.1191], device='cuda:0'), in_proj_covar=tensor([0.0326, 0.0448, 0.0354, 0.0403, 0.0427, 0.0408, 0.0363, 0.0348], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 01:42:28,126 INFO [finetune.py:992] (0/2) Epoch 5, batch 1450, loss[loss=0.1629, simple_loss=0.2352, pruned_loss=0.04528, over 12004.00 frames. ], tot_loss[loss=0.174, simple_loss=0.2621, pruned_loss=0.04294, over 2381608.55 frames. ], batch size: 28, lr: 4.72e-03, grad_scale: 16.0 2023-05-16 01:42:40,611 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=150402.0, num_to_drop=1, layers_to_drop={0} 2023-05-16 01:42:51,677 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=4.82 vs. limit=5.0 2023-05-16 01:43:00,975 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.806e+02 2.801e+02 3.263e+02 4.020e+02 6.215e+02, threshold=6.526e+02, percent-clipped=0.0 2023-05-16 01:43:03,770 INFO [finetune.py:992] (0/2) Epoch 5, batch 1500, loss[loss=0.2003, simple_loss=0.297, pruned_loss=0.05181, over 12349.00 frames. ], tot_loss[loss=0.1748, simple_loss=0.2627, pruned_loss=0.04347, over 2376166.92 frames. ], batch size: 36, lr: 4.72e-03, grad_scale: 16.0 2023-05-16 01:43:03,849 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=150435.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 01:43:14,393 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=150450.0, num_to_drop=1, layers_to_drop={0} 2023-05-16 01:43:39,642 INFO [finetune.py:992] (0/2) Epoch 5, batch 1550, loss[loss=0.1572, simple_loss=0.25, pruned_loss=0.03218, over 12116.00 frames. ], tot_loss[loss=0.1748, simple_loss=0.2627, pruned_loss=0.04344, over 2380250.24 frames. ], batch size: 33, lr: 4.72e-03, grad_scale: 16.0 2023-05-16 01:43:57,505 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.4848, 4.9877, 5.4533, 4.7297, 4.9939, 4.7981, 5.4559, 5.0968], device='cuda:0'), covar=tensor([0.0203, 0.0343, 0.0222, 0.0226, 0.0333, 0.0251, 0.0182, 0.0225], device='cuda:0'), in_proj_covar=tensor([0.0230, 0.0237, 0.0256, 0.0231, 0.0230, 0.0228, 0.0207, 0.0192], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 01:44:12,913 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.840e+02 2.822e+02 3.242e+02 4.160e+02 8.190e+02, threshold=6.485e+02, percent-clipped=2.0 2023-05-16 01:44:15,895 INFO [finetune.py:992] (0/2) Epoch 5, batch 1600, loss[loss=0.1699, simple_loss=0.2626, pruned_loss=0.03857, over 11806.00 frames. ], tot_loss[loss=0.1743, simple_loss=0.2625, pruned_loss=0.04306, over 2381787.42 frames. ], batch size: 44, lr: 4.72e-03, grad_scale: 16.0 2023-05-16 01:44:17,496 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=150537.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 01:44:51,100 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.7941, 2.6582, 3.3161, 4.5241, 2.6572, 4.5192, 4.6441, 4.7753], device='cuda:0'), covar=tensor([0.0122, 0.1176, 0.0442, 0.0153, 0.1204, 0.0253, 0.0164, 0.0102], device='cuda:0'), in_proj_covar=tensor([0.0117, 0.0201, 0.0183, 0.0114, 0.0188, 0.0174, 0.0165, 0.0119], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 01:44:51,627 INFO [finetune.py:992] (0/2) Epoch 5, batch 1650, loss[loss=0.1542, simple_loss=0.2319, pruned_loss=0.03824, over 11756.00 frames. ], tot_loss[loss=0.1745, simple_loss=0.2622, pruned_loss=0.04342, over 2375301.20 frames. ], batch size: 26, lr: 4.72e-03, grad_scale: 16.0 2023-05-16 01:45:01,233 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=150598.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 01:45:15,554 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.6480, 3.1525, 5.2317, 2.6752, 2.5367, 3.8345, 2.9625, 3.8593], device='cuda:0'), covar=tensor([0.0445, 0.1210, 0.0286, 0.1220, 0.2088, 0.1281, 0.1576, 0.1042], device='cuda:0'), in_proj_covar=tensor([0.0229, 0.0228, 0.0230, 0.0177, 0.0234, 0.0279, 0.0221, 0.0245], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:0') 2023-05-16 01:45:24,695 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=150630.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 01:45:25,335 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.928e+02 2.880e+02 3.247e+02 4.025e+02 9.318e+02, threshold=6.493e+02, percent-clipped=2.0 2023-05-16 01:45:28,895 INFO [finetune.py:992] (0/2) Epoch 5, batch 1700, loss[loss=0.1718, simple_loss=0.2638, pruned_loss=0.03987, over 12354.00 frames. ], tot_loss[loss=0.1751, simple_loss=0.2629, pruned_loss=0.04368, over 2380331.66 frames. ], batch size: 36, lr: 4.72e-03, grad_scale: 16.0 2023-05-16 01:46:04,298 INFO [finetune.py:992] (0/2) Epoch 5, batch 1750, loss[loss=0.2201, simple_loss=0.3059, pruned_loss=0.06716, over 12344.00 frames. ], tot_loss[loss=0.1747, simple_loss=0.2625, pruned_loss=0.04347, over 2375645.28 frames. ], batch size: 36, lr: 4.72e-03, grad_scale: 16.0 2023-05-16 01:46:04,605 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.7931, 2.8855, 4.6062, 4.8702, 2.8449, 2.6742, 3.0531, 2.1212], device='cuda:0'), covar=tensor([0.1325, 0.2788, 0.0402, 0.0313, 0.1157, 0.1942, 0.2310, 0.3604], device='cuda:0'), in_proj_covar=tensor([0.0269, 0.0366, 0.0258, 0.0279, 0.0246, 0.0277, 0.0345, 0.0350], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-16 01:46:36,894 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.084e+02 2.891e+02 3.374e+02 3.862e+02 1.303e+03, threshold=6.748e+02, percent-clipped=4.0 2023-05-16 01:46:37,653 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=150732.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 01:46:39,491 INFO [finetune.py:992] (0/2) Epoch 5, batch 1800, loss[loss=0.1738, simple_loss=0.2654, pruned_loss=0.04105, over 11690.00 frames. ], tot_loss[loss=0.174, simple_loss=0.2615, pruned_loss=0.04323, over 2385788.69 frames. ], batch size: 48, lr: 4.72e-03, grad_scale: 16.0 2023-05-16 01:46:39,623 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=150735.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 01:46:41,937 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.6442, 2.5821, 3.7390, 4.6685, 3.9376, 4.5755, 4.0062, 3.4013], device='cuda:0'), covar=tensor([0.0030, 0.0355, 0.0127, 0.0023, 0.0129, 0.0058, 0.0084, 0.0279], device='cuda:0'), in_proj_covar=tensor([0.0080, 0.0119, 0.0099, 0.0072, 0.0097, 0.0109, 0.0086, 0.0133], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 01:46:48,158 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.6732, 2.5039, 3.9066, 4.1217, 2.8384, 2.5935, 2.7264, 2.2386], device='cuda:0'), covar=tensor([0.1427, 0.2981, 0.0597, 0.0456, 0.1126, 0.2041, 0.2595, 0.3674], device='cuda:0'), in_proj_covar=tensor([0.0272, 0.0369, 0.0261, 0.0281, 0.0248, 0.0278, 0.0348, 0.0352], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-16 01:47:13,416 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=150783.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 01:47:15,443 INFO [finetune.py:992] (0/2) Epoch 5, batch 1850, loss[loss=0.2115, simple_loss=0.3008, pruned_loss=0.06113, over 12034.00 frames. ], tot_loss[loss=0.1746, simple_loss=0.2624, pruned_loss=0.04339, over 2384928.67 frames. ], batch size: 45, lr: 4.72e-03, grad_scale: 16.0 2023-05-16 01:47:18,039 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=4.53 vs. limit=5.0 2023-05-16 01:47:21,224 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=150793.0, num_to_drop=1, layers_to_drop={1} 2023-05-16 01:47:31,353 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.8490, 3.7284, 3.7130, 3.8695, 3.5542, 3.9439, 3.9120, 4.0589], device='cuda:0'), covar=tensor([0.0263, 0.0177, 0.0209, 0.0379, 0.0676, 0.0322, 0.0179, 0.0234], device='cuda:0'), in_proj_covar=tensor([0.0180, 0.0181, 0.0179, 0.0232, 0.0228, 0.0199, 0.0164, 0.0216], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0005, 0.0004, 0.0005, 0.0004, 0.0005], device='cuda:0') 2023-05-16 01:47:38,758 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.4597, 3.0877, 4.8610, 2.6465, 2.8318, 3.7279, 3.1258, 3.7498], device='cuda:0'), covar=tensor([0.0454, 0.1085, 0.0302, 0.1053, 0.1603, 0.1308, 0.1258, 0.1090], device='cuda:0'), in_proj_covar=tensor([0.0225, 0.0224, 0.0226, 0.0174, 0.0228, 0.0273, 0.0217, 0.0242], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:0') 2023-05-16 01:47:46,020 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.39 vs. limit=2.0 2023-05-16 01:47:48,477 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.938e+02 2.918e+02 3.629e+02 4.209e+02 8.227e+02, threshold=7.258e+02, percent-clipped=1.0 2023-05-16 01:47:51,368 INFO [finetune.py:992] (0/2) Epoch 5, batch 1900, loss[loss=0.1738, simple_loss=0.2702, pruned_loss=0.03871, over 12347.00 frames. ], tot_loss[loss=0.1745, simple_loss=0.2623, pruned_loss=0.0433, over 2389083.34 frames. ], batch size: 35, lr: 4.72e-03, grad_scale: 16.0 2023-05-16 01:47:53,090 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.0463, 2.3418, 3.5112, 2.8496, 3.3657, 3.0701, 2.3742, 3.5011], device='cuda:0'), covar=tensor([0.0130, 0.0375, 0.0156, 0.0239, 0.0179, 0.0168, 0.0393, 0.0107], device='cuda:0'), in_proj_covar=tensor([0.0158, 0.0192, 0.0168, 0.0171, 0.0192, 0.0146, 0.0184, 0.0163], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 01:48:26,686 INFO [finetune.py:992] (0/2) Epoch 5, batch 1950, loss[loss=0.1449, simple_loss=0.2287, pruned_loss=0.03049, over 11829.00 frames. ], tot_loss[loss=0.1762, simple_loss=0.2639, pruned_loss=0.04423, over 2382847.05 frames. ], batch size: 26, lr: 4.72e-03, grad_scale: 16.0 2023-05-16 01:48:29,115 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.0728, 4.7200, 4.8203, 4.9392, 4.7159, 4.9205, 4.8186, 2.6129], device='cuda:0'), covar=tensor([0.0076, 0.0058, 0.0079, 0.0057, 0.0051, 0.0088, 0.0069, 0.0734], device='cuda:0'), in_proj_covar=tensor([0.0063, 0.0070, 0.0074, 0.0067, 0.0055, 0.0085, 0.0072, 0.0089], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-16 01:48:32,451 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=150893.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 01:48:35,782 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=4.55 vs. limit=5.0 2023-05-16 01:48:37,531 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=150900.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 01:48:51,032 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.6598, 2.2855, 3.7499, 4.6145, 3.9195, 4.5672, 4.0211, 3.1101], device='cuda:0'), covar=tensor([0.0026, 0.0420, 0.0109, 0.0031, 0.0121, 0.0059, 0.0075, 0.0355], device='cuda:0'), in_proj_covar=tensor([0.0080, 0.0121, 0.0100, 0.0072, 0.0098, 0.0110, 0.0087, 0.0134], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 01:48:59,294 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=150930.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 01:48:59,868 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.924e+02 2.802e+02 3.415e+02 4.185e+02 7.760e+02, threshold=6.830e+02, percent-clipped=3.0 2023-05-16 01:49:01,461 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.7292, 2.5396, 3.8374, 4.7066, 3.9528, 4.6554, 4.0945, 3.3123], device='cuda:0'), covar=tensor([0.0025, 0.0372, 0.0109, 0.0027, 0.0130, 0.0057, 0.0090, 0.0318], device='cuda:0'), in_proj_covar=tensor([0.0080, 0.0122, 0.0100, 0.0072, 0.0099, 0.0111, 0.0087, 0.0135], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 01:49:03,432 INFO [finetune.py:992] (0/2) Epoch 5, batch 2000, loss[loss=0.1889, simple_loss=0.2851, pruned_loss=0.04635, over 12056.00 frames. ], tot_loss[loss=0.1762, simple_loss=0.264, pruned_loss=0.04424, over 2385959.50 frames. ], batch size: 42, lr: 4.72e-03, grad_scale: 16.0 2023-05-16 01:49:05,314 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=3.10 vs. limit=5.0 2023-05-16 01:49:19,583 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.53 vs. limit=2.0 2023-05-16 01:49:22,278 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=150961.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 01:49:34,357 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=150978.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 01:49:35,939 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.2285, 3.0107, 4.6906, 2.4465, 2.7587, 3.5372, 3.0853, 3.6865], device='cuda:0'), covar=tensor([0.0471, 0.1180, 0.0260, 0.1207, 0.1711, 0.1232, 0.1287, 0.0954], device='cuda:0'), in_proj_covar=tensor([0.0229, 0.0228, 0.0230, 0.0178, 0.0232, 0.0279, 0.0221, 0.0246], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:0') 2023-05-16 01:49:39,317 INFO [finetune.py:992] (0/2) Epoch 5, batch 2050, loss[loss=0.1744, simple_loss=0.2624, pruned_loss=0.04321, over 12343.00 frames. ], tot_loss[loss=0.1757, simple_loss=0.2631, pruned_loss=0.04417, over 2385194.85 frames. ], batch size: 31, lr: 4.72e-03, grad_scale: 16.0 2023-05-16 01:50:04,876 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.66 vs. limit=2.0 2023-05-16 01:50:12,131 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.726e+02 2.900e+02 3.345e+02 3.884e+02 7.651e+02, threshold=6.691e+02, percent-clipped=2.0 2023-05-16 01:50:15,080 INFO [finetune.py:992] (0/2) Epoch 5, batch 2100, loss[loss=0.1674, simple_loss=0.2606, pruned_loss=0.03709, over 12183.00 frames. ], tot_loss[loss=0.1762, simple_loss=0.2636, pruned_loss=0.0444, over 2378923.48 frames. ], batch size: 35, lr: 4.72e-03, grad_scale: 16.0 2023-05-16 01:50:24,781 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2023-05-16 01:50:33,156 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.7016, 3.2648, 5.1556, 2.7252, 2.8370, 3.8294, 3.2374, 3.8895], device='cuda:0'), covar=tensor([0.0446, 0.1154, 0.0261, 0.1093, 0.1853, 0.1441, 0.1288, 0.1101], device='cuda:0'), in_proj_covar=tensor([0.0226, 0.0225, 0.0227, 0.0176, 0.0230, 0.0276, 0.0219, 0.0244], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:0') 2023-05-16 01:50:33,177 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.3403, 4.6572, 4.1177, 5.0438, 4.5985, 3.0578, 4.2195, 3.0375], device='cuda:0'), covar=tensor([0.0767, 0.0831, 0.1351, 0.0324, 0.0968, 0.1464, 0.1063, 0.3269], device='cuda:0'), in_proj_covar=tensor([0.0290, 0.0363, 0.0343, 0.0254, 0.0351, 0.0258, 0.0328, 0.0352], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 01:50:44,968 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.6154, 3.3061, 5.1307, 2.5317, 2.7317, 3.8949, 3.1474, 3.9630], device='cuda:0'), covar=tensor([0.0394, 0.1048, 0.0196, 0.1142, 0.1771, 0.1157, 0.1300, 0.0812], device='cuda:0'), in_proj_covar=tensor([0.0226, 0.0225, 0.0227, 0.0176, 0.0230, 0.0276, 0.0219, 0.0244], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:0') 2023-05-16 01:50:51,738 INFO [finetune.py:992] (0/2) Epoch 5, batch 2150, loss[loss=0.1658, simple_loss=0.2497, pruned_loss=0.04096, over 12038.00 frames. ], tot_loss[loss=0.1761, simple_loss=0.2638, pruned_loss=0.04417, over 2388277.94 frames. ], batch size: 31, lr: 4.72e-03, grad_scale: 16.0 2023-05-16 01:50:53,957 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=151088.0, num_to_drop=1, layers_to_drop={3} 2023-05-16 01:51:05,705 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.5029, 4.8268, 3.0875, 2.6760, 4.1914, 2.4990, 4.2064, 3.2857], device='cuda:0'), covar=tensor([0.0569, 0.0408, 0.0840, 0.1398, 0.0215, 0.1193, 0.0394, 0.0698], device='cuda:0'), in_proj_covar=tensor([0.0181, 0.0241, 0.0168, 0.0194, 0.0134, 0.0176, 0.0186, 0.0170], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 01:51:23,880 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.028e+02 2.729e+02 3.385e+02 4.097e+02 8.908e+02, threshold=6.771e+02, percent-clipped=1.0 2023-05-16 01:51:26,696 INFO [finetune.py:992] (0/2) Epoch 5, batch 2200, loss[loss=0.1871, simple_loss=0.2717, pruned_loss=0.05127, over 12088.00 frames. ], tot_loss[loss=0.1768, simple_loss=0.2644, pruned_loss=0.04457, over 2380453.73 frames. ], batch size: 32, lr: 4.72e-03, grad_scale: 16.0 2023-05-16 01:51:58,843 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.1845, 2.5252, 3.7123, 3.0890, 3.4491, 3.2681, 2.5316, 3.5576], device='cuda:0'), covar=tensor([0.0104, 0.0274, 0.0116, 0.0199, 0.0133, 0.0136, 0.0325, 0.0117], device='cuda:0'), in_proj_covar=tensor([0.0161, 0.0194, 0.0170, 0.0173, 0.0194, 0.0149, 0.0186, 0.0165], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 01:52:00,863 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.9040, 5.7350, 5.4173, 5.3026, 5.8205, 5.1560, 5.2785, 5.3683], device='cuda:0'), covar=tensor([0.1424, 0.0987, 0.0879, 0.1865, 0.1016, 0.1924, 0.1800, 0.1200], device='cuda:0'), in_proj_covar=tensor([0.0330, 0.0452, 0.0356, 0.0405, 0.0432, 0.0412, 0.0366, 0.0352], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 01:52:01,702 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=151184.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 01:52:02,199 INFO [finetune.py:992] (0/2) Epoch 5, batch 2250, loss[loss=0.1551, simple_loss=0.2513, pruned_loss=0.0295, over 12152.00 frames. ], tot_loss[loss=0.1768, simple_loss=0.2644, pruned_loss=0.04455, over 2377990.70 frames. ], batch size: 34, lr: 4.72e-03, grad_scale: 16.0 2023-05-16 01:52:05,512 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.33 vs. limit=2.0 2023-05-16 01:52:08,031 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=151193.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 01:52:10,131 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.3618, 5.1495, 5.3395, 5.2991, 4.6471, 4.8584, 4.7699, 5.2456], device='cuda:0'), covar=tensor([0.0871, 0.0779, 0.0904, 0.0817, 0.2755, 0.1795, 0.0674, 0.1293], device='cuda:0'), in_proj_covar=tensor([0.0482, 0.0623, 0.0544, 0.0577, 0.0770, 0.0688, 0.0510, 0.0462], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0005, 0.0005, 0.0004, 0.0004], device='cuda:0') 2023-05-16 01:52:32,576 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.42 vs. limit=2.0 2023-05-16 01:52:36,363 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.801e+02 2.934e+02 3.390e+02 3.969e+02 1.202e+03, threshold=6.780e+02, percent-clipped=1.0 2023-05-16 01:52:39,303 INFO [finetune.py:992] (0/2) Epoch 5, batch 2300, loss[loss=0.1809, simple_loss=0.2756, pruned_loss=0.04314, over 12113.00 frames. ], tot_loss[loss=0.1768, simple_loss=0.2644, pruned_loss=0.04462, over 2379282.06 frames. ], batch size: 39, lr: 4.72e-03, grad_scale: 16.0 2023-05-16 01:52:43,005 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.5352, 3.2346, 3.0603, 2.8644, 2.6353, 2.5581, 3.2476, 2.0980], device='cuda:0'), covar=tensor([0.0274, 0.0141, 0.0162, 0.0171, 0.0343, 0.0303, 0.0121, 0.0436], device='cuda:0'), in_proj_covar=tensor([0.0185, 0.0156, 0.0151, 0.0179, 0.0200, 0.0192, 0.0158, 0.0200], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 01:52:43,503 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=151241.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 01:52:46,370 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=151245.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 01:52:54,110 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=151256.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 01:53:09,051 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.6016, 2.1917, 2.9554, 2.5760, 2.7819, 2.8138, 2.1989, 2.9252], device='cuda:0'), covar=tensor([0.0124, 0.0288, 0.0177, 0.0224, 0.0201, 0.0148, 0.0320, 0.0147], device='cuda:0'), in_proj_covar=tensor([0.0161, 0.0195, 0.0171, 0.0174, 0.0195, 0.0149, 0.0187, 0.0167], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 01:53:12,479 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=151282.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 01:53:14,407 INFO [finetune.py:992] (0/2) Epoch 5, batch 2350, loss[loss=0.1855, simple_loss=0.2728, pruned_loss=0.04909, over 12140.00 frames. ], tot_loss[loss=0.176, simple_loss=0.2636, pruned_loss=0.04427, over 2383749.07 frames. ], batch size: 38, lr: 4.71e-03, grad_scale: 16.0 2023-05-16 01:53:18,467 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.61 vs. limit=2.0 2023-05-16 01:53:23,915 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=151298.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 01:53:46,946 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.999e+02 2.953e+02 3.417e+02 4.004e+02 7.292e+02, threshold=6.835e+02, percent-clipped=1.0 2023-05-16 01:53:49,772 INFO [finetune.py:992] (0/2) Epoch 5, batch 2400, loss[loss=0.1924, simple_loss=0.2872, pruned_loss=0.04887, over 12279.00 frames. ], tot_loss[loss=0.1766, simple_loss=0.2641, pruned_loss=0.04453, over 2374531.85 frames. ], batch size: 37, lr: 4.71e-03, grad_scale: 16.0 2023-05-16 01:53:55,428 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=151343.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 01:54:07,628 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=151359.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 01:54:26,428 INFO [finetune.py:992] (0/2) Epoch 5, batch 2450, loss[loss=0.1538, simple_loss=0.233, pruned_loss=0.03736, over 12186.00 frames. ], tot_loss[loss=0.1754, simple_loss=0.2626, pruned_loss=0.04411, over 2377506.06 frames. ], batch size: 29, lr: 4.71e-03, grad_scale: 16.0 2023-05-16 01:54:28,676 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=151388.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 01:54:59,077 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.105e+02 2.885e+02 3.251e+02 3.670e+02 6.500e+02, threshold=6.501e+02, percent-clipped=0.0 2023-05-16 01:55:01,840 INFO [finetune.py:992] (0/2) Epoch 5, batch 2500, loss[loss=0.1673, simple_loss=0.2512, pruned_loss=0.04166, over 12121.00 frames. ], tot_loss[loss=0.1749, simple_loss=0.2623, pruned_loss=0.04378, over 2380259.58 frames. ], batch size: 30, lr: 4.71e-03, grad_scale: 16.0 2023-05-16 01:55:02,622 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=151436.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 01:55:28,325 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.7674, 2.2894, 3.1540, 2.6979, 3.0620, 2.9532, 2.3669, 3.1796], device='cuda:0'), covar=tensor([0.0121, 0.0275, 0.0157, 0.0221, 0.0143, 0.0144, 0.0299, 0.0117], device='cuda:0'), in_proj_covar=tensor([0.0161, 0.0194, 0.0172, 0.0173, 0.0194, 0.0149, 0.0187, 0.0166], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 01:55:37,280 INFO [finetune.py:992] (0/2) Epoch 5, batch 2550, loss[loss=0.1478, simple_loss=0.2308, pruned_loss=0.03237, over 12043.00 frames. ], tot_loss[loss=0.1755, simple_loss=0.2629, pruned_loss=0.04408, over 2380956.49 frames. ], batch size: 28, lr: 4.71e-03, grad_scale: 16.0 2023-05-16 01:56:02,382 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.35 vs. limit=2.0 2023-05-16 01:56:09,561 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.4132, 4.9130, 5.3875, 4.6804, 4.9708, 4.7682, 5.3854, 4.9634], device='cuda:0'), covar=tensor([0.0249, 0.0339, 0.0244, 0.0248, 0.0317, 0.0305, 0.0178, 0.0240], device='cuda:0'), in_proj_covar=tensor([0.0236, 0.0240, 0.0261, 0.0234, 0.0234, 0.0232, 0.0211, 0.0197], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 01:56:10,859 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.911e+02 2.895e+02 3.302e+02 4.132e+02 5.921e+02, threshold=6.604e+02, percent-clipped=0.0 2023-05-16 01:56:13,778 INFO [finetune.py:992] (0/2) Epoch 5, batch 2600, loss[loss=0.1816, simple_loss=0.2684, pruned_loss=0.04735, over 11650.00 frames. ], tot_loss[loss=0.1762, simple_loss=0.2636, pruned_loss=0.04438, over 2378360.08 frames. ], batch size: 48, lr: 4.71e-03, grad_scale: 16.0 2023-05-16 01:56:17,467 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=151540.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 01:56:19,767 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.5145, 2.4770, 3.2300, 4.3665, 2.0682, 4.4035, 4.4595, 4.5683], device='cuda:0'), covar=tensor([0.0148, 0.1223, 0.0477, 0.0160, 0.1439, 0.0210, 0.0131, 0.0108], device='cuda:0'), in_proj_covar=tensor([0.0118, 0.0203, 0.0185, 0.0114, 0.0189, 0.0176, 0.0168, 0.0121], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 01:56:27,573 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.0535, 4.7218, 4.8764, 5.0085, 4.7693, 4.9540, 4.8055, 2.6170], device='cuda:0'), covar=tensor([0.0131, 0.0063, 0.0078, 0.0053, 0.0047, 0.0081, 0.0078, 0.0739], device='cuda:0'), in_proj_covar=tensor([0.0064, 0.0071, 0.0075, 0.0068, 0.0056, 0.0086, 0.0074, 0.0090], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 01:56:28,783 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=151556.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 01:56:49,299 INFO [finetune.py:992] (0/2) Epoch 5, batch 2650, loss[loss=0.1713, simple_loss=0.2647, pruned_loss=0.03897, over 12370.00 frames. ], tot_loss[loss=0.1752, simple_loss=0.2627, pruned_loss=0.04386, over 2377269.34 frames. ], batch size: 35, lr: 4.71e-03, grad_scale: 16.0 2023-05-16 01:56:53,834 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.5142, 4.5746, 4.1928, 5.0668, 4.7241, 3.2694, 4.4097, 3.1755], device='cuda:0'), covar=tensor([0.0639, 0.0842, 0.1365, 0.0368, 0.0860, 0.1270, 0.0830, 0.2926], device='cuda:0'), in_proj_covar=tensor([0.0289, 0.0363, 0.0341, 0.0255, 0.0349, 0.0257, 0.0325, 0.0350], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 01:57:03,151 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=151604.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 01:57:21,908 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.196e+02 2.905e+02 3.336e+02 3.698e+02 6.432e+02, threshold=6.672e+02, percent-clipped=0.0 2023-05-16 01:57:25,385 INFO [finetune.py:992] (0/2) Epoch 5, batch 2700, loss[loss=0.1507, simple_loss=0.2344, pruned_loss=0.03348, over 11785.00 frames. ], tot_loss[loss=0.1747, simple_loss=0.2623, pruned_loss=0.04359, over 2379875.71 frames. ], batch size: 26, lr: 4.71e-03, grad_scale: 16.0 2023-05-16 01:57:27,656 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=151638.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 01:57:38,728 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=151654.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 01:58:00,891 INFO [finetune.py:992] (0/2) Epoch 5, batch 2750, loss[loss=0.1801, simple_loss=0.2691, pruned_loss=0.04553, over 11260.00 frames. ], tot_loss[loss=0.1742, simple_loss=0.2617, pruned_loss=0.04334, over 2383340.66 frames. ], batch size: 55, lr: 4.71e-03, grad_scale: 16.0 2023-05-16 01:58:03,268 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=151688.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 01:58:33,494 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.720e+02 2.923e+02 3.297e+02 3.991e+02 1.077e+03, threshold=6.595e+02, percent-clipped=4.0 2023-05-16 01:58:36,333 INFO [finetune.py:992] (0/2) Epoch 5, batch 2800, loss[loss=0.1914, simple_loss=0.2801, pruned_loss=0.05137, over 12356.00 frames. ], tot_loss[loss=0.1748, simple_loss=0.2626, pruned_loss=0.0435, over 2378826.81 frames. ], batch size: 38, lr: 4.71e-03, grad_scale: 32.0 2023-05-16 01:58:46,299 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=151749.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 01:59:12,492 INFO [finetune.py:992] (0/2) Epoch 5, batch 2850, loss[loss=0.1431, simple_loss=0.2227, pruned_loss=0.03181, over 12340.00 frames. ], tot_loss[loss=0.1742, simple_loss=0.2616, pruned_loss=0.04343, over 2385863.26 frames. ], batch size: 30, lr: 4.71e-03, grad_scale: 32.0 2023-05-16 01:59:24,012 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=151801.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 01:59:36,945 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.1470, 3.9482, 3.9387, 4.3683, 3.0009, 3.8714, 2.5183, 4.0406], device='cuda:0'), covar=tensor([0.1517, 0.0689, 0.0931, 0.0561, 0.1045, 0.0591, 0.1827, 0.1202], device='cuda:0'), in_proj_covar=tensor([0.0221, 0.0258, 0.0290, 0.0346, 0.0235, 0.0236, 0.0254, 0.0381], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 01:59:46,491 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.013e+02 3.095e+02 3.461e+02 4.162e+02 1.203e+03, threshold=6.923e+02, percent-clipped=2.0 2023-05-16 01:59:48,470 INFO [finetune.py:992] (0/2) Epoch 5, batch 2900, loss[loss=0.1872, simple_loss=0.2798, pruned_loss=0.04733, over 11803.00 frames. ], tot_loss[loss=0.1741, simple_loss=0.2616, pruned_loss=0.04331, over 2391484.63 frames. ], batch size: 44, lr: 4.71e-03, grad_scale: 16.0 2023-05-16 01:59:51,963 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=151840.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 02:00:07,667 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=151862.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 02:00:23,590 INFO [finetune.py:992] (0/2) Epoch 5, batch 2950, loss[loss=0.1664, simple_loss=0.2459, pruned_loss=0.04343, over 11992.00 frames. ], tot_loss[loss=0.1746, simple_loss=0.2619, pruned_loss=0.0436, over 2381740.67 frames. ], batch size: 28, lr: 4.71e-03, grad_scale: 16.0 2023-05-16 02:00:25,808 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=151888.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 02:00:27,051 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.97 vs. limit=2.0 2023-05-16 02:00:29,160 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.90 vs. limit=2.0 2023-05-16 02:00:57,498 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.930e+02 2.810e+02 3.358e+02 4.108e+02 1.104e+03, threshold=6.716e+02, percent-clipped=4.0 2023-05-16 02:00:59,694 INFO [finetune.py:992] (0/2) Epoch 5, batch 3000, loss[loss=0.1697, simple_loss=0.2592, pruned_loss=0.04012, over 12180.00 frames. ], tot_loss[loss=0.1738, simple_loss=0.2612, pruned_loss=0.04316, over 2384465.29 frames. ], batch size: 35, lr: 4.71e-03, grad_scale: 16.0 2023-05-16 02:00:59,695 INFO [finetune.py:1017] (0/2) Computing validation loss 2023-05-16 02:01:17,520 INFO [finetune.py:1026] (0/2) Epoch 5, validation: loss=0.328, simple_loss=0.4027, pruned_loss=0.1267, over 1020973.00 frames. 2023-05-16 02:01:17,521 INFO [finetune.py:1027] (0/2) Maximum memory allocated so far is 12745MB 2023-05-16 02:01:19,832 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=151938.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 02:01:31,073 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=151954.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 02:01:41,171 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.6585, 2.8687, 4.3693, 4.6980, 3.0004, 2.7140, 3.0296, 2.1474], device='cuda:0'), covar=tensor([0.1373, 0.2700, 0.0515, 0.0403, 0.1131, 0.1975, 0.2301, 0.3671], device='cuda:0'), in_proj_covar=tensor([0.0273, 0.0371, 0.0262, 0.0285, 0.0250, 0.0281, 0.0351, 0.0356], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-16 02:01:53,018 INFO [finetune.py:992] (0/2) Epoch 5, batch 3050, loss[loss=0.1748, simple_loss=0.2535, pruned_loss=0.04804, over 12347.00 frames. ], tot_loss[loss=0.1739, simple_loss=0.2616, pruned_loss=0.04314, over 2369747.22 frames. ], batch size: 30, lr: 4.71e-03, grad_scale: 16.0 2023-05-16 02:01:53,792 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=151986.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 02:02:03,978 INFO [checkpoint.py:75] (0/2) Saving checkpoint to pruned_transducer_stateless7/exp_giga_finetune/checkpoint-52000.pt 2023-05-16 02:02:08,629 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=152002.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 02:02:30,575 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.140e+02 2.948e+02 3.429e+02 4.088e+02 7.361e+02, threshold=6.857e+02, percent-clipped=1.0 2023-05-16 02:02:32,704 INFO [finetune.py:992] (0/2) Epoch 5, batch 3100, loss[loss=0.1509, simple_loss=0.2379, pruned_loss=0.03191, over 12183.00 frames. ], tot_loss[loss=0.1737, simple_loss=0.2613, pruned_loss=0.04303, over 2370048.49 frames. ], batch size: 29, lr: 4.71e-03, grad_scale: 16.0 2023-05-16 02:02:39,064 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=152044.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 02:03:08,992 INFO [finetune.py:992] (0/2) Epoch 5, batch 3150, loss[loss=0.1417, simple_loss=0.2312, pruned_loss=0.02605, over 12013.00 frames. ], tot_loss[loss=0.1736, simple_loss=0.2611, pruned_loss=0.04305, over 2362249.90 frames. ], batch size: 28, lr: 4.71e-03, grad_scale: 16.0 2023-05-16 02:03:42,599 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.792e+02 2.785e+02 3.310e+02 3.978e+02 6.675e+02, threshold=6.621e+02, percent-clipped=0.0 2023-05-16 02:03:44,787 INFO [finetune.py:992] (0/2) Epoch 5, batch 3200, loss[loss=0.1515, simple_loss=0.2345, pruned_loss=0.0342, over 12336.00 frames. ], tot_loss[loss=0.1728, simple_loss=0.2603, pruned_loss=0.04265, over 2366233.88 frames. ], batch size: 30, lr: 4.71e-03, grad_scale: 16.0 2023-05-16 02:03:59,791 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.7030, 2.6007, 3.8178, 4.6217, 3.9302, 4.6628, 4.0154, 3.3570], device='cuda:0'), covar=tensor([0.0026, 0.0328, 0.0116, 0.0036, 0.0105, 0.0043, 0.0089, 0.0257], device='cuda:0'), in_proj_covar=tensor([0.0079, 0.0119, 0.0099, 0.0071, 0.0097, 0.0110, 0.0086, 0.0131], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 02:04:00,400 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=152157.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 02:04:04,665 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.5181, 2.1507, 3.1400, 4.3265, 2.0074, 4.2968, 4.4663, 4.5906], device='cuda:0'), covar=tensor([0.0097, 0.1317, 0.0497, 0.0139, 0.1372, 0.0238, 0.0103, 0.0068], device='cuda:0'), in_proj_covar=tensor([0.0117, 0.0200, 0.0182, 0.0113, 0.0186, 0.0174, 0.0166, 0.0119], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 02:04:18,032 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.6061, 2.4806, 3.2671, 4.5140, 2.1715, 4.4618, 4.5630, 4.7135], device='cuda:0'), covar=tensor([0.0116, 0.1233, 0.0456, 0.0104, 0.1395, 0.0216, 0.0127, 0.0063], device='cuda:0'), in_proj_covar=tensor([0.0117, 0.0200, 0.0182, 0.0113, 0.0186, 0.0174, 0.0166, 0.0119], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 02:04:20,686 INFO [finetune.py:992] (0/2) Epoch 5, batch 3250, loss[loss=0.142, simple_loss=0.2288, pruned_loss=0.02758, over 12180.00 frames. ], tot_loss[loss=0.1728, simple_loss=0.2602, pruned_loss=0.04268, over 2372816.40 frames. ], batch size: 31, lr: 4.71e-03, grad_scale: 16.0 2023-05-16 02:04:50,254 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=152225.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 02:04:51,692 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.0755, 5.0140, 4.8610, 4.9579, 4.5231, 5.0712, 5.0246, 5.3365], device='cuda:0'), covar=tensor([0.0191, 0.0133, 0.0182, 0.0286, 0.0757, 0.0252, 0.0144, 0.0159], device='cuda:0'), in_proj_covar=tensor([0.0180, 0.0184, 0.0180, 0.0233, 0.0232, 0.0199, 0.0165, 0.0214], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0005, 0.0005, 0.0005, 0.0004, 0.0005], device='cuda:0') 2023-05-16 02:04:55,204 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.967e+02 2.801e+02 3.170e+02 3.939e+02 9.922e+02, threshold=6.341e+02, percent-clipped=3.0 2023-05-16 02:04:57,385 INFO [finetune.py:992] (0/2) Epoch 5, batch 3300, loss[loss=0.1488, simple_loss=0.2303, pruned_loss=0.03362, over 12167.00 frames. ], tot_loss[loss=0.1733, simple_loss=0.2609, pruned_loss=0.04281, over 2378090.17 frames. ], batch size: 29, lr: 4.71e-03, grad_scale: 16.0 2023-05-16 02:05:24,212 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=152273.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 02:05:32,445 INFO [finetune.py:992] (0/2) Epoch 5, batch 3350, loss[loss=0.1733, simple_loss=0.2592, pruned_loss=0.04369, over 12321.00 frames. ], tot_loss[loss=0.1735, simple_loss=0.2613, pruned_loss=0.04284, over 2370777.79 frames. ], batch size: 34, lr: 4.70e-03, grad_scale: 16.0 2023-05-16 02:05:33,411 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=152286.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 02:06:06,645 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.167e+02 2.793e+02 3.433e+02 4.351e+02 2.420e+03, threshold=6.867e+02, percent-clipped=7.0 2023-05-16 02:06:08,228 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=152334.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 02:06:08,765 INFO [finetune.py:992] (0/2) Epoch 5, batch 3400, loss[loss=0.1595, simple_loss=0.2432, pruned_loss=0.03787, over 12252.00 frames. ], tot_loss[loss=0.1738, simple_loss=0.2617, pruned_loss=0.04293, over 2371706.51 frames. ], batch size: 32, lr: 4.70e-03, grad_scale: 16.0 2023-05-16 02:06:15,339 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=152344.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 02:06:20,762 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.95 vs. limit=2.0 2023-05-16 02:06:45,190 INFO [finetune.py:992] (0/2) Epoch 5, batch 3450, loss[loss=0.1827, simple_loss=0.2707, pruned_loss=0.04736, over 12115.00 frames. ], tot_loss[loss=0.1729, simple_loss=0.2607, pruned_loss=0.04255, over 2362725.88 frames. ], batch size: 33, lr: 4.70e-03, grad_scale: 16.0 2023-05-16 02:06:47,550 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.8002, 2.9204, 4.8188, 4.9472, 2.9299, 2.8230, 3.1273, 2.2416], device='cuda:0'), covar=tensor([0.1342, 0.2883, 0.0398, 0.0340, 0.1154, 0.1907, 0.2367, 0.3618], device='cuda:0'), in_proj_covar=tensor([0.0272, 0.0369, 0.0260, 0.0284, 0.0249, 0.0279, 0.0348, 0.0353], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-16 02:06:48,282 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2023-05-16 02:06:50,189 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=152392.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 02:07:18,638 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.957e+02 2.749e+02 3.256e+02 3.945e+02 8.450e+02, threshold=6.513e+02, percent-clipped=1.0 2023-05-16 02:07:20,729 INFO [finetune.py:992] (0/2) Epoch 5, batch 3500, loss[loss=0.1433, simple_loss=0.2206, pruned_loss=0.03297, over 11831.00 frames. ], tot_loss[loss=0.1739, simple_loss=0.2619, pruned_loss=0.04294, over 2368011.33 frames. ], batch size: 26, lr: 4.70e-03, grad_scale: 16.0 2023-05-16 02:07:27,885 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=152445.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 02:07:36,951 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=152457.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 02:07:56,775 INFO [finetune.py:992] (0/2) Epoch 5, batch 3550, loss[loss=0.1836, simple_loss=0.2666, pruned_loss=0.05037, over 12350.00 frames. ], tot_loss[loss=0.1742, simple_loss=0.2618, pruned_loss=0.04324, over 2354850.55 frames. ], batch size: 35, lr: 4.70e-03, grad_scale: 16.0 2023-05-16 02:08:10,010 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.93 vs. limit=2.0 2023-05-16 02:08:11,087 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=152505.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 02:08:12,000 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=152506.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 02:08:23,766 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=5.05 vs. limit=5.0 2023-05-16 02:08:31,209 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.938e+02 2.698e+02 3.317e+02 3.795e+02 6.447e+02, threshold=6.634e+02, percent-clipped=0.0 2023-05-16 02:08:33,456 INFO [finetune.py:992] (0/2) Epoch 5, batch 3600, loss[loss=0.1537, simple_loss=0.2361, pruned_loss=0.03562, over 12176.00 frames. ], tot_loss[loss=0.1739, simple_loss=0.2614, pruned_loss=0.04322, over 2360050.01 frames. ], batch size: 31, lr: 4.70e-03, grad_scale: 16.0 2023-05-16 02:09:06,110 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=152581.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 02:09:06,898 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=152582.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 02:09:08,895 INFO [finetune.py:992] (0/2) Epoch 5, batch 3650, loss[loss=0.1579, simple_loss=0.2491, pruned_loss=0.03337, over 12293.00 frames. ], tot_loss[loss=0.1738, simple_loss=0.2613, pruned_loss=0.04312, over 2363863.75 frames. ], batch size: 34, lr: 4.70e-03, grad_scale: 16.0 2023-05-16 02:09:11,861 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=152589.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 02:09:25,600 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.3893, 4.8639, 5.3303, 4.6352, 4.9764, 4.7471, 5.3649, 4.9746], device='cuda:0'), covar=tensor([0.0220, 0.0365, 0.0235, 0.0249, 0.0317, 0.0285, 0.0188, 0.0301], device='cuda:0'), in_proj_covar=tensor([0.0234, 0.0241, 0.0260, 0.0235, 0.0232, 0.0233, 0.0212, 0.0196], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 02:09:41,261 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=152629.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 02:09:43,326 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.139e+02 2.862e+02 3.312e+02 3.947e+02 9.238e+02, threshold=6.625e+02, percent-clipped=2.0 2023-05-16 02:09:45,490 INFO [finetune.py:992] (0/2) Epoch 5, batch 3700, loss[loss=0.1579, simple_loss=0.2327, pruned_loss=0.04153, over 11827.00 frames. ], tot_loss[loss=0.1736, simple_loss=0.2614, pruned_loss=0.0429, over 2371576.58 frames. ], batch size: 26, lr: 4.70e-03, grad_scale: 16.0 2023-05-16 02:09:51,431 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=152643.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 02:09:56,394 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=152650.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 02:10:22,095 INFO [finetune.py:992] (0/2) Epoch 5, batch 3750, loss[loss=0.1646, simple_loss=0.2607, pruned_loss=0.03426, over 10707.00 frames. ], tot_loss[loss=0.1735, simple_loss=0.2611, pruned_loss=0.04301, over 2369759.30 frames. ], batch size: 68, lr: 4.70e-03, grad_scale: 16.0 2023-05-16 02:10:51,729 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=152726.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 02:10:55,754 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.124e+02 2.824e+02 3.202e+02 3.601e+02 7.259e+02, threshold=6.404e+02, percent-clipped=1.0 2023-05-16 02:10:57,913 INFO [finetune.py:992] (0/2) Epoch 5, batch 3800, loss[loss=0.1794, simple_loss=0.2709, pruned_loss=0.04397, over 12113.00 frames. ], tot_loss[loss=0.1739, simple_loss=0.2615, pruned_loss=0.04311, over 2374397.18 frames. ], batch size: 38, lr: 4.70e-03, grad_scale: 16.0 2023-05-16 02:11:02,344 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([6.3371, 6.1526, 6.0735, 5.5105, 5.3445, 6.1840, 5.7737, 5.5127], device='cuda:0'), covar=tensor([0.0531, 0.1016, 0.0630, 0.1480, 0.0565, 0.0663, 0.1309, 0.1038], device='cuda:0'), in_proj_covar=tensor([0.0583, 0.0512, 0.0490, 0.0599, 0.0385, 0.0672, 0.0735, 0.0542], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0004, 0.0004, 0.0003], device='cuda:0') 2023-05-16 02:11:33,959 INFO [finetune.py:992] (0/2) Epoch 5, batch 3850, loss[loss=0.1601, simple_loss=0.2483, pruned_loss=0.03595, over 12098.00 frames. ], tot_loss[loss=0.1744, simple_loss=0.2623, pruned_loss=0.04328, over 2374917.80 frames. ], batch size: 33, lr: 4.70e-03, grad_scale: 16.0 2023-05-16 02:11:35,599 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=152787.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 02:11:37,069 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.2730, 2.3223, 3.0886, 4.1231, 2.2555, 4.2113, 4.1366, 4.3576], device='cuda:0'), covar=tensor([0.0122, 0.1206, 0.0474, 0.0136, 0.1231, 0.0214, 0.0179, 0.0089], device='cuda:0'), in_proj_covar=tensor([0.0118, 0.0199, 0.0182, 0.0114, 0.0186, 0.0174, 0.0167, 0.0120], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 02:11:46,298 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=152801.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 02:12:08,245 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.037e+02 2.857e+02 3.300e+02 4.012e+02 8.037e+02, threshold=6.600e+02, percent-clipped=3.0 2023-05-16 02:12:10,397 INFO [finetune.py:992] (0/2) Epoch 5, batch 3900, loss[loss=0.1736, simple_loss=0.2684, pruned_loss=0.03943, over 11996.00 frames. ], tot_loss[loss=0.1749, simple_loss=0.2623, pruned_loss=0.0437, over 2366418.89 frames. ], batch size: 40, lr: 4.70e-03, grad_scale: 16.0 2023-05-16 02:12:38,887 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=3.41 vs. limit=5.0 2023-05-16 02:12:43,690 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=152881.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 02:12:46,370 INFO [finetune.py:992] (0/2) Epoch 5, batch 3950, loss[loss=0.1426, simple_loss=0.2305, pruned_loss=0.02736, over 12125.00 frames. ], tot_loss[loss=0.1739, simple_loss=0.2614, pruned_loss=0.04318, over 2370380.22 frames. ], batch size: 30, lr: 4.70e-03, grad_scale: 16.0 2023-05-16 02:13:18,183 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=152929.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 02:13:18,228 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=152929.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 02:13:20,243 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.976e+02 2.843e+02 3.254e+02 3.920e+02 8.560e+02, threshold=6.509e+02, percent-clipped=1.0 2023-05-16 02:13:22,298 INFO [finetune.py:992] (0/2) Epoch 5, batch 4000, loss[loss=0.2091, simple_loss=0.2941, pruned_loss=0.06211, over 10378.00 frames. ], tot_loss[loss=0.1748, simple_loss=0.2626, pruned_loss=0.04356, over 2362763.84 frames. ], batch size: 68, lr: 4.70e-03, grad_scale: 16.0 2023-05-16 02:13:24,892 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=152938.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 02:13:29,741 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=152945.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 02:13:52,675 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=152977.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 02:13:58,469 INFO [finetune.py:992] (0/2) Epoch 5, batch 4050, loss[loss=0.1532, simple_loss=0.2419, pruned_loss=0.03228, over 12239.00 frames. ], tot_loss[loss=0.1757, simple_loss=0.2634, pruned_loss=0.04397, over 2359102.06 frames. ], batch size: 32, lr: 4.70e-03, grad_scale: 16.0 2023-05-16 02:14:31,793 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.109e+02 3.060e+02 3.470e+02 4.175e+02 7.154e+02, threshold=6.940e+02, percent-clipped=2.0 2023-05-16 02:14:33,952 INFO [finetune.py:992] (0/2) Epoch 5, batch 4100, loss[loss=0.1528, simple_loss=0.2504, pruned_loss=0.02765, over 12368.00 frames. ], tot_loss[loss=0.1752, simple_loss=0.2629, pruned_loss=0.04378, over 2355203.72 frames. ], batch size: 35, lr: 4.70e-03, grad_scale: 16.0 2023-05-16 02:15:08,760 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=153082.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 02:15:10,811 INFO [finetune.py:992] (0/2) Epoch 5, batch 4150, loss[loss=0.1659, simple_loss=0.2727, pruned_loss=0.02958, over 12289.00 frames. ], tot_loss[loss=0.175, simple_loss=0.2628, pruned_loss=0.0436, over 2359584.36 frames. ], batch size: 37, lr: 4.70e-03, grad_scale: 16.0 2023-05-16 02:15:22,151 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=153101.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 02:15:27,931 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.5801, 2.3812, 3.3455, 4.3619, 2.2864, 4.4256, 4.5109, 4.6182], device='cuda:0'), covar=tensor([0.0109, 0.1285, 0.0474, 0.0166, 0.1345, 0.0264, 0.0138, 0.0088], device='cuda:0'), in_proj_covar=tensor([0.0119, 0.0205, 0.0186, 0.0117, 0.0190, 0.0178, 0.0171, 0.0123], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 02:15:33,609 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=153117.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 02:15:44,066 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.883e+02 3.037e+02 3.431e+02 4.052e+02 6.200e+02, threshold=6.862e+02, percent-clipped=0.0 2023-05-16 02:15:46,155 INFO [finetune.py:992] (0/2) Epoch 5, batch 4200, loss[loss=0.1951, simple_loss=0.2831, pruned_loss=0.05357, over 12042.00 frames. ], tot_loss[loss=0.1759, simple_loss=0.2638, pruned_loss=0.044, over 2357365.71 frames. ], batch size: 40, lr: 4.70e-03, grad_scale: 16.0 2023-05-16 02:15:56,133 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=153149.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 02:16:17,204 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=153178.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 02:16:21,886 INFO [finetune.py:992] (0/2) Epoch 5, batch 4250, loss[loss=0.1752, simple_loss=0.2657, pruned_loss=0.04231, over 12092.00 frames. ], tot_loss[loss=0.1752, simple_loss=0.2628, pruned_loss=0.04377, over 2359418.65 frames. ], batch size: 32, lr: 4.70e-03, grad_scale: 16.0 2023-05-16 02:16:56,568 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.049e+02 2.788e+02 3.231e+02 3.915e+02 1.320e+03, threshold=6.462e+02, percent-clipped=1.0 2023-05-16 02:16:58,686 INFO [finetune.py:992] (0/2) Epoch 5, batch 4300, loss[loss=0.1563, simple_loss=0.2444, pruned_loss=0.03412, over 12101.00 frames. ], tot_loss[loss=0.1741, simple_loss=0.2616, pruned_loss=0.04331, over 2373395.37 frames. ], batch size: 33, lr: 4.70e-03, grad_scale: 16.0 2023-05-16 02:17:00,993 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=153238.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 02:17:05,902 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=153245.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 02:17:29,195 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=153278.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 02:17:33,852 INFO [finetune.py:992] (0/2) Epoch 5, batch 4350, loss[loss=0.2012, simple_loss=0.3004, pruned_loss=0.05104, over 12046.00 frames. ], tot_loss[loss=0.1748, simple_loss=0.2623, pruned_loss=0.04369, over 2371580.79 frames. ], batch size: 37, lr: 4.70e-03, grad_scale: 16.0 2023-05-16 02:17:34,676 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=153286.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 02:17:39,579 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=153293.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 02:17:54,627 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.6882, 3.7098, 3.4342, 3.2932, 3.0316, 2.9943, 3.8044, 2.5417], device='cuda:0'), covar=tensor([0.0287, 0.0134, 0.0140, 0.0152, 0.0339, 0.0273, 0.0102, 0.0379], device='cuda:0'), in_proj_covar=tensor([0.0185, 0.0156, 0.0150, 0.0180, 0.0198, 0.0192, 0.0158, 0.0199], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 02:17:55,317 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.7759, 4.5039, 4.4304, 4.6902, 4.4384, 4.6640, 4.5232, 2.2663], device='cuda:0'), covar=tensor([0.0188, 0.0111, 0.0158, 0.0143, 0.0105, 0.0206, 0.0162, 0.1046], device='cuda:0'), in_proj_covar=tensor([0.0065, 0.0073, 0.0077, 0.0070, 0.0057, 0.0088, 0.0076, 0.0092], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 02:18:07,773 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.085e+02 2.827e+02 3.260e+02 4.214e+02 1.164e+03, threshold=6.520e+02, percent-clipped=1.0 2023-05-16 02:18:09,894 INFO [finetune.py:992] (0/2) Epoch 5, batch 4400, loss[loss=0.1835, simple_loss=0.2712, pruned_loss=0.04791, over 11309.00 frames. ], tot_loss[loss=0.1752, simple_loss=0.2622, pruned_loss=0.04405, over 2355061.98 frames. ], batch size: 55, lr: 4.69e-03, grad_scale: 16.0 2023-05-16 02:18:12,988 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=153339.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 02:18:16,526 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.1024, 5.9696, 5.5614, 5.4771, 6.0312, 5.3690, 5.6268, 5.5665], device='cuda:0'), covar=tensor([0.1419, 0.0975, 0.0996, 0.2012, 0.1055, 0.2258, 0.1563, 0.1119], device='cuda:0'), in_proj_covar=tensor([0.0337, 0.0460, 0.0366, 0.0409, 0.0443, 0.0417, 0.0376, 0.0359], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 02:18:26,466 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.9353, 4.8633, 4.6983, 4.8277, 4.3791, 4.9496, 4.8285, 5.2176], device='cuda:0'), covar=tensor([0.0209, 0.0151, 0.0198, 0.0276, 0.0810, 0.0333, 0.0175, 0.0169], device='cuda:0'), in_proj_covar=tensor([0.0180, 0.0184, 0.0181, 0.0232, 0.0232, 0.0200, 0.0164, 0.0212], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0005, 0.0005, 0.0005, 0.0004, 0.0005], device='cuda:0') 2023-05-16 02:18:44,241 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=153382.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 02:18:46,221 INFO [finetune.py:992] (0/2) Epoch 5, batch 4450, loss[loss=0.1668, simple_loss=0.2653, pruned_loss=0.03413, over 12080.00 frames. ], tot_loss[loss=0.1737, simple_loss=0.2611, pruned_loss=0.04319, over 2367795.93 frames. ], batch size: 42, lr: 4.69e-03, grad_scale: 16.0 2023-05-16 02:19:12,112 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=153421.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 02:19:18,357 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=153430.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 02:19:19,750 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.237e+02 2.916e+02 3.318e+02 4.100e+02 6.864e+02, threshold=6.635e+02, percent-clipped=2.0 2023-05-16 02:19:21,849 INFO [finetune.py:992] (0/2) Epoch 5, batch 4500, loss[loss=0.1736, simple_loss=0.2635, pruned_loss=0.04184, over 12111.00 frames. ], tot_loss[loss=0.1749, simple_loss=0.2619, pruned_loss=0.04391, over 2364008.10 frames. ], batch size: 33, lr: 4.69e-03, grad_scale: 16.0 2023-05-16 02:19:49,055 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=153473.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 02:19:56,151 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=153482.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 02:19:58,023 INFO [finetune.py:992] (0/2) Epoch 5, batch 4550, loss[loss=0.1678, simple_loss=0.2623, pruned_loss=0.03665, over 12156.00 frames. ], tot_loss[loss=0.1748, simple_loss=0.2621, pruned_loss=0.04368, over 2369932.72 frames. ], batch size: 36, lr: 4.69e-03, grad_scale: 16.0 2023-05-16 02:20:09,708 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=5.33 vs. limit=5.0 2023-05-16 02:20:31,536 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.058e+02 2.893e+02 3.389e+02 4.281e+02 1.179e+03, threshold=6.777e+02, percent-clipped=3.0 2023-05-16 02:20:33,611 INFO [finetune.py:992] (0/2) Epoch 5, batch 4600, loss[loss=0.2206, simple_loss=0.3008, pruned_loss=0.07017, over 11838.00 frames. ], tot_loss[loss=0.1753, simple_loss=0.2627, pruned_loss=0.04393, over 2371196.30 frames. ], batch size: 44, lr: 4.69e-03, grad_scale: 16.0 2023-05-16 02:20:35,924 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.5760, 4.1109, 4.3309, 4.4113, 3.1972, 4.0335, 2.8518, 4.1061], device='cuda:0'), covar=tensor([0.1418, 0.0724, 0.0749, 0.0598, 0.1164, 0.0590, 0.1709, 0.1170], device='cuda:0'), in_proj_covar=tensor([0.0220, 0.0258, 0.0289, 0.0346, 0.0235, 0.0232, 0.0253, 0.0377], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 02:21:08,775 INFO [finetune.py:992] (0/2) Epoch 5, batch 4650, loss[loss=0.183, simple_loss=0.2602, pruned_loss=0.05292, over 12325.00 frames. ], tot_loss[loss=0.1743, simple_loss=0.2617, pruned_loss=0.04346, over 2374394.98 frames. ], batch size: 31, lr: 4.69e-03, grad_scale: 16.0 2023-05-16 02:21:08,978 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.1575, 4.7808, 4.9152, 5.1475, 4.7163, 4.9971, 4.7866, 2.8181], device='cuda:0'), covar=tensor([0.0098, 0.0059, 0.0074, 0.0048, 0.0053, 0.0083, 0.0079, 0.0656], device='cuda:0'), in_proj_covar=tensor([0.0065, 0.0072, 0.0076, 0.0069, 0.0057, 0.0088, 0.0075, 0.0091], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 02:21:43,116 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.786e+02 2.784e+02 3.285e+02 3.986e+02 8.403e+02, threshold=6.570e+02, percent-clipped=1.0 2023-05-16 02:21:44,625 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=153634.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 02:21:45,314 INFO [finetune.py:992] (0/2) Epoch 5, batch 4700, loss[loss=0.1816, simple_loss=0.2776, pruned_loss=0.04286, over 12152.00 frames. ], tot_loss[loss=0.1747, simple_loss=0.2622, pruned_loss=0.04359, over 2377338.14 frames. ], batch size: 36, lr: 4.69e-03, grad_scale: 16.0 2023-05-16 02:22:21,130 INFO [finetune.py:992] (0/2) Epoch 5, batch 4750, loss[loss=0.2092, simple_loss=0.2911, pruned_loss=0.06363, over 12040.00 frames. ], tot_loss[loss=0.1753, simple_loss=0.2628, pruned_loss=0.04395, over 2372409.06 frames. ], batch size: 42, lr: 4.69e-03, grad_scale: 16.0 2023-05-16 02:22:29,862 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.2919, 6.1315, 5.7120, 5.6783, 6.2403, 5.5068, 5.8308, 5.6896], device='cuda:0'), covar=tensor([0.1383, 0.0991, 0.1079, 0.2027, 0.0865, 0.2168, 0.1448, 0.1119], device='cuda:0'), in_proj_covar=tensor([0.0341, 0.0466, 0.0370, 0.0416, 0.0445, 0.0423, 0.0380, 0.0364], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 02:22:37,306 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.96 vs. limit=2.0 2023-05-16 02:22:45,562 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([6.1465, 6.0783, 5.8664, 5.5000, 5.1740, 6.0262, 5.6520, 5.3544], device='cuda:0'), covar=tensor([0.0506, 0.0858, 0.0592, 0.1542, 0.0604, 0.0575, 0.1271, 0.0969], device='cuda:0'), in_proj_covar=tensor([0.0590, 0.0519, 0.0495, 0.0610, 0.0393, 0.0679, 0.0745, 0.0547], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0004, 0.0004, 0.0003], device='cuda:0') 2023-05-16 02:22:54,599 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.975e+02 2.771e+02 3.141e+02 3.707e+02 1.127e+03, threshold=6.281e+02, percent-clipped=2.0 2023-05-16 02:22:56,736 INFO [finetune.py:992] (0/2) Epoch 5, batch 4800, loss[loss=0.1547, simple_loss=0.2348, pruned_loss=0.03729, over 12282.00 frames. ], tot_loss[loss=0.1749, simple_loss=0.2619, pruned_loss=0.04391, over 2368956.70 frames. ], batch size: 28, lr: 4.69e-03, grad_scale: 16.0 2023-05-16 02:22:58,659 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.88 vs. limit=2.0 2023-05-16 02:23:24,871 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=153773.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 02:23:26,574 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=4.58 vs. limit=5.0 2023-05-16 02:23:27,585 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=153777.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 02:23:32,992 INFO [finetune.py:992] (0/2) Epoch 5, batch 4850, loss[loss=0.188, simple_loss=0.2785, pruned_loss=0.04876, over 11803.00 frames. ], tot_loss[loss=0.1751, simple_loss=0.2623, pruned_loss=0.04396, over 2373905.52 frames. ], batch size: 44, lr: 4.69e-03, grad_scale: 16.0 2023-05-16 02:23:55,207 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=3.37 vs. limit=5.0 2023-05-16 02:23:59,786 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=153821.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 02:24:06,071 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.2637, 6.0981, 5.6462, 5.7417, 6.1724, 5.6021, 5.7077, 5.7602], device='cuda:0'), covar=tensor([0.1443, 0.0808, 0.0879, 0.1610, 0.0963, 0.1861, 0.1649, 0.0950], device='cuda:0'), in_proj_covar=tensor([0.0343, 0.0469, 0.0372, 0.0418, 0.0446, 0.0423, 0.0382, 0.0363], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 02:24:07,332 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.927e+02 2.827e+02 3.174e+02 3.763e+02 6.877e+02, threshold=6.348e+02, percent-clipped=0.0 2023-05-16 02:24:09,555 INFO [finetune.py:992] (0/2) Epoch 5, batch 4900, loss[loss=0.1998, simple_loss=0.2893, pruned_loss=0.0551, over 12346.00 frames. ], tot_loss[loss=0.1753, simple_loss=0.2625, pruned_loss=0.04411, over 2378800.22 frames. ], batch size: 36, lr: 4.69e-03, grad_scale: 32.0 2023-05-16 02:24:35,407 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.43 vs. limit=2.0 2023-05-16 02:24:44,825 INFO [finetune.py:992] (0/2) Epoch 5, batch 4950, loss[loss=0.1992, simple_loss=0.2937, pruned_loss=0.05229, over 12127.00 frames. ], tot_loss[loss=0.175, simple_loss=0.2624, pruned_loss=0.04376, over 2384261.22 frames. ], batch size: 39, lr: 4.69e-03, grad_scale: 32.0 2023-05-16 02:25:00,294 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.9980, 5.7368, 5.3310, 5.3794, 5.8911, 5.1648, 5.4707, 5.4275], device='cuda:0'), covar=tensor([0.1213, 0.0925, 0.1014, 0.1761, 0.1008, 0.2163, 0.1549, 0.1086], device='cuda:0'), in_proj_covar=tensor([0.0343, 0.0471, 0.0374, 0.0419, 0.0448, 0.0425, 0.0382, 0.0365], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 02:25:12,367 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=153923.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 02:25:18,478 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.582e+02 2.987e+02 3.492e+02 4.100e+02 7.921e+02, threshold=6.983e+02, percent-clipped=7.0 2023-05-16 02:25:19,977 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=153934.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 02:25:20,603 INFO [finetune.py:992] (0/2) Epoch 5, batch 5000, loss[loss=0.1801, simple_loss=0.2744, pruned_loss=0.04287, over 11636.00 frames. ], tot_loss[loss=0.1749, simple_loss=0.2624, pruned_loss=0.04373, over 2377351.77 frames. ], batch size: 48, lr: 4.69e-03, grad_scale: 32.0 2023-05-16 02:25:54,914 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=153982.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 02:25:56,518 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=153984.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 02:25:57,018 INFO [finetune.py:992] (0/2) Epoch 5, batch 5050, loss[loss=0.1801, simple_loss=0.2675, pruned_loss=0.04636, over 11591.00 frames. ], tot_loss[loss=0.1752, simple_loss=0.2626, pruned_loss=0.04393, over 2368165.41 frames. ], batch size: 48, lr: 4.69e-03, grad_scale: 16.0 2023-05-16 02:26:08,010 INFO [checkpoint.py:75] (0/2) Saving checkpoint to pruned_transducer_stateless7/exp_giga_finetune/checkpoint-54000.pt 2023-05-16 02:26:12,543 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.0920, 2.5465, 3.7228, 3.0989, 3.5090, 3.1899, 2.5793, 3.6417], device='cuda:0'), covar=tensor([0.0130, 0.0330, 0.0122, 0.0232, 0.0124, 0.0156, 0.0370, 0.0111], device='cuda:0'), in_proj_covar=tensor([0.0166, 0.0198, 0.0176, 0.0176, 0.0199, 0.0153, 0.0190, 0.0169], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 02:26:34,374 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.141e+02 2.897e+02 3.384e+02 4.051e+02 8.052e+02, threshold=6.767e+02, percent-clipped=3.0 2023-05-16 02:26:35,819 INFO [finetune.py:992] (0/2) Epoch 5, batch 5100, loss[loss=0.1827, simple_loss=0.2735, pruned_loss=0.04592, over 12127.00 frames. ], tot_loss[loss=0.1749, simple_loss=0.2623, pruned_loss=0.04374, over 2369003.20 frames. ], batch size: 42, lr: 4.69e-03, grad_scale: 16.0 2023-05-16 02:26:46,792 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=154049.0, num_to_drop=1, layers_to_drop={1} 2023-05-16 02:27:05,953 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=154076.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 02:27:06,659 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=154077.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 02:27:11,582 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=154084.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 02:27:12,098 INFO [finetune.py:992] (0/2) Epoch 5, batch 5150, loss[loss=0.1419, simple_loss=0.2254, pruned_loss=0.02925, over 12273.00 frames. ], tot_loss[loss=0.1754, simple_loss=0.2632, pruned_loss=0.04383, over 2369286.90 frames. ], batch size: 28, lr: 4.69e-03, grad_scale: 16.0 2023-05-16 02:27:30,745 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=154110.0, num_to_drop=1, layers_to_drop={0} 2023-05-16 02:27:36,346 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.8264, 3.8321, 3.7833, 3.9099, 3.5345, 3.5195, 3.5990, 3.8004], device='cuda:0'), covar=tensor([0.1697, 0.0909, 0.2162, 0.0991, 0.2476, 0.2076, 0.0809, 0.1233], device='cuda:0'), in_proj_covar=tensor([0.0498, 0.0639, 0.0555, 0.0596, 0.0790, 0.0704, 0.0519, 0.0475], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0005, 0.0004, 0.0004, 0.0005, 0.0005, 0.0004, 0.0004], device='cuda:0') 2023-05-16 02:27:41,113 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=154125.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 02:27:46,557 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.052e+02 2.960e+02 3.349e+02 3.965e+02 6.910e+02, threshold=6.698e+02, percent-clipped=1.0 2023-05-16 02:27:48,071 INFO [finetune.py:992] (0/2) Epoch 5, batch 5200, loss[loss=0.166, simple_loss=0.2445, pruned_loss=0.0438, over 11988.00 frames. ], tot_loss[loss=0.1758, simple_loss=0.2636, pruned_loss=0.04397, over 2363432.63 frames. ], batch size: 28, lr: 4.69e-03, grad_scale: 16.0 2023-05-16 02:27:49,672 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=154137.0, num_to_drop=1, layers_to_drop={0} 2023-05-16 02:27:55,415 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=154145.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 02:28:04,544 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.1103, 5.9817, 5.5292, 5.5713, 6.0819, 5.3946, 5.6114, 5.6125], device='cuda:0'), covar=tensor([0.1219, 0.0895, 0.0862, 0.1679, 0.0847, 0.2039, 0.1536, 0.1088], device='cuda:0'), in_proj_covar=tensor([0.0341, 0.0468, 0.0374, 0.0416, 0.0447, 0.0423, 0.0380, 0.0366], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 02:28:06,604 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.2392, 6.0404, 5.5957, 5.5874, 6.1529, 5.4715, 5.6834, 5.6566], device='cuda:0'), covar=tensor([0.1340, 0.0854, 0.0828, 0.1661, 0.0839, 0.1938, 0.1403, 0.1082], device='cuda:0'), in_proj_covar=tensor([0.0341, 0.0468, 0.0374, 0.0416, 0.0447, 0.0423, 0.0380, 0.0366], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 02:28:23,649 INFO [finetune.py:992] (0/2) Epoch 5, batch 5250, loss[loss=0.1668, simple_loss=0.2577, pruned_loss=0.03799, over 12343.00 frames. ], tot_loss[loss=0.1745, simple_loss=0.2621, pruned_loss=0.04343, over 2368990.33 frames. ], batch size: 36, lr: 4.69e-03, grad_scale: 16.0 2023-05-16 02:28:50,151 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=154221.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 02:28:58,487 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.993e+02 2.850e+02 3.342e+02 3.874e+02 6.540e+02, threshold=6.683e+02, percent-clipped=0.0 2023-05-16 02:28:59,940 INFO [finetune.py:992] (0/2) Epoch 5, batch 5300, loss[loss=0.1717, simple_loss=0.2643, pruned_loss=0.03954, over 12262.00 frames. ], tot_loss[loss=0.1746, simple_loss=0.2624, pruned_loss=0.04344, over 2376482.81 frames. ], batch size: 37, lr: 4.69e-03, grad_scale: 16.0 2023-05-16 02:29:31,844 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=154279.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 02:29:34,023 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=154282.0, num_to_drop=1, layers_to_drop={0} 2023-05-16 02:29:35,961 INFO [finetune.py:992] (0/2) Epoch 5, batch 5350, loss[loss=0.1639, simple_loss=0.257, pruned_loss=0.03541, over 12264.00 frames. ], tot_loss[loss=0.1749, simple_loss=0.2624, pruned_loss=0.04372, over 2363994.47 frames. ], batch size: 37, lr: 4.69e-03, grad_scale: 16.0 2023-05-16 02:29:36,854 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=154286.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 02:29:40,338 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=154291.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 02:30:07,092 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.53 vs. limit=2.0 2023-05-16 02:30:09,562 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.028e+02 2.896e+02 3.518e+02 4.242e+02 8.242e+02, threshold=7.036e+02, percent-clipped=1.0 2023-05-16 02:30:11,027 INFO [finetune.py:992] (0/2) Epoch 5, batch 5400, loss[loss=0.1556, simple_loss=0.2441, pruned_loss=0.03353, over 12335.00 frames. ], tot_loss[loss=0.1753, simple_loss=0.2628, pruned_loss=0.04392, over 2358616.16 frames. ], batch size: 31, lr: 4.69e-03, grad_scale: 16.0 2023-05-16 02:30:12,609 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=154337.0, num_to_drop=1, layers_to_drop={0} 2023-05-16 02:30:18,872 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.6307, 4.2693, 4.5881, 4.0827, 4.3185, 4.1279, 4.6063, 4.1466], device='cuda:0'), covar=tensor([0.0258, 0.0348, 0.0272, 0.0237, 0.0321, 0.0304, 0.0214, 0.0696], device='cuda:0'), in_proj_covar=tensor([0.0235, 0.0240, 0.0260, 0.0234, 0.0233, 0.0233, 0.0212, 0.0198], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 02:30:20,269 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=154347.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 02:30:23,879 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=154352.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 02:30:47,171 INFO [finetune.py:992] (0/2) Epoch 5, batch 5450, loss[loss=0.2037, simple_loss=0.2885, pruned_loss=0.05943, over 12131.00 frames. ], tot_loss[loss=0.1759, simple_loss=0.263, pruned_loss=0.04436, over 2344292.59 frames. ], batch size: 39, lr: 4.68e-03, grad_scale: 16.0 2023-05-16 02:30:57,401 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=154398.0, num_to_drop=1, layers_to_drop={1} 2023-05-16 02:31:02,442 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=154405.0, num_to_drop=1, layers_to_drop={1} 2023-05-16 02:31:21,676 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=154432.0, num_to_drop=1, layers_to_drop={2} 2023-05-16 02:31:22,248 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.125e+02 2.884e+02 3.458e+02 3.962e+02 7.721e+02, threshold=6.916e+02, percent-clipped=3.0 2023-05-16 02:31:23,712 INFO [finetune.py:992] (0/2) Epoch 5, batch 5500, loss[loss=0.1801, simple_loss=0.2772, pruned_loss=0.04151, over 11591.00 frames. ], tot_loss[loss=0.1744, simple_loss=0.2617, pruned_loss=0.04357, over 2358823.18 frames. ], batch size: 48, lr: 4.68e-03, grad_scale: 16.0 2023-05-16 02:31:27,378 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=154440.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 02:31:50,346 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=2.91 vs. limit=5.0 2023-05-16 02:31:51,516 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=154474.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 02:31:59,094 INFO [finetune.py:992] (0/2) Epoch 5, batch 5550, loss[loss=0.1424, simple_loss=0.2261, pruned_loss=0.02936, over 12007.00 frames. ], tot_loss[loss=0.1752, simple_loss=0.2624, pruned_loss=0.04401, over 2361586.47 frames. ], batch size: 28, lr: 4.68e-03, grad_scale: 16.0 2023-05-16 02:32:34,218 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.964e+02 2.833e+02 3.428e+02 4.181e+02 7.252e+02, threshold=6.856e+02, percent-clipped=1.0 2023-05-16 02:32:35,630 INFO [finetune.py:992] (0/2) Epoch 5, batch 5600, loss[loss=0.1882, simple_loss=0.2783, pruned_loss=0.04904, over 12170.00 frames. ], tot_loss[loss=0.1762, simple_loss=0.2633, pruned_loss=0.04453, over 2348154.50 frames. ], batch size: 35, lr: 4.68e-03, grad_scale: 16.0 2023-05-16 02:32:35,825 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.0641, 4.7575, 4.8396, 4.8599, 4.6164, 4.9588, 4.7255, 2.7232], device='cuda:0'), covar=tensor([0.0093, 0.0066, 0.0076, 0.0069, 0.0057, 0.0071, 0.0138, 0.0667], device='cuda:0'), in_proj_covar=tensor([0.0065, 0.0072, 0.0076, 0.0070, 0.0058, 0.0087, 0.0075, 0.0091], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 02:32:35,867 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=154535.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 02:32:39,391 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.2066, 2.4051, 3.3129, 4.2265, 3.6809, 4.1520, 3.7199, 3.0172], device='cuda:0'), covar=tensor([0.0031, 0.0379, 0.0124, 0.0036, 0.0126, 0.0064, 0.0087, 0.0315], device='cuda:0'), in_proj_covar=tensor([0.0080, 0.0120, 0.0099, 0.0073, 0.0098, 0.0109, 0.0086, 0.0133], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 02:33:05,508 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=154577.0, num_to_drop=1, layers_to_drop={1} 2023-05-16 02:33:06,981 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=154579.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 02:33:11,135 INFO [finetune.py:992] (0/2) Epoch 5, batch 5650, loss[loss=0.1564, simple_loss=0.2417, pruned_loss=0.03549, over 12365.00 frames. ], tot_loss[loss=0.1758, simple_loss=0.263, pruned_loss=0.04426, over 2358042.51 frames. ], batch size: 30, lr: 4.68e-03, grad_scale: 16.0 2023-05-16 02:33:29,131 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.94 vs. limit=2.0 2023-05-16 02:33:40,779 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=154627.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 02:33:41,612 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=154628.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 02:33:45,680 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.024e+02 2.849e+02 3.278e+02 3.834e+02 6.939e+02, threshold=6.556e+02, percent-clipped=1.0 2023-05-16 02:33:47,071 INFO [finetune.py:992] (0/2) Epoch 5, batch 5700, loss[loss=0.1778, simple_loss=0.2703, pruned_loss=0.0427, over 12247.00 frames. ], tot_loss[loss=0.1778, simple_loss=0.265, pruned_loss=0.04529, over 2359003.38 frames. ], batch size: 32, lr: 4.68e-03, grad_scale: 16.0 2023-05-16 02:33:47,302 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.4035, 3.2884, 4.8530, 2.5622, 2.7653, 3.6414, 3.2292, 3.7404], device='cuda:0'), covar=tensor([0.0452, 0.0981, 0.0275, 0.1145, 0.1723, 0.1272, 0.1197, 0.1055], device='cuda:0'), in_proj_covar=tensor([0.0233, 0.0228, 0.0231, 0.0181, 0.0235, 0.0281, 0.0224, 0.0248], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:0') 2023-05-16 02:33:52,032 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=154642.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 02:33:55,645 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=154647.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 02:34:23,025 INFO [finetune.py:992] (0/2) Epoch 5, batch 5750, loss[loss=0.1787, simple_loss=0.2583, pruned_loss=0.04952, over 12346.00 frames. ], tot_loss[loss=0.1775, simple_loss=0.265, pruned_loss=0.04503, over 2364626.98 frames. ], batch size: 31, lr: 4.68e-03, grad_scale: 16.0 2023-05-16 02:34:26,137 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=154689.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 02:34:28,946 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=154693.0, num_to_drop=1, layers_to_drop={2} 2023-05-16 02:34:37,328 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=154705.0, num_to_drop=1, layers_to_drop={1} 2023-05-16 02:34:43,252 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.1492, 2.6296, 3.7396, 3.0352, 3.4925, 3.1813, 2.5397, 3.5941], device='cuda:0'), covar=tensor([0.0133, 0.0318, 0.0122, 0.0254, 0.0156, 0.0174, 0.0361, 0.0100], device='cuda:0'), in_proj_covar=tensor([0.0165, 0.0197, 0.0175, 0.0175, 0.0199, 0.0152, 0.0189, 0.0168], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 02:34:48,107 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=154720.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 02:34:56,622 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=154732.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 02:34:57,130 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.863e+02 2.814e+02 3.331e+02 3.980e+02 7.563e+02, threshold=6.661e+02, percent-clipped=4.0 2023-05-16 02:34:58,565 INFO [finetune.py:992] (0/2) Epoch 5, batch 5800, loss[loss=0.1835, simple_loss=0.277, pruned_loss=0.04502, over 12159.00 frames. ], tot_loss[loss=0.1775, simple_loss=0.265, pruned_loss=0.04497, over 2357821.10 frames. ], batch size: 36, lr: 4.68e-03, grad_scale: 16.0 2023-05-16 02:35:02,395 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=154740.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 02:35:11,475 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=154753.0, num_to_drop=1, layers_to_drop={0} 2023-05-16 02:35:31,691 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=154780.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 02:35:32,528 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=154781.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 02:35:35,159 INFO [finetune.py:992] (0/2) Epoch 5, batch 5850, loss[loss=0.1862, simple_loss=0.2817, pruned_loss=0.04535, over 12168.00 frames. ], tot_loss[loss=0.1762, simple_loss=0.2638, pruned_loss=0.04432, over 2369492.53 frames. ], batch size: 36, lr: 4.68e-03, grad_scale: 16.0 2023-05-16 02:35:37,412 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=154788.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 02:36:08,439 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=154830.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 02:36:10,427 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.068e+02 2.903e+02 3.495e+02 4.253e+02 2.305e+03, threshold=6.990e+02, percent-clipped=5.0 2023-05-16 02:36:11,803 INFO [finetune.py:992] (0/2) Epoch 5, batch 5900, loss[loss=0.1887, simple_loss=0.2769, pruned_loss=0.05027, over 11790.00 frames. ], tot_loss[loss=0.1765, simple_loss=0.2641, pruned_loss=0.04449, over 2362350.99 frames. ], batch size: 44, lr: 4.68e-03, grad_scale: 16.0 2023-05-16 02:36:32,051 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.6517, 2.7196, 4.5894, 5.0336, 3.4287, 2.6753, 2.9587, 2.0624], device='cuda:0'), covar=tensor([0.1411, 0.3496, 0.0442, 0.0266, 0.0871, 0.2088, 0.2529, 0.3968], device='cuda:0'), in_proj_covar=tensor([0.0272, 0.0370, 0.0263, 0.0282, 0.0252, 0.0280, 0.0350, 0.0354], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-16 02:36:41,931 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=154877.0, num_to_drop=1, layers_to_drop={0} 2023-05-16 02:36:47,282 INFO [finetune.py:992] (0/2) Epoch 5, batch 5950, loss[loss=0.2439, simple_loss=0.3132, pruned_loss=0.08733, over 7939.00 frames. ], tot_loss[loss=0.1757, simple_loss=0.2632, pruned_loss=0.04406, over 2360282.96 frames. ], batch size: 98, lr: 4.68e-03, grad_scale: 16.0 2023-05-16 02:37:16,061 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=154925.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 02:37:21,772 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.010e+02 2.820e+02 3.339e+02 4.158e+02 6.543e+02, threshold=6.678e+02, percent-clipped=0.0 2023-05-16 02:37:23,223 INFO [finetune.py:992] (0/2) Epoch 5, batch 6000, loss[loss=0.1728, simple_loss=0.264, pruned_loss=0.04078, over 12351.00 frames. ], tot_loss[loss=0.1767, simple_loss=0.2642, pruned_loss=0.04461, over 2355720.48 frames. ], batch size: 36, lr: 4.68e-03, grad_scale: 16.0 2023-05-16 02:37:23,224 INFO [finetune.py:1017] (0/2) Computing validation loss 2023-05-16 02:37:34,112 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.8919, 4.7004, 4.9056, 4.8066, 4.5507, 4.7030, 4.7390, 2.6812], device='cuda:0'), covar=tensor([0.0145, 0.0055, 0.0057, 0.0055, 0.0058, 0.0117, 0.0061, 0.0822], device='cuda:0'), in_proj_covar=tensor([0.0066, 0.0074, 0.0077, 0.0071, 0.0059, 0.0089, 0.0077, 0.0093], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 02:37:41,095 INFO [finetune.py:1026] (0/2) Epoch 5, validation: loss=0.3218, simple_loss=0.3996, pruned_loss=0.122, over 1020973.00 frames. 2023-05-16 02:37:41,096 INFO [finetune.py:1027] (0/2) Maximum memory allocated so far is 12745MB 2023-05-16 02:37:46,177 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=154942.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 02:37:49,651 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=154947.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 02:38:10,015 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=154976.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 02:38:15,629 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=154984.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 02:38:16,235 INFO [finetune.py:992] (0/2) Epoch 5, batch 6050, loss[loss=0.1736, simple_loss=0.2494, pruned_loss=0.04885, over 11997.00 frames. ], tot_loss[loss=0.1768, simple_loss=0.2641, pruned_loss=0.04471, over 2357016.41 frames. ], batch size: 28, lr: 4.68e-03, grad_scale: 16.0 2023-05-16 02:38:19,790 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=154990.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 02:38:22,007 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=154993.0, num_to_drop=1, layers_to_drop={2} 2023-05-16 02:38:23,334 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=154995.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 02:38:51,370 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.786e+02 2.886e+02 3.414e+02 4.219e+02 1.110e+03, threshold=6.829e+02, percent-clipped=1.0 2023-05-16 02:38:52,864 INFO [finetune.py:992] (0/2) Epoch 5, batch 6100, loss[loss=0.1535, simple_loss=0.2309, pruned_loss=0.03798, over 12021.00 frames. ], tot_loss[loss=0.1761, simple_loss=0.264, pruned_loss=0.04411, over 2364170.13 frames. ], batch size: 28, lr: 4.68e-03, grad_scale: 16.0 2023-05-16 02:38:54,455 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=155037.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 02:38:57,101 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=155041.0, num_to_drop=1, layers_to_drop={1} 2023-05-16 02:39:02,909 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.34 vs. limit=2.0 2023-05-16 02:39:17,578 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=155069.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 02:39:22,579 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=155076.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 02:39:29,021 INFO [finetune.py:992] (0/2) Epoch 5, batch 6150, loss[loss=0.1835, simple_loss=0.2692, pruned_loss=0.04895, over 11813.00 frames. ], tot_loss[loss=0.1767, simple_loss=0.265, pruned_loss=0.0442, over 2361261.82 frames. ], batch size: 44, lr: 4.68e-03, grad_scale: 16.0 2023-05-16 02:39:32,456 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.91 vs. limit=2.0 2023-05-16 02:39:37,262 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.36 vs. limit=2.0 2023-05-16 02:39:49,005 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.4383, 2.2983, 3.5335, 4.4964, 3.9201, 4.3526, 3.8248, 3.0986], device='cuda:0'), covar=tensor([0.0037, 0.0450, 0.0130, 0.0032, 0.0116, 0.0078, 0.0120, 0.0364], device='cuda:0'), in_proj_covar=tensor([0.0082, 0.0122, 0.0100, 0.0073, 0.0099, 0.0110, 0.0088, 0.0134], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 02:40:01,095 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=155130.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 02:40:01,176 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=155130.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 02:40:03,126 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.118e+02 2.887e+02 3.492e+02 4.080e+02 6.055e+02, threshold=6.984e+02, percent-clipped=0.0 2023-05-16 02:40:04,468 INFO [finetune.py:992] (0/2) Epoch 5, batch 6200, loss[loss=0.1966, simple_loss=0.2778, pruned_loss=0.05773, over 12093.00 frames. ], tot_loss[loss=0.177, simple_loss=0.2654, pruned_loss=0.04425, over 2370012.50 frames. ], batch size: 42, lr: 4.68e-03, grad_scale: 16.0 2023-05-16 02:40:35,837 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=155178.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 02:40:39,557 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.3943, 4.9054, 5.3172, 4.6179, 4.9816, 4.7384, 5.4023, 5.0568], device='cuda:0'), covar=tensor([0.0225, 0.0338, 0.0265, 0.0264, 0.0316, 0.0339, 0.0195, 0.0278], device='cuda:0'), in_proj_covar=tensor([0.0233, 0.0241, 0.0259, 0.0236, 0.0232, 0.0235, 0.0211, 0.0197], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 02:40:40,793 INFO [finetune.py:992] (0/2) Epoch 5, batch 6250, loss[loss=0.1883, simple_loss=0.2761, pruned_loss=0.05029, over 12286.00 frames. ], tot_loss[loss=0.1766, simple_loss=0.2649, pruned_loss=0.04418, over 2368698.70 frames. ], batch size: 33, lr: 4.68e-03, grad_scale: 16.0 2023-05-16 02:40:57,351 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.4076, 4.3710, 4.3321, 4.6265, 3.1106, 4.1864, 2.8868, 4.1837], device='cuda:0'), covar=tensor([0.1550, 0.0568, 0.0763, 0.0492, 0.1105, 0.0570, 0.1610, 0.1454], device='cuda:0'), in_proj_covar=tensor([0.0222, 0.0258, 0.0293, 0.0349, 0.0237, 0.0234, 0.0255, 0.0380], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 02:41:07,308 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.0422, 2.3935, 3.6311, 2.9650, 3.5271, 3.0542, 2.3846, 3.5002], device='cuda:0'), covar=tensor([0.0133, 0.0385, 0.0160, 0.0224, 0.0139, 0.0200, 0.0396, 0.0148], device='cuda:0'), in_proj_covar=tensor([0.0163, 0.0194, 0.0174, 0.0174, 0.0199, 0.0151, 0.0187, 0.0166], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 02:41:07,977 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.6055, 5.4253, 5.5087, 5.5875, 5.2134, 5.2394, 5.0790, 5.5416], device='cuda:0'), covar=tensor([0.0602, 0.0465, 0.0622, 0.0543, 0.1522, 0.1160, 0.0431, 0.0822], device='cuda:0'), in_proj_covar=tensor([0.0493, 0.0634, 0.0547, 0.0586, 0.0781, 0.0699, 0.0511, 0.0469], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0005, 0.0004, 0.0004, 0.0005, 0.0005, 0.0004, 0.0004], device='cuda:0') 2023-05-16 02:41:08,798 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.6599, 2.6748, 3.6313, 4.7578, 3.8943, 4.5980, 3.8986, 3.3084], device='cuda:0'), covar=tensor([0.0027, 0.0377, 0.0113, 0.0025, 0.0144, 0.0069, 0.0092, 0.0326], device='cuda:0'), in_proj_covar=tensor([0.0082, 0.0122, 0.0101, 0.0073, 0.0100, 0.0110, 0.0089, 0.0135], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 02:41:15,322 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.921e+02 2.915e+02 3.294e+02 3.991e+02 1.581e+03, threshold=6.587e+02, percent-clipped=1.0 2023-05-16 02:41:16,787 INFO [finetune.py:992] (0/2) Epoch 5, batch 6300, loss[loss=0.1482, simple_loss=0.2296, pruned_loss=0.03334, over 12248.00 frames. ], tot_loss[loss=0.1762, simple_loss=0.2648, pruned_loss=0.04378, over 2371367.87 frames. ], batch size: 28, lr: 4.68e-03, grad_scale: 16.0 2023-05-16 02:41:18,753 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.42 vs. limit=2.0 2023-05-16 02:41:51,776 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=155284.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 02:41:52,384 INFO [finetune.py:992] (0/2) Epoch 5, batch 6350, loss[loss=0.1792, simple_loss=0.2678, pruned_loss=0.04534, over 12144.00 frames. ], tot_loss[loss=0.176, simple_loss=0.2643, pruned_loss=0.04383, over 2367230.54 frames. ], batch size: 36, lr: 4.68e-03, grad_scale: 16.0 2023-05-16 02:42:13,099 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.38 vs. limit=2.0 2023-05-16 02:42:25,964 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=155332.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 02:42:26,024 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=155332.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 02:42:26,609 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.039e+02 2.968e+02 3.532e+02 4.284e+02 2.270e+03, threshold=7.063e+02, percent-clipped=5.0 2023-05-16 02:42:28,034 INFO [finetune.py:992] (0/2) Epoch 5, batch 6400, loss[loss=0.1768, simple_loss=0.2657, pruned_loss=0.04391, over 12061.00 frames. ], tot_loss[loss=0.1761, simple_loss=0.2644, pruned_loss=0.04389, over 2366362.28 frames. ], batch size: 40, lr: 4.68e-03, grad_scale: 16.0 2023-05-16 02:42:31,293 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.85 vs. limit=2.0 2023-05-16 02:42:57,639 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=155376.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 02:43:03,958 INFO [finetune.py:992] (0/2) Epoch 5, batch 6450, loss[loss=0.1886, simple_loss=0.2758, pruned_loss=0.05065, over 12155.00 frames. ], tot_loss[loss=0.1759, simple_loss=0.2642, pruned_loss=0.04383, over 2366410.14 frames. ], batch size: 34, lr: 4.67e-03, grad_scale: 16.0 2023-05-16 02:43:14,841 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.4886, 4.7468, 2.9641, 2.6018, 4.0968, 2.7748, 3.9383, 3.5102], device='cuda:0'), covar=tensor([0.0581, 0.0503, 0.1047, 0.1430, 0.0257, 0.1126, 0.0451, 0.0653], device='cuda:0'), in_proj_covar=tensor([0.0182, 0.0244, 0.0172, 0.0192, 0.0136, 0.0176, 0.0189, 0.0169], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 02:43:21,758 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.6579, 3.4631, 5.0673, 2.7113, 2.7609, 3.7883, 3.2219, 3.9371], device='cuda:0'), covar=tensor([0.0454, 0.1030, 0.0302, 0.1094, 0.1924, 0.1598, 0.1337, 0.1067], device='cuda:0'), in_proj_covar=tensor([0.0232, 0.0227, 0.0231, 0.0180, 0.0235, 0.0280, 0.0224, 0.0247], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:0') 2023-05-16 02:43:31,763 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=155424.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 02:43:32,492 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=155425.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 02:43:38,060 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.153e+02 3.057e+02 3.788e+02 4.404e+02 2.061e+03, threshold=7.575e+02, percent-clipped=3.0 2023-05-16 02:43:39,498 INFO [finetune.py:992] (0/2) Epoch 5, batch 6500, loss[loss=0.1516, simple_loss=0.2422, pruned_loss=0.0305, over 12348.00 frames. ], tot_loss[loss=0.1761, simple_loss=0.2641, pruned_loss=0.04402, over 2365793.87 frames. ], batch size: 31, lr: 4.67e-03, grad_scale: 16.0 2023-05-16 02:43:55,963 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=155457.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 02:44:15,818 INFO [finetune.py:992] (0/2) Epoch 5, batch 6550, loss[loss=0.1551, simple_loss=0.2382, pruned_loss=0.03599, over 12128.00 frames. ], tot_loss[loss=0.1766, simple_loss=0.2645, pruned_loss=0.04429, over 2371324.57 frames. ], batch size: 30, lr: 4.67e-03, grad_scale: 16.0 2023-05-16 02:44:18,906 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.2723, 4.7084, 2.8860, 2.6121, 4.0280, 2.5721, 3.9633, 3.3518], device='cuda:0'), covar=tensor([0.0696, 0.0403, 0.1137, 0.1443, 0.0246, 0.1366, 0.0437, 0.0706], device='cuda:0'), in_proj_covar=tensor([0.0183, 0.0245, 0.0172, 0.0192, 0.0137, 0.0176, 0.0189, 0.0170], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 02:44:39,906 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=155518.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 02:44:42,815 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.1892, 4.7190, 4.9641, 4.9444, 4.7686, 5.0426, 4.8444, 2.4740], device='cuda:0'), covar=tensor([0.0102, 0.0083, 0.0097, 0.0076, 0.0062, 0.0099, 0.0086, 0.0773], device='cuda:0'), in_proj_covar=tensor([0.0065, 0.0073, 0.0077, 0.0070, 0.0058, 0.0089, 0.0076, 0.0091], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 02:44:50,328 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.753e+02 2.898e+02 3.292e+02 3.853e+02 7.914e+02, threshold=6.584e+02, percent-clipped=1.0 2023-05-16 02:44:51,713 INFO [finetune.py:992] (0/2) Epoch 5, batch 6600, loss[loss=0.2134, simple_loss=0.2905, pruned_loss=0.06819, over 7997.00 frames. ], tot_loss[loss=0.1764, simple_loss=0.2647, pruned_loss=0.04407, over 2372789.18 frames. ], batch size: 97, lr: 4.67e-03, grad_scale: 16.0 2023-05-16 02:45:08,138 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.9223, 3.3855, 5.5192, 2.7537, 2.6848, 4.3998, 3.4969, 4.2890], device='cuda:0'), covar=tensor([0.0390, 0.1060, 0.0155, 0.1102, 0.1899, 0.1051, 0.1205, 0.0761], device='cuda:0'), in_proj_covar=tensor([0.0233, 0.0230, 0.0233, 0.0182, 0.0237, 0.0283, 0.0226, 0.0250], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:0') 2023-05-16 02:45:27,007 INFO [finetune.py:992] (0/2) Epoch 5, batch 6650, loss[loss=0.2077, simple_loss=0.2911, pruned_loss=0.06214, over 12155.00 frames. ], tot_loss[loss=0.1768, simple_loss=0.2648, pruned_loss=0.04435, over 2371304.46 frames. ], batch size: 34, lr: 4.67e-03, grad_scale: 16.0 2023-05-16 02:45:30,272 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=2.86 vs. limit=5.0 2023-05-16 02:45:42,441 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=155605.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 02:46:01,707 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=155632.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 02:46:02,197 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.122e+02 2.948e+02 3.456e+02 4.032e+02 7.773e+02, threshold=6.912e+02, percent-clipped=4.0 2023-05-16 02:46:03,599 INFO [finetune.py:992] (0/2) Epoch 5, batch 6700, loss[loss=0.16, simple_loss=0.2437, pruned_loss=0.03818, over 12340.00 frames. ], tot_loss[loss=0.1764, simple_loss=0.2645, pruned_loss=0.04412, over 2371825.47 frames. ], batch size: 30, lr: 4.67e-03, grad_scale: 16.0 2023-05-16 02:46:20,064 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=155657.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 02:46:21,541 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.3980, 3.6743, 3.5595, 3.2110, 3.0908, 2.9895, 3.6874, 2.2652], device='cuda:0'), covar=tensor([0.0381, 0.0107, 0.0113, 0.0169, 0.0268, 0.0260, 0.0104, 0.0448], device='cuda:0'), in_proj_covar=tensor([0.0186, 0.0158, 0.0150, 0.0181, 0.0200, 0.0194, 0.0159, 0.0200], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 02:46:26,337 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=155666.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 02:46:36,194 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=155680.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 02:46:39,416 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.97 vs. limit=2.0 2023-05-16 02:46:39,690 INFO [finetune.py:992] (0/2) Epoch 5, batch 6750, loss[loss=0.159, simple_loss=0.2462, pruned_loss=0.03589, over 12350.00 frames. ], tot_loss[loss=0.1765, simple_loss=0.2646, pruned_loss=0.04421, over 2374710.29 frames. ], batch size: 36, lr: 4.67e-03, grad_scale: 16.0 2023-05-16 02:46:44,693 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=2.31 vs. limit=5.0 2023-05-16 02:47:03,125 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=155718.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 02:47:08,195 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=155725.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 02:47:13,779 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.940e+02 2.812e+02 3.211e+02 3.815e+02 1.253e+03, threshold=6.422e+02, percent-clipped=3.0 2023-05-16 02:47:15,212 INFO [finetune.py:992] (0/2) Epoch 5, batch 6800, loss[loss=0.1602, simple_loss=0.2384, pruned_loss=0.04096, over 11818.00 frames. ], tot_loss[loss=0.1765, simple_loss=0.2647, pruned_loss=0.04419, over 2373361.39 frames. ], batch size: 26, lr: 4.67e-03, grad_scale: 16.0 2023-05-16 02:47:42,950 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=155773.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 02:47:51,030 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=155784.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 02:47:51,559 INFO [finetune.py:992] (0/2) Epoch 5, batch 6850, loss[loss=0.1793, simple_loss=0.2682, pruned_loss=0.0452, over 12091.00 frames. ], tot_loss[loss=0.1755, simple_loss=0.2633, pruned_loss=0.04384, over 2379921.44 frames. ], batch size: 32, lr: 4.67e-03, grad_scale: 16.0 2023-05-16 02:48:12,468 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=155813.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 02:48:26,677 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.847e+02 2.816e+02 3.262e+02 4.172e+02 9.329e+02, threshold=6.524e+02, percent-clipped=4.0 2023-05-16 02:48:28,087 INFO [finetune.py:992] (0/2) Epoch 5, batch 6900, loss[loss=0.1522, simple_loss=0.2332, pruned_loss=0.03555, over 12008.00 frames. ], tot_loss[loss=0.1751, simple_loss=0.2629, pruned_loss=0.0436, over 2376916.29 frames. ], batch size: 28, lr: 4.67e-03, grad_scale: 16.0 2023-05-16 02:48:35,275 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=155845.0, num_to_drop=1, layers_to_drop={2} 2023-05-16 02:49:03,617 INFO [finetune.py:992] (0/2) Epoch 5, batch 6950, loss[loss=0.1499, simple_loss=0.2343, pruned_loss=0.03273, over 12017.00 frames. ], tot_loss[loss=0.1754, simple_loss=0.2632, pruned_loss=0.04377, over 2375093.85 frames. ], batch size: 31, lr: 4.67e-03, grad_scale: 16.0 2023-05-16 02:49:08,065 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([6.2094, 6.1217, 5.8975, 5.4187, 5.2531, 6.0914, 5.6834, 5.4117], device='cuda:0'), covar=tensor([0.0588, 0.0975, 0.0582, 0.1290, 0.0623, 0.0669, 0.1386, 0.1012], device='cuda:0'), in_proj_covar=tensor([0.0572, 0.0503, 0.0482, 0.0593, 0.0389, 0.0668, 0.0730, 0.0537], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0004, 0.0004, 0.0003], device='cuda:0') 2023-05-16 02:49:37,949 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.951e+02 2.826e+02 3.292e+02 4.057e+02 7.257e+02, threshold=6.584e+02, percent-clipped=2.0 2023-05-16 02:49:39,388 INFO [finetune.py:992] (0/2) Epoch 5, batch 7000, loss[loss=0.1765, simple_loss=0.2608, pruned_loss=0.04609, over 12262.00 frames. ], tot_loss[loss=0.1756, simple_loss=0.2631, pruned_loss=0.04405, over 2382152.70 frames. ], batch size: 32, lr: 4.67e-03, grad_scale: 16.0 2023-05-16 02:49:52,585 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.0626, 4.6387, 4.8260, 4.9265, 4.6725, 4.9347, 4.7568, 2.3728], device='cuda:0'), covar=tensor([0.0106, 0.0057, 0.0080, 0.0056, 0.0052, 0.0099, 0.0067, 0.0774], device='cuda:0'), in_proj_covar=tensor([0.0065, 0.0073, 0.0077, 0.0070, 0.0058, 0.0089, 0.0076, 0.0091], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 02:49:58,616 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=155961.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 02:50:15,804 INFO [finetune.py:992] (0/2) Epoch 5, batch 7050, loss[loss=0.1683, simple_loss=0.2603, pruned_loss=0.03819, over 12300.00 frames. ], tot_loss[loss=0.1758, simple_loss=0.2635, pruned_loss=0.04402, over 2386054.79 frames. ], batch size: 34, lr: 4.67e-03, grad_scale: 32.0 2023-05-16 02:50:26,625 INFO [checkpoint.py:75] (0/2) Saving checkpoint to pruned_transducer_stateless7/exp_giga_finetune/checkpoint-56000.pt 2023-05-16 02:50:39,244 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=156013.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 02:50:44,063 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.95 vs. limit=2.0 2023-05-16 02:50:50,804 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.9837, 3.4568, 5.4053, 3.1296, 3.0991, 4.0942, 3.7185, 4.0610], device='cuda:0'), covar=tensor([0.0410, 0.1086, 0.0267, 0.1024, 0.1762, 0.1350, 0.1140, 0.1065], device='cuda:0'), in_proj_covar=tensor([0.0236, 0.0231, 0.0234, 0.0183, 0.0239, 0.0286, 0.0227, 0.0252], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-16 02:50:53,336 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.045e+02 2.953e+02 3.641e+02 4.488e+02 8.703e+02, threshold=7.282e+02, percent-clipped=6.0 2023-05-16 02:50:54,729 INFO [finetune.py:992] (0/2) Epoch 5, batch 7100, loss[loss=0.1984, simple_loss=0.2754, pruned_loss=0.06067, over 11148.00 frames. ], tot_loss[loss=0.1758, simple_loss=0.2631, pruned_loss=0.04432, over 2383498.04 frames. ], batch size: 55, lr: 4.67e-03, grad_scale: 32.0 2023-05-16 02:51:20,345 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=156070.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 02:51:23,438 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.40 vs. limit=2.0 2023-05-16 02:51:31,375 INFO [finetune.py:992] (0/2) Epoch 5, batch 7150, loss[loss=0.1781, simple_loss=0.2708, pruned_loss=0.04274, over 12375.00 frames. ], tot_loss[loss=0.1761, simple_loss=0.2637, pruned_loss=0.04426, over 2383685.53 frames. ], batch size: 38, lr: 4.67e-03, grad_scale: 32.0 2023-05-16 02:51:51,447 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=156113.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 02:52:04,312 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=156131.0, num_to_drop=1, layers_to_drop={3} 2023-05-16 02:52:05,469 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.314e+02 2.951e+02 3.337e+02 3.935e+02 8.219e+02, threshold=6.675e+02, percent-clipped=0.0 2023-05-16 02:52:06,936 INFO [finetune.py:992] (0/2) Epoch 5, batch 7200, loss[loss=0.1582, simple_loss=0.2472, pruned_loss=0.03456, over 12334.00 frames. ], tot_loss[loss=0.1753, simple_loss=0.2634, pruned_loss=0.04364, over 2382537.89 frames. ], batch size: 30, lr: 4.67e-03, grad_scale: 32.0 2023-05-16 02:52:10,499 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=156140.0, num_to_drop=1, layers_to_drop={0} 2023-05-16 02:52:21,836 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.2634, 2.6371, 3.8925, 3.3853, 3.7698, 3.3843, 2.7671, 3.7523], device='cuda:0'), covar=tensor([0.0111, 0.0300, 0.0114, 0.0170, 0.0105, 0.0142, 0.0286, 0.0098], device='cuda:0'), in_proj_covar=tensor([0.0165, 0.0194, 0.0176, 0.0174, 0.0198, 0.0152, 0.0188, 0.0168], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 02:52:25,170 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=156161.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 02:52:27,164 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.86 vs. limit=2.0 2023-05-16 02:52:42,329 INFO [finetune.py:992] (0/2) Epoch 5, batch 7250, loss[loss=0.1711, simple_loss=0.2617, pruned_loss=0.04029, over 11628.00 frames. ], tot_loss[loss=0.1761, simple_loss=0.2639, pruned_loss=0.04417, over 2369827.81 frames. ], batch size: 48, lr: 4.67e-03, grad_scale: 32.0 2023-05-16 02:53:05,305 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=156216.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 02:53:17,791 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.880e+02 2.791e+02 3.354e+02 4.246e+02 6.489e+02, threshold=6.708e+02, percent-clipped=1.0 2023-05-16 02:53:19,263 INFO [finetune.py:992] (0/2) Epoch 5, batch 7300, loss[loss=0.1719, simple_loss=0.2701, pruned_loss=0.03687, over 11705.00 frames. ], tot_loss[loss=0.1755, simple_loss=0.2636, pruned_loss=0.04374, over 2374540.39 frames. ], batch size: 48, lr: 4.67e-03, grad_scale: 32.0 2023-05-16 02:53:19,459 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.9442, 4.9223, 4.7131, 4.7946, 4.3588, 4.9264, 4.8665, 5.1904], device='cuda:0'), covar=tensor([0.0240, 0.0143, 0.0232, 0.0279, 0.0794, 0.0256, 0.0157, 0.0166], device='cuda:0'), in_proj_covar=tensor([0.0179, 0.0184, 0.0182, 0.0230, 0.0231, 0.0199, 0.0166, 0.0215], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0005, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:0') 2023-05-16 02:53:37,824 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=156261.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 02:53:49,230 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=156277.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 02:53:54,751 INFO [finetune.py:992] (0/2) Epoch 5, batch 7350, loss[loss=0.1544, simple_loss=0.2537, pruned_loss=0.02756, over 12305.00 frames. ], tot_loss[loss=0.1742, simple_loss=0.2623, pruned_loss=0.04302, over 2374514.99 frames. ], batch size: 34, lr: 4.67e-03, grad_scale: 32.0 2023-05-16 02:53:58,555 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=156290.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 02:54:11,918 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=156309.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 02:54:13,023 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.94 vs. limit=2.0 2023-05-16 02:54:14,889 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=156313.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 02:54:21,259 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.6971, 2.5645, 3.7225, 4.8058, 4.1590, 4.6595, 4.1109, 3.3417], device='cuda:0'), covar=tensor([0.0030, 0.0389, 0.0114, 0.0024, 0.0094, 0.0052, 0.0077, 0.0322], device='cuda:0'), in_proj_covar=tensor([0.0081, 0.0120, 0.0100, 0.0072, 0.0098, 0.0109, 0.0088, 0.0133], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 02:54:29,434 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.034e+02 2.962e+02 3.365e+02 3.856e+02 1.085e+03, threshold=6.730e+02, percent-clipped=2.0 2023-05-16 02:54:30,883 INFO [finetune.py:992] (0/2) Epoch 5, batch 7400, loss[loss=0.1684, simple_loss=0.2639, pruned_loss=0.03649, over 12340.00 frames. ], tot_loss[loss=0.1748, simple_loss=0.2628, pruned_loss=0.04336, over 2366167.17 frames. ], batch size: 31, lr: 4.67e-03, grad_scale: 32.0 2023-05-16 02:54:42,526 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=156351.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 02:54:49,605 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=156361.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 02:55:04,245 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.0021, 4.8234, 4.8833, 4.9445, 4.5932, 4.6233, 4.4542, 4.9011], device='cuda:0'), covar=tensor([0.0649, 0.0607, 0.0853, 0.0666, 0.1828, 0.1357, 0.0540, 0.0992], device='cuda:0'), in_proj_covar=tensor([0.0498, 0.0643, 0.0552, 0.0596, 0.0793, 0.0707, 0.0517, 0.0479], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0005, 0.0004, 0.0004, 0.0005, 0.0005, 0.0004, 0.0004], device='cuda:0') 2023-05-16 02:55:06,910 INFO [finetune.py:992] (0/2) Epoch 5, batch 7450, loss[loss=0.1929, simple_loss=0.2816, pruned_loss=0.05214, over 12129.00 frames. ], tot_loss[loss=0.1743, simple_loss=0.2623, pruned_loss=0.04318, over 2376292.27 frames. ], batch size: 38, lr: 4.66e-03, grad_scale: 32.0 2023-05-16 02:55:36,532 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=156426.0, num_to_drop=1, layers_to_drop={1} 2023-05-16 02:55:41,211 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.591e+02 2.806e+02 3.421e+02 3.972e+02 8.442e+02, threshold=6.841e+02, percent-clipped=3.0 2023-05-16 02:55:42,621 INFO [finetune.py:992] (0/2) Epoch 5, batch 7500, loss[loss=0.1638, simple_loss=0.2372, pruned_loss=0.0452, over 11830.00 frames. ], tot_loss[loss=0.1751, simple_loss=0.2628, pruned_loss=0.04375, over 2374536.12 frames. ], batch size: 26, lr: 4.66e-03, grad_scale: 32.0 2023-05-16 02:55:46,289 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=156440.0, num_to_drop=1, layers_to_drop={2} 2023-05-16 02:56:18,458 INFO [finetune.py:992] (0/2) Epoch 5, batch 7550, loss[loss=0.1677, simple_loss=0.2576, pruned_loss=0.03885, over 12151.00 frames. ], tot_loss[loss=0.1757, simple_loss=0.2632, pruned_loss=0.04413, over 2369868.99 frames. ], batch size: 36, lr: 4.66e-03, grad_scale: 32.0 2023-05-16 02:56:20,476 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=156488.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 02:56:52,759 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.785e+02 2.893e+02 3.396e+02 4.183e+02 9.129e+02, threshold=6.793e+02, percent-clipped=2.0 2023-05-16 02:56:54,232 INFO [finetune.py:992] (0/2) Epoch 5, batch 7600, loss[loss=0.1455, simple_loss=0.2275, pruned_loss=0.03177, over 12138.00 frames. ], tot_loss[loss=0.1755, simple_loss=0.2631, pruned_loss=0.04396, over 2376916.94 frames. ], batch size: 30, lr: 4.66e-03, grad_scale: 32.0 2023-05-16 02:57:20,283 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=156572.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 02:57:29,555 INFO [finetune.py:992] (0/2) Epoch 5, batch 7650, loss[loss=0.2167, simple_loss=0.2979, pruned_loss=0.06781, over 10824.00 frames. ], tot_loss[loss=0.1764, simple_loss=0.2639, pruned_loss=0.04446, over 2369464.42 frames. ], batch size: 69, lr: 4.66e-03, grad_scale: 32.0 2023-05-16 02:58:04,219 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.926e+02 2.831e+02 3.260e+02 4.176e+02 6.963e+02, threshold=6.520e+02, percent-clipped=1.0 2023-05-16 02:58:05,706 INFO [finetune.py:992] (0/2) Epoch 5, batch 7700, loss[loss=0.1784, simple_loss=0.2672, pruned_loss=0.04482, over 12037.00 frames. ], tot_loss[loss=0.1776, simple_loss=0.2651, pruned_loss=0.04504, over 2367022.95 frames. ], batch size: 42, lr: 4.66e-03, grad_scale: 32.0 2023-05-16 02:58:13,605 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=156646.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 02:58:41,771 INFO [finetune.py:992] (0/2) Epoch 5, batch 7750, loss[loss=0.2004, simple_loss=0.2894, pruned_loss=0.05572, over 11805.00 frames. ], tot_loss[loss=0.179, simple_loss=0.2663, pruned_loss=0.04588, over 2352314.14 frames. ], batch size: 44, lr: 4.66e-03, grad_scale: 32.0 2023-05-16 02:59:10,756 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=156726.0, num_to_drop=1, layers_to_drop={2} 2023-05-16 02:59:12,961 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.0155, 4.8092, 4.7934, 5.0184, 4.7150, 4.9407, 4.8164, 2.7258], device='cuda:0'), covar=tensor([0.0134, 0.0055, 0.0088, 0.0055, 0.0063, 0.0105, 0.0067, 0.0688], device='cuda:0'), in_proj_covar=tensor([0.0065, 0.0072, 0.0077, 0.0070, 0.0058, 0.0088, 0.0075, 0.0090], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 02:59:15,574 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.120e+02 2.973e+02 3.437e+02 3.966e+02 7.070e+02, threshold=6.873e+02, percent-clipped=1.0 2023-05-16 02:59:17,012 INFO [finetune.py:992] (0/2) Epoch 5, batch 7800, loss[loss=0.1624, simple_loss=0.2476, pruned_loss=0.03862, over 12361.00 frames. ], tot_loss[loss=0.1783, simple_loss=0.2655, pruned_loss=0.04551, over 2363650.59 frames. ], batch size: 35, lr: 4.66e-03, grad_scale: 32.0 2023-05-16 02:59:27,179 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=156749.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 02:59:30,257 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.7453, 2.8810, 4.5244, 4.7179, 2.9236, 2.6161, 3.0291, 2.1215], device='cuda:0'), covar=tensor([0.1362, 0.3056, 0.0431, 0.0345, 0.1211, 0.2075, 0.2351, 0.3746], device='cuda:0'), in_proj_covar=tensor([0.0269, 0.0366, 0.0260, 0.0281, 0.0251, 0.0279, 0.0347, 0.0350], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-16 02:59:45,719 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=156774.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 02:59:53,419 INFO [finetune.py:992] (0/2) Epoch 5, batch 7850, loss[loss=0.1505, simple_loss=0.2381, pruned_loss=0.03146, over 12253.00 frames. ], tot_loss[loss=0.1779, simple_loss=0.2653, pruned_loss=0.04524, over 2360827.90 frames. ], batch size: 32, lr: 4.66e-03, grad_scale: 32.0 2023-05-16 03:00:00,342 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2023-05-16 03:00:00,974 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=2.72 vs. limit=5.0 2023-05-16 03:00:12,452 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=156810.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 03:00:28,452 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.049e+02 2.736e+02 3.172e+02 3.920e+02 5.453e+02, threshold=6.343e+02, percent-clipped=0.0 2023-05-16 03:00:29,800 INFO [finetune.py:992] (0/2) Epoch 5, batch 7900, loss[loss=0.1766, simple_loss=0.2566, pruned_loss=0.04831, over 12102.00 frames. ], tot_loss[loss=0.1771, simple_loss=0.2648, pruned_loss=0.04475, over 2372343.35 frames. ], batch size: 33, lr: 4.66e-03, grad_scale: 32.0 2023-05-16 03:00:31,368 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.8481, 4.5817, 4.6623, 4.8489, 4.5117, 4.8394, 4.6456, 2.7930], device='cuda:0'), covar=tensor([0.0106, 0.0065, 0.0090, 0.0057, 0.0069, 0.0084, 0.0079, 0.0637], device='cuda:0'), in_proj_covar=tensor([0.0063, 0.0071, 0.0075, 0.0068, 0.0057, 0.0086, 0.0074, 0.0088], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 03:00:55,981 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.2004, 5.1718, 4.9310, 5.0930, 4.6315, 5.1434, 5.1580, 5.4261], device='cuda:0'), covar=tensor([0.0151, 0.0126, 0.0202, 0.0305, 0.0700, 0.0259, 0.0133, 0.0125], device='cuda:0'), in_proj_covar=tensor([0.0179, 0.0185, 0.0181, 0.0230, 0.0232, 0.0200, 0.0166, 0.0215], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0005, 0.0005, 0.0005, 0.0004, 0.0005], device='cuda:0') 2023-05-16 03:00:55,992 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=156872.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 03:01:05,103 INFO [finetune.py:992] (0/2) Epoch 5, batch 7950, loss[loss=0.1818, simple_loss=0.2684, pruned_loss=0.04763, over 12369.00 frames. ], tot_loss[loss=0.1762, simple_loss=0.2641, pruned_loss=0.04418, over 2376715.18 frames. ], batch size: 38, lr: 4.66e-03, grad_scale: 32.0 2023-05-16 03:01:05,282 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=156885.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 03:01:31,029 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=156920.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 03:01:36,807 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.4798, 5.0755, 5.4220, 4.7224, 5.1479, 4.8099, 5.4252, 5.1243], device='cuda:0'), covar=tensor([0.0271, 0.0350, 0.0351, 0.0256, 0.0284, 0.0319, 0.0274, 0.0235], device='cuda:0'), in_proj_covar=tensor([0.0237, 0.0247, 0.0266, 0.0239, 0.0236, 0.0238, 0.0215, 0.0204], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 03:01:39,195 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 2023-05-16 03:01:40,152 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.241e+02 2.948e+02 3.353e+02 3.956e+02 7.512e+02, threshold=6.705e+02, percent-clipped=2.0 2023-05-16 03:01:41,643 INFO [finetune.py:992] (0/2) Epoch 5, batch 8000, loss[loss=0.1845, simple_loss=0.275, pruned_loss=0.04693, over 12048.00 frames. ], tot_loss[loss=0.1755, simple_loss=0.2626, pruned_loss=0.04416, over 2376853.50 frames. ], batch size: 37, lr: 4.66e-03, grad_scale: 32.0 2023-05-16 03:01:44,529 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.0960, 2.4881, 3.6286, 3.0677, 3.5962, 3.0998, 2.5719, 3.5405], device='cuda:0'), covar=tensor([0.0145, 0.0351, 0.0185, 0.0262, 0.0162, 0.0210, 0.0328, 0.0136], device='cuda:0'), in_proj_covar=tensor([0.0166, 0.0197, 0.0177, 0.0177, 0.0201, 0.0154, 0.0189, 0.0171], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 03:01:49,400 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=156946.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 03:01:49,460 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=156946.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 03:02:17,227 INFO [finetune.py:992] (0/2) Epoch 5, batch 8050, loss[loss=0.2017, simple_loss=0.2809, pruned_loss=0.0612, over 12138.00 frames. ], tot_loss[loss=0.1763, simple_loss=0.2636, pruned_loss=0.04446, over 2378529.26 frames. ], batch size: 39, lr: 4.66e-03, grad_scale: 32.0 2023-05-16 03:02:23,660 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=156994.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 03:02:35,420 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=157010.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 03:02:51,177 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.866e+02 2.775e+02 3.369e+02 4.080e+02 5.929e+02, threshold=6.738e+02, percent-clipped=0.0 2023-05-16 03:02:52,696 INFO [finetune.py:992] (0/2) Epoch 5, batch 8100, loss[loss=0.1559, simple_loss=0.2346, pruned_loss=0.03858, over 12280.00 frames. ], tot_loss[loss=0.177, simple_loss=0.2636, pruned_loss=0.04521, over 2365762.65 frames. ], batch size: 28, lr: 4.66e-03, grad_scale: 32.0 2023-05-16 03:03:13,823 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.85 vs. limit=2.0 2023-05-16 03:03:19,227 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=157071.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 03:03:29,023 INFO [finetune.py:992] (0/2) Epoch 5, batch 8150, loss[loss=0.1467, simple_loss=0.2305, pruned_loss=0.03146, over 12361.00 frames. ], tot_loss[loss=0.1772, simple_loss=0.2637, pruned_loss=0.04535, over 2366763.95 frames. ], batch size: 30, lr: 4.66e-03, grad_scale: 32.0 2023-05-16 03:03:43,836 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=157105.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 03:03:58,043 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.2718, 4.6298, 2.9500, 2.6103, 4.0189, 2.5391, 3.9333, 3.0244], device='cuda:0'), covar=tensor([0.0726, 0.0552, 0.1114, 0.1514, 0.0256, 0.1421, 0.0504, 0.0891], device='cuda:0'), in_proj_covar=tensor([0.0188, 0.0251, 0.0177, 0.0198, 0.0139, 0.0181, 0.0196, 0.0173], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 03:04:03,571 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.287e+02 2.872e+02 3.313e+02 3.901e+02 6.271e+02, threshold=6.625e+02, percent-clipped=0.0 2023-05-16 03:04:05,057 INFO [finetune.py:992] (0/2) Epoch 5, batch 8200, loss[loss=0.1644, simple_loss=0.2556, pruned_loss=0.03657, over 12020.00 frames. ], tot_loss[loss=0.1762, simple_loss=0.2632, pruned_loss=0.04458, over 2369579.60 frames. ], batch size: 31, lr: 4.66e-03, grad_scale: 32.0 2023-05-16 03:04:10,158 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=157142.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 03:04:29,305 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.4125, 4.9096, 5.3876, 4.7446, 5.0408, 4.8796, 5.4477, 5.0787], device='cuda:0'), covar=tensor([0.0220, 0.0323, 0.0264, 0.0225, 0.0282, 0.0242, 0.0196, 0.0264], device='cuda:0'), in_proj_covar=tensor([0.0238, 0.0249, 0.0269, 0.0241, 0.0237, 0.0239, 0.0217, 0.0204], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 03:04:30,039 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.4303, 3.3382, 3.1709, 2.9726, 2.7153, 2.6671, 3.2788, 2.2142], device='cuda:0'), covar=tensor([0.0310, 0.0161, 0.0194, 0.0178, 0.0357, 0.0289, 0.0155, 0.0416], device='cuda:0'), in_proj_covar=tensor([0.0191, 0.0160, 0.0153, 0.0184, 0.0205, 0.0195, 0.0162, 0.0204], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 03:04:40,631 INFO [finetune.py:992] (0/2) Epoch 5, batch 8250, loss[loss=0.1885, simple_loss=0.2743, pruned_loss=0.05136, over 12165.00 frames. ], tot_loss[loss=0.1768, simple_loss=0.2639, pruned_loss=0.04487, over 2363849.43 frames. ], batch size: 31, lr: 4.66e-03, grad_scale: 32.0 2023-05-16 03:04:41,523 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.9170, 4.5163, 4.5801, 4.7934, 4.6297, 4.8827, 4.6424, 2.5026], device='cuda:0'), covar=tensor([0.0098, 0.0075, 0.0110, 0.0070, 0.0053, 0.0089, 0.0127, 0.0763], device='cuda:0'), in_proj_covar=tensor([0.0063, 0.0070, 0.0074, 0.0068, 0.0056, 0.0085, 0.0074, 0.0088], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-16 03:04:54,456 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=157203.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 03:05:15,221 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.879e+02 2.813e+02 3.377e+02 4.175e+02 1.079e+03, threshold=6.754e+02, percent-clipped=4.0 2023-05-16 03:05:16,754 INFO [finetune.py:992] (0/2) Epoch 5, batch 8300, loss[loss=0.1675, simple_loss=0.2552, pruned_loss=0.03993, over 12032.00 frames. ], tot_loss[loss=0.1761, simple_loss=0.2634, pruned_loss=0.04442, over 2366355.13 frames. ], batch size: 31, lr: 4.66e-03, grad_scale: 32.0 2023-05-16 03:05:21,670 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=157241.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 03:05:25,365 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.0787, 4.2637, 4.1907, 4.4085, 3.0106, 3.9586, 2.4479, 4.0931], device='cuda:0'), covar=tensor([0.1605, 0.0603, 0.0803, 0.0580, 0.1037, 0.0546, 0.1887, 0.1193], device='cuda:0'), in_proj_covar=tensor([0.0221, 0.0257, 0.0291, 0.0345, 0.0233, 0.0233, 0.0252, 0.0377], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 03:05:52,684 INFO [finetune.py:992] (0/2) Epoch 5, batch 8350, loss[loss=0.1897, simple_loss=0.2768, pruned_loss=0.05128, over 12166.00 frames. ], tot_loss[loss=0.1755, simple_loss=0.2633, pruned_loss=0.04388, over 2376521.50 frames. ], batch size: 39, lr: 4.66e-03, grad_scale: 16.0 2023-05-16 03:06:27,096 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.169e+02 2.785e+02 3.372e+02 4.179e+02 8.157e+02, threshold=6.744e+02, percent-clipped=2.0 2023-05-16 03:06:27,754 INFO [finetune.py:992] (0/2) Epoch 5, batch 8400, loss[loss=0.1801, simple_loss=0.2754, pruned_loss=0.0424, over 12140.00 frames. ], tot_loss[loss=0.1767, simple_loss=0.2644, pruned_loss=0.0445, over 2371795.71 frames. ], batch size: 39, lr: 4.66e-03, grad_scale: 16.0 2023-05-16 03:06:43,533 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=157356.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 03:06:50,573 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=157366.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 03:07:01,841 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=157382.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 03:07:03,858 INFO [finetune.py:992] (0/2) Epoch 5, batch 8450, loss[loss=0.1904, simple_loss=0.2797, pruned_loss=0.05056, over 12130.00 frames. ], tot_loss[loss=0.1771, simple_loss=0.2651, pruned_loss=0.04452, over 2377056.29 frames. ], batch size: 39, lr: 4.65e-03, grad_scale: 16.0 2023-05-16 03:07:18,262 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=157404.0, num_to_drop=1, layers_to_drop={1} 2023-05-16 03:07:18,957 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=157405.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 03:07:27,426 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=157417.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 03:07:38,201 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.3645, 4.9045, 3.1849, 2.6806, 4.2189, 3.1068, 4.1982, 3.5914], device='cuda:0'), covar=tensor([0.0680, 0.0435, 0.0948, 0.1457, 0.0217, 0.0992, 0.0386, 0.0664], device='cuda:0'), in_proj_covar=tensor([0.0186, 0.0249, 0.0175, 0.0196, 0.0137, 0.0179, 0.0194, 0.0171], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 03:07:39,350 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.967e+02 2.701e+02 3.200e+02 3.996e+02 1.013e+03, threshold=6.400e+02, percent-clipped=3.0 2023-05-16 03:07:40,011 INFO [finetune.py:992] (0/2) Epoch 5, batch 8500, loss[loss=0.1729, simple_loss=0.2527, pruned_loss=0.04659, over 12185.00 frames. ], tot_loss[loss=0.1773, simple_loss=0.265, pruned_loss=0.04481, over 2370539.84 frames. ], batch size: 31, lr: 4.65e-03, grad_scale: 16.0 2023-05-16 03:07:45,785 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=157443.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 03:07:52,617 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=157453.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 03:07:59,750 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.2569, 5.2769, 5.1294, 5.1878, 4.6837, 5.2834, 5.2479, 5.5147], device='cuda:0'), covar=tensor([0.0174, 0.0117, 0.0160, 0.0232, 0.0718, 0.0341, 0.0164, 0.0162], device='cuda:0'), in_proj_covar=tensor([0.0181, 0.0186, 0.0184, 0.0233, 0.0234, 0.0203, 0.0168, 0.0218], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0005, 0.0005, 0.0005, 0.0004, 0.0005], device='cuda:0') 2023-05-16 03:08:01,247 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=157465.0, num_to_drop=1, layers_to_drop={0} 2023-05-16 03:08:15,985 INFO [finetune.py:992] (0/2) Epoch 5, batch 8550, loss[loss=0.1528, simple_loss=0.2385, pruned_loss=0.03359, over 12118.00 frames. ], tot_loss[loss=0.1767, simple_loss=0.2644, pruned_loss=0.04452, over 2375060.57 frames. ], batch size: 30, lr: 4.65e-03, grad_scale: 16.0 2023-05-16 03:08:25,351 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=157498.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 03:08:28,449 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.0937, 4.5788, 4.1130, 4.9597, 4.5254, 2.7866, 4.1505, 3.1531], device='cuda:0'), covar=tensor([0.0947, 0.0750, 0.1383, 0.0414, 0.1061, 0.1736, 0.1030, 0.3036], device='cuda:0'), in_proj_covar=tensor([0.0296, 0.0367, 0.0344, 0.0267, 0.0352, 0.0261, 0.0330, 0.0357], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 03:08:31,233 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=157506.0, num_to_drop=1, layers_to_drop={1} 2023-05-16 03:08:51,042 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.096e+02 2.972e+02 3.489e+02 4.078e+02 7.596e+02, threshold=6.977e+02, percent-clipped=3.0 2023-05-16 03:08:51,787 INFO [finetune.py:992] (0/2) Epoch 5, batch 8600, loss[loss=0.1648, simple_loss=0.2569, pruned_loss=0.03633, over 12306.00 frames. ], tot_loss[loss=0.1764, simple_loss=0.2641, pruned_loss=0.04435, over 2372940.08 frames. ], batch size: 34, lr: 4.65e-03, grad_scale: 16.0 2023-05-16 03:08:56,988 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=157541.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 03:09:15,476 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=157567.0, num_to_drop=1, layers_to_drop={2} 2023-05-16 03:09:16,155 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=157568.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 03:09:25,637 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.92 vs. limit=2.0 2023-05-16 03:09:28,006 INFO [finetune.py:992] (0/2) Epoch 5, batch 8650, loss[loss=0.158, simple_loss=0.248, pruned_loss=0.03403, over 12088.00 frames. ], tot_loss[loss=0.1762, simple_loss=0.2639, pruned_loss=0.04425, over 2368407.03 frames. ], batch size: 32, lr: 4.65e-03, grad_scale: 16.0 2023-05-16 03:09:30,937 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=157589.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 03:09:59,783 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=157629.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 03:10:02,964 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.068e+02 3.050e+02 3.438e+02 4.172e+02 7.635e+02, threshold=6.877e+02, percent-clipped=1.0 2023-05-16 03:10:03,664 INFO [finetune.py:992] (0/2) Epoch 5, batch 8700, loss[loss=0.1803, simple_loss=0.2724, pruned_loss=0.04406, over 12146.00 frames. ], tot_loss[loss=0.1764, simple_loss=0.2637, pruned_loss=0.04452, over 2365484.87 frames. ], batch size: 36, lr: 4.65e-03, grad_scale: 16.0 2023-05-16 03:10:25,504 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=157666.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 03:10:39,264 INFO [finetune.py:992] (0/2) Epoch 5, batch 8750, loss[loss=0.1435, simple_loss=0.2279, pruned_loss=0.0296, over 12021.00 frames. ], tot_loss[loss=0.1762, simple_loss=0.2632, pruned_loss=0.04459, over 2370785.14 frames. ], batch size: 28, lr: 4.65e-03, grad_scale: 16.0 2023-05-16 03:10:56,570 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.3124, 4.8840, 5.2319, 4.5808, 4.8919, 4.6844, 5.3052, 4.9510], device='cuda:0'), covar=tensor([0.0259, 0.0339, 0.0309, 0.0277, 0.0321, 0.0329, 0.0257, 0.0319], device='cuda:0'), in_proj_covar=tensor([0.0237, 0.0248, 0.0268, 0.0239, 0.0236, 0.0239, 0.0215, 0.0203], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 03:10:58,717 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=157712.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 03:11:00,077 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=157714.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 03:11:14,027 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.074e+02 2.815e+02 3.326e+02 4.142e+02 6.744e+02, threshold=6.652e+02, percent-clipped=0.0 2023-05-16 03:11:14,762 INFO [finetune.py:992] (0/2) Epoch 5, batch 8800, loss[loss=0.151, simple_loss=0.2442, pruned_loss=0.02889, over 12103.00 frames. ], tot_loss[loss=0.1762, simple_loss=0.2636, pruned_loss=0.04443, over 2376206.75 frames. ], batch size: 33, lr: 4.65e-03, grad_scale: 16.0 2023-05-16 03:11:17,030 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=157738.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 03:11:32,418 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=157760.0, num_to_drop=1, layers_to_drop={1} 2023-05-16 03:11:40,359 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.2801, 5.0084, 5.1250, 5.2074, 4.7976, 4.8997, 4.7161, 5.1517], device='cuda:0'), covar=tensor([0.0515, 0.0599, 0.0745, 0.0538, 0.1745, 0.1113, 0.0485, 0.0818], device='cuda:0'), in_proj_covar=tensor([0.0490, 0.0642, 0.0549, 0.0586, 0.0785, 0.0699, 0.0514, 0.0473], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0005, 0.0004, 0.0004, 0.0005, 0.0005, 0.0004, 0.0004], device='cuda:0') 2023-05-16 03:11:45,383 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([6.3209, 6.2093, 6.1192, 5.6736, 5.3468, 6.2246, 5.8609, 5.5775], device='cuda:0'), covar=tensor([0.0616, 0.0933, 0.0558, 0.1245, 0.0566, 0.0648, 0.1433, 0.0907], device='cuda:0'), in_proj_covar=tensor([0.0571, 0.0510, 0.0487, 0.0598, 0.0389, 0.0672, 0.0733, 0.0536], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0004, 0.0004, 0.0003], device='cuda:0') 2023-05-16 03:11:50,852 INFO [finetune.py:992] (0/2) Epoch 5, batch 8850, loss[loss=0.2102, simple_loss=0.2962, pruned_loss=0.06208, over 11787.00 frames. ], tot_loss[loss=0.1779, simple_loss=0.265, pruned_loss=0.04543, over 2371787.70 frames. ], batch size: 44, lr: 4.65e-03, grad_scale: 16.0 2023-05-16 03:12:00,227 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=157798.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 03:12:18,795 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.45 vs. limit=2.0 2023-05-16 03:12:26,201 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.157e+02 3.014e+02 3.460e+02 4.025e+02 6.937e+02, threshold=6.919e+02, percent-clipped=1.0 2023-05-16 03:12:26,940 INFO [finetune.py:992] (0/2) Epoch 5, batch 8900, loss[loss=0.1695, simple_loss=0.2642, pruned_loss=0.03738, over 12158.00 frames. ], tot_loss[loss=0.1781, simple_loss=0.2654, pruned_loss=0.04538, over 2370080.17 frames. ], batch size: 34, lr: 4.65e-03, grad_scale: 16.0 2023-05-16 03:12:29,513 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.80 vs. limit=2.0 2023-05-16 03:12:34,724 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=157846.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 03:12:39,993 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=157853.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 03:12:46,256 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=157862.0, num_to_drop=1, layers_to_drop={1} 2023-05-16 03:13:02,489 INFO [finetune.py:992] (0/2) Epoch 5, batch 8950, loss[loss=0.1648, simple_loss=0.2582, pruned_loss=0.03571, over 12063.00 frames. ], tot_loss[loss=0.177, simple_loss=0.2645, pruned_loss=0.04479, over 2372872.04 frames. ], batch size: 40, lr: 4.65e-03, grad_scale: 16.0 2023-05-16 03:13:24,063 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=157914.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 03:13:31,093 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=157924.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 03:13:37,963 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.872e+02 2.977e+02 3.480e+02 4.159e+02 6.142e+02, threshold=6.960e+02, percent-clipped=0.0 2023-05-16 03:13:38,635 INFO [finetune.py:992] (0/2) Epoch 5, batch 9000, loss[loss=0.1821, simple_loss=0.2764, pruned_loss=0.04396, over 12188.00 frames. ], tot_loss[loss=0.1758, simple_loss=0.2637, pruned_loss=0.04389, over 2377567.20 frames. ], batch size: 35, lr: 4.65e-03, grad_scale: 16.0 2023-05-16 03:13:38,636 INFO [finetune.py:1017] (0/2) Computing validation loss 2023-05-16 03:13:57,466 INFO [finetune.py:1026] (0/2) Epoch 5, validation: loss=0.3387, simple_loss=0.4077, pruned_loss=0.1348, over 1020973.00 frames. 2023-05-16 03:13:57,467 INFO [finetune.py:1027] (0/2) Maximum memory allocated so far is 12745MB 2023-05-16 03:13:59,004 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.5972, 2.3687, 3.8640, 4.5989, 4.1052, 4.3709, 4.0250, 3.2793], device='cuda:0'), covar=tensor([0.0036, 0.0502, 0.0104, 0.0040, 0.0085, 0.0086, 0.0082, 0.0327], device='cuda:0'), in_proj_covar=tensor([0.0081, 0.0119, 0.0100, 0.0073, 0.0097, 0.0110, 0.0088, 0.0133], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 03:14:32,905 INFO [finetune.py:992] (0/2) Epoch 5, batch 9050, loss[loss=0.1843, simple_loss=0.2719, pruned_loss=0.04831, over 12344.00 frames. ], tot_loss[loss=0.1756, simple_loss=0.2635, pruned_loss=0.04383, over 2381952.22 frames. ], batch size: 36, lr: 4.65e-03, grad_scale: 16.0 2023-05-16 03:14:40,790 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.7017, 2.6861, 3.7634, 4.6012, 4.0769, 4.6260, 3.8130, 3.2077], device='cuda:0'), covar=tensor([0.0026, 0.0349, 0.0103, 0.0039, 0.0078, 0.0054, 0.0101, 0.0317], device='cuda:0'), in_proj_covar=tensor([0.0081, 0.0120, 0.0100, 0.0073, 0.0097, 0.0110, 0.0089, 0.0133], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 03:14:44,486 INFO [checkpoint.py:75] (0/2) Saving checkpoint to pruned_transducer_stateless7/exp_giga_finetune/checkpoint-58000.pt 2023-05-16 03:14:49,440 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.4729, 4.2614, 4.1859, 4.5375, 3.2442, 4.0911, 2.8232, 4.2745], device='cuda:0'), covar=tensor([0.1336, 0.0603, 0.0975, 0.0572, 0.0982, 0.0501, 0.1662, 0.1430], device='cuda:0'), in_proj_covar=tensor([0.0218, 0.0255, 0.0290, 0.0343, 0.0230, 0.0231, 0.0251, 0.0373], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 03:14:55,829 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=158012.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 03:14:58,153 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.2798, 4.5545, 4.2017, 4.9968, 4.6582, 2.8425, 4.3011, 3.0398], device='cuda:0'), covar=tensor([0.0769, 0.0789, 0.1201, 0.0350, 0.0867, 0.1538, 0.0942, 0.3195], device='cuda:0'), in_proj_covar=tensor([0.0297, 0.0372, 0.0346, 0.0268, 0.0355, 0.0262, 0.0333, 0.0358], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 03:15:12,123 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.758e+02 2.887e+02 3.613e+02 4.458e+02 1.304e+03, threshold=7.226e+02, percent-clipped=2.0 2023-05-16 03:15:12,142 INFO [finetune.py:992] (0/2) Epoch 5, batch 9100, loss[loss=0.1924, simple_loss=0.2772, pruned_loss=0.05379, over 12147.00 frames. ], tot_loss[loss=0.1756, simple_loss=0.2635, pruned_loss=0.04381, over 2384804.74 frames. ], batch size: 34, lr: 4.65e-03, grad_scale: 8.0 2023-05-16 03:15:14,547 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=158038.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 03:15:16,713 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.9169, 3.3398, 5.2884, 2.8651, 2.8659, 3.8946, 3.2767, 3.8572], device='cuda:0'), covar=tensor([0.0369, 0.1109, 0.0241, 0.1036, 0.1925, 0.1318, 0.1295, 0.1081], device='cuda:0'), in_proj_covar=tensor([0.0230, 0.0224, 0.0231, 0.0176, 0.0233, 0.0280, 0.0221, 0.0246], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:0') 2023-05-16 03:15:30,074 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=158060.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 03:15:30,186 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=158060.0, num_to_drop=1, layers_to_drop={1} 2023-05-16 03:15:39,182 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.1100, 4.1130, 4.1478, 4.3764, 3.0254, 4.0464, 2.5233, 4.1157], device='cuda:0'), covar=tensor([0.1662, 0.0653, 0.0794, 0.0630, 0.1072, 0.0520, 0.1794, 0.1214], device='cuda:0'), in_proj_covar=tensor([0.0218, 0.0253, 0.0287, 0.0342, 0.0229, 0.0230, 0.0249, 0.0371], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 03:15:47,992 INFO [finetune.py:992] (0/2) Epoch 5, batch 9150, loss[loss=0.1609, simple_loss=0.2424, pruned_loss=0.03965, over 12295.00 frames. ], tot_loss[loss=0.1754, simple_loss=0.2633, pruned_loss=0.04374, over 2384980.15 frames. ], batch size: 28, lr: 4.65e-03, grad_scale: 8.0 2023-05-16 03:15:48,786 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=158086.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 03:16:04,408 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=158108.0, num_to_drop=1, layers_to_drop={1} 2023-05-16 03:16:08,710 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.0291, 4.6486, 4.8429, 4.9206, 4.8035, 4.8983, 4.8175, 2.7335], device='cuda:0'), covar=tensor([0.0094, 0.0072, 0.0072, 0.0066, 0.0048, 0.0092, 0.0085, 0.0713], device='cuda:0'), in_proj_covar=tensor([0.0064, 0.0072, 0.0076, 0.0069, 0.0057, 0.0087, 0.0075, 0.0090], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 03:16:14,129 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=158122.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 03:16:23,215 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.857e+02 3.039e+02 3.389e+02 4.190e+02 1.167e+03, threshold=6.777e+02, percent-clipped=1.0 2023-05-16 03:16:23,235 INFO [finetune.py:992] (0/2) Epoch 5, batch 9200, loss[loss=0.178, simple_loss=0.2662, pruned_loss=0.04492, over 12045.00 frames. ], tot_loss[loss=0.1759, simple_loss=0.2637, pruned_loss=0.04404, over 2384385.49 frames. ], batch size: 40, lr: 4.65e-03, grad_scale: 8.0 2023-05-16 03:16:43,435 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=158162.0, num_to_drop=1, layers_to_drop={2} 2023-05-16 03:16:47,739 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=2.62 vs. limit=5.0 2023-05-16 03:16:58,182 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=158183.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 03:16:59,348 INFO [finetune.py:992] (0/2) Epoch 5, batch 9250, loss[loss=0.1929, simple_loss=0.2768, pruned_loss=0.05451, over 12380.00 frames. ], tot_loss[loss=0.1756, simple_loss=0.2632, pruned_loss=0.04396, over 2387619.82 frames. ], batch size: 38, lr: 4.65e-03, grad_scale: 8.0 2023-05-16 03:17:16,821 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=158209.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 03:17:17,577 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=158210.0, num_to_drop=1, layers_to_drop={1} 2023-05-16 03:17:27,932 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=158224.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 03:17:34,435 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.7167, 2.9940, 4.7722, 5.0005, 2.8827, 2.8115, 2.9887, 2.2702], device='cuda:0'), covar=tensor([0.1386, 0.2765, 0.0444, 0.0337, 0.1161, 0.1928, 0.2573, 0.3662], device='cuda:0'), in_proj_covar=tensor([0.0271, 0.0366, 0.0262, 0.0284, 0.0251, 0.0280, 0.0348, 0.0351], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-16 03:17:35,584 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.926e+02 2.936e+02 3.408e+02 4.274e+02 9.884e+02, threshold=6.817e+02, percent-clipped=6.0 2023-05-16 03:17:35,607 INFO [finetune.py:992] (0/2) Epoch 5, batch 9300, loss[loss=0.1721, simple_loss=0.2524, pruned_loss=0.04588, over 12289.00 frames. ], tot_loss[loss=0.1762, simple_loss=0.264, pruned_loss=0.04417, over 2378914.02 frames. ], batch size: 33, lr: 4.65e-03, grad_scale: 8.0 2023-05-16 03:17:46,432 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.7391, 2.8143, 3.6302, 4.6832, 4.1487, 4.6659, 3.9427, 3.1257], device='cuda:0'), covar=tensor([0.0026, 0.0327, 0.0126, 0.0028, 0.0095, 0.0052, 0.0084, 0.0310], device='cuda:0'), in_proj_covar=tensor([0.0082, 0.0121, 0.0102, 0.0074, 0.0098, 0.0111, 0.0089, 0.0135], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 03:18:02,086 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=158272.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 03:18:05,540 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.2879, 4.6939, 4.2222, 5.0447, 4.5481, 2.4490, 4.1635, 3.0350], device='cuda:0'), covar=tensor([0.0780, 0.0687, 0.1161, 0.0429, 0.0997, 0.1797, 0.1113, 0.2934], device='cuda:0'), in_proj_covar=tensor([0.0296, 0.0370, 0.0345, 0.0268, 0.0352, 0.0262, 0.0331, 0.0357], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 03:18:11,516 INFO [finetune.py:992] (0/2) Epoch 5, batch 9350, loss[loss=0.1787, simple_loss=0.2753, pruned_loss=0.04108, over 12138.00 frames. ], tot_loss[loss=0.1752, simple_loss=0.263, pruned_loss=0.04372, over 2387460.39 frames. ], batch size: 34, lr: 4.65e-03, grad_scale: 8.0 2023-05-16 03:18:12,433 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.3271, 4.8664, 5.2839, 4.6059, 4.9305, 4.7012, 5.3623, 4.9823], device='cuda:0'), covar=tensor([0.0248, 0.0371, 0.0267, 0.0235, 0.0308, 0.0311, 0.0186, 0.0258], device='cuda:0'), in_proj_covar=tensor([0.0236, 0.0249, 0.0267, 0.0239, 0.0235, 0.0239, 0.0216, 0.0203], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 03:18:20,353 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.33 vs. limit=2.0 2023-05-16 03:18:30,688 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.3407, 4.3044, 4.1768, 4.2596, 3.8906, 4.3522, 4.3334, 4.5072], device='cuda:0'), covar=tensor([0.0281, 0.0168, 0.0213, 0.0347, 0.0782, 0.0348, 0.0186, 0.0229], device='cuda:0'), in_proj_covar=tensor([0.0182, 0.0186, 0.0184, 0.0234, 0.0234, 0.0203, 0.0168, 0.0218], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0005, 0.0005, 0.0005, 0.0004, 0.0005], device='cuda:0') 2023-05-16 03:18:46,868 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.853e+02 2.787e+02 3.259e+02 4.117e+02 8.884e+02, threshold=6.518e+02, percent-clipped=2.0 2023-05-16 03:18:46,887 INFO [finetune.py:992] (0/2) Epoch 5, batch 9400, loss[loss=0.1541, simple_loss=0.2404, pruned_loss=0.03388, over 12353.00 frames. ], tot_loss[loss=0.1749, simple_loss=0.2628, pruned_loss=0.04349, over 2385506.59 frames. ], batch size: 30, lr: 4.65e-03, grad_scale: 8.0 2023-05-16 03:19:15,461 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.51 vs. limit=2.0 2023-05-16 03:19:22,466 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=158384.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 03:19:23,007 INFO [finetune.py:992] (0/2) Epoch 5, batch 9450, loss[loss=0.1571, simple_loss=0.2332, pruned_loss=0.04049, over 11860.00 frames. ], tot_loss[loss=0.1753, simple_loss=0.2631, pruned_loss=0.04375, over 2381976.00 frames. ], batch size: 26, lr: 4.64e-03, grad_scale: 8.0 2023-05-16 03:19:36,298 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.6264, 4.7566, 4.2758, 5.2241, 4.6782, 3.0868, 4.3779, 3.3474], device='cuda:0'), covar=tensor([0.0583, 0.0755, 0.1337, 0.0304, 0.0949, 0.1392, 0.0972, 0.2792], device='cuda:0'), in_proj_covar=tensor([0.0295, 0.0368, 0.0344, 0.0267, 0.0351, 0.0260, 0.0330, 0.0354], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 03:19:37,980 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.72 vs. limit=2.0 2023-05-16 03:19:38,418 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.7462, 2.7056, 4.0846, 4.3296, 2.9833, 2.6780, 2.7858, 2.1364], device='cuda:0'), covar=tensor([0.1282, 0.2368, 0.0525, 0.0387, 0.0995, 0.1885, 0.2375, 0.3548], device='cuda:0'), in_proj_covar=tensor([0.0272, 0.0367, 0.0264, 0.0285, 0.0252, 0.0282, 0.0350, 0.0354], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-16 03:19:59,140 INFO [finetune.py:992] (0/2) Epoch 5, batch 9500, loss[loss=0.1857, simple_loss=0.2721, pruned_loss=0.04964, over 12103.00 frames. ], tot_loss[loss=0.1755, simple_loss=0.2632, pruned_loss=0.04389, over 2378265.60 frames. ], batch size: 39, lr: 4.64e-03, grad_scale: 4.0 2023-05-16 03:19:59,873 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.979e+02 2.755e+02 3.329e+02 4.230e+02 9.398e+02, threshold=6.658e+02, percent-clipped=3.0 2023-05-16 03:20:06,417 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=158445.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 03:20:29,342 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=158478.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 03:20:34,328 INFO [finetune.py:992] (0/2) Epoch 5, batch 9550, loss[loss=0.1762, simple_loss=0.2665, pruned_loss=0.04294, over 11124.00 frames. ], tot_loss[loss=0.1757, simple_loss=0.2633, pruned_loss=0.04403, over 2372298.24 frames. ], batch size: 55, lr: 4.64e-03, grad_scale: 4.0 2023-05-16 03:20:51,532 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=158509.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 03:20:55,632 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.3530, 3.6057, 3.5138, 4.0568, 2.9274, 3.6291, 2.4832, 3.6651], device='cuda:0'), covar=tensor([0.1257, 0.0720, 0.1292, 0.0889, 0.1005, 0.0561, 0.1591, 0.0998], device='cuda:0'), in_proj_covar=tensor([0.0219, 0.0256, 0.0291, 0.0346, 0.0231, 0.0233, 0.0250, 0.0374], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 03:20:59,152 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.5771, 4.4952, 4.3454, 4.4226, 4.1034, 4.5987, 4.5608, 4.7483], device='cuda:0'), covar=tensor([0.0225, 0.0162, 0.0209, 0.0335, 0.0750, 0.0317, 0.0161, 0.0181], device='cuda:0'), in_proj_covar=tensor([0.0180, 0.0185, 0.0183, 0.0232, 0.0233, 0.0202, 0.0166, 0.0217], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0005, 0.0005, 0.0005, 0.0004, 0.0005], device='cuda:0') 2023-05-16 03:21:10,489 INFO [finetune.py:992] (0/2) Epoch 5, batch 9600, loss[loss=0.1483, simple_loss=0.2335, pruned_loss=0.03149, over 12084.00 frames. ], tot_loss[loss=0.1761, simple_loss=0.2637, pruned_loss=0.04422, over 2372059.47 frames. ], batch size: 32, lr: 4.64e-03, grad_scale: 8.0 2023-05-16 03:21:11,202 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.823e+02 3.116e+02 3.732e+02 4.442e+02 1.315e+03, threshold=7.464e+02, percent-clipped=4.0 2023-05-16 03:21:11,409 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.2520, 5.2310, 5.0137, 5.1858, 4.7571, 5.1571, 5.2028, 5.4557], device='cuda:0'), covar=tensor([0.0178, 0.0124, 0.0187, 0.0274, 0.0729, 0.0293, 0.0128, 0.0145], device='cuda:0'), in_proj_covar=tensor([0.0180, 0.0184, 0.0182, 0.0231, 0.0232, 0.0202, 0.0166, 0.0216], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0005, 0.0004, 0.0005, 0.0004, 0.0005], device='cuda:0') 2023-05-16 03:21:26,389 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=158557.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 03:21:46,387 INFO [finetune.py:992] (0/2) Epoch 5, batch 9650, loss[loss=0.1579, simple_loss=0.2361, pruned_loss=0.03988, over 12289.00 frames. ], tot_loss[loss=0.1757, simple_loss=0.2633, pruned_loss=0.04406, over 2372161.58 frames. ], batch size: 28, lr: 4.64e-03, grad_scale: 8.0 2023-05-16 03:21:58,280 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([6.0535, 6.0108, 5.7704, 5.3449, 5.2246, 5.9694, 5.5665, 5.3173], device='cuda:0'), covar=tensor([0.0603, 0.0890, 0.0694, 0.1484, 0.0529, 0.0626, 0.1356, 0.1005], device='cuda:0'), in_proj_covar=tensor([0.0574, 0.0515, 0.0494, 0.0603, 0.0390, 0.0676, 0.0737, 0.0535], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0004, 0.0004, 0.0003], device='cuda:0') 2023-05-16 03:22:22,042 INFO [finetune.py:992] (0/2) Epoch 5, batch 9700, loss[loss=0.1729, simple_loss=0.2495, pruned_loss=0.0481, over 12332.00 frames. ], tot_loss[loss=0.1765, simple_loss=0.2644, pruned_loss=0.04429, over 2371105.33 frames. ], batch size: 30, lr: 4.64e-03, grad_scale: 8.0 2023-05-16 03:22:22,703 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.968e+02 2.939e+02 3.389e+02 3.950e+02 7.191e+02, threshold=6.777e+02, percent-clipped=0.0 2023-05-16 03:22:26,782 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.89 vs. limit=2.0 2023-05-16 03:22:34,943 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.1812, 5.9324, 5.6062, 5.4387, 6.1101, 5.3792, 5.5874, 5.5428], device='cuda:0'), covar=tensor([0.1437, 0.0783, 0.0814, 0.1983, 0.0789, 0.1784, 0.1426, 0.0947], device='cuda:0'), in_proj_covar=tensor([0.0345, 0.0462, 0.0368, 0.0418, 0.0445, 0.0425, 0.0383, 0.0362], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 03:22:58,178 INFO [finetune.py:992] (0/2) Epoch 5, batch 9750, loss[loss=0.197, simple_loss=0.2858, pruned_loss=0.05415, over 12049.00 frames. ], tot_loss[loss=0.176, simple_loss=0.2638, pruned_loss=0.04412, over 2375274.94 frames. ], batch size: 40, lr: 4.64e-03, grad_scale: 8.0 2023-05-16 03:23:34,530 INFO [finetune.py:992] (0/2) Epoch 5, batch 9800, loss[loss=0.1774, simple_loss=0.2699, pruned_loss=0.04247, over 12293.00 frames. ], tot_loss[loss=0.1768, simple_loss=0.2647, pruned_loss=0.04444, over 2372386.50 frames. ], batch size: 34, lr: 4.64e-03, grad_scale: 8.0 2023-05-16 03:23:35,228 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.870e+02 2.850e+02 3.458e+02 4.279e+02 8.665e+02, threshold=6.916e+02, percent-clipped=3.0 2023-05-16 03:23:38,180 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=158740.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 03:23:38,322 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.2314, 2.5369, 3.7729, 3.1862, 3.5384, 3.3120, 2.6605, 3.6084], device='cuda:0'), covar=tensor([0.0131, 0.0307, 0.0107, 0.0210, 0.0139, 0.0141, 0.0304, 0.0111], device='cuda:0'), in_proj_covar=tensor([0.0165, 0.0196, 0.0177, 0.0176, 0.0201, 0.0153, 0.0187, 0.0169], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 03:24:00,713 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.2194, 4.9658, 5.2278, 5.1819, 4.3929, 4.5506, 4.6654, 4.9949], device='cuda:0'), covar=tensor([0.1045, 0.1030, 0.0816, 0.0872, 0.3404, 0.2294, 0.0761, 0.1633], device='cuda:0'), in_proj_covar=tensor([0.0500, 0.0650, 0.0553, 0.0597, 0.0798, 0.0709, 0.0524, 0.0481], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0005, 0.0004, 0.0004, 0.0005, 0.0005, 0.0004, 0.0004], device='cuda:0') 2023-05-16 03:24:04,315 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=158777.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 03:24:04,946 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=158778.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 03:24:08,514 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=158783.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 03:24:09,796 INFO [finetune.py:992] (0/2) Epoch 5, batch 9850, loss[loss=0.1668, simple_loss=0.2589, pruned_loss=0.03733, over 12158.00 frames. ], tot_loss[loss=0.1761, simple_loss=0.2636, pruned_loss=0.04429, over 2375308.10 frames. ], batch size: 34, lr: 4.64e-03, grad_scale: 8.0 2023-05-16 03:24:39,909 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=158826.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 03:24:46,324 INFO [finetune.py:992] (0/2) Epoch 5, batch 9900, loss[loss=0.1675, simple_loss=0.2554, pruned_loss=0.03977, over 11820.00 frames. ], tot_loss[loss=0.1764, simple_loss=0.2638, pruned_loss=0.04456, over 2374360.01 frames. ], batch size: 44, lr: 4.64e-03, grad_scale: 8.0 2023-05-16 03:24:46,549 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=158835.0, num_to_drop=1, layers_to_drop={0} 2023-05-16 03:24:47,058 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.995e+02 2.906e+02 3.304e+02 4.242e+02 8.853e+02, threshold=6.608e+02, percent-clipped=3.0 2023-05-16 03:24:48,627 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=158838.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 03:24:52,931 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=158844.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 03:25:22,454 INFO [finetune.py:992] (0/2) Epoch 5, batch 9950, loss[loss=0.2429, simple_loss=0.3173, pruned_loss=0.08429, over 8624.00 frames. ], tot_loss[loss=0.1753, simple_loss=0.2629, pruned_loss=0.0439, over 2378150.75 frames. ], batch size: 98, lr: 4.64e-03, grad_scale: 8.0 2023-05-16 03:25:30,736 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=158896.0, num_to_drop=1, layers_to_drop={2} 2023-05-16 03:25:36,198 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.0841, 5.8250, 5.4439, 5.4197, 5.9849, 5.2939, 5.5176, 5.4417], device='cuda:0'), covar=tensor([0.1391, 0.0907, 0.0961, 0.1820, 0.0875, 0.1953, 0.1611, 0.0990], device='cuda:0'), in_proj_covar=tensor([0.0346, 0.0464, 0.0366, 0.0418, 0.0446, 0.0427, 0.0381, 0.0361], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 03:25:43,786 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.87 vs. limit=2.0 2023-05-16 03:25:58,036 INFO [finetune.py:992] (0/2) Epoch 5, batch 10000, loss[loss=0.1619, simple_loss=0.2581, pruned_loss=0.03288, over 12155.00 frames. ], tot_loss[loss=0.1752, simple_loss=0.263, pruned_loss=0.04376, over 2383988.95 frames. ], batch size: 36, lr: 4.64e-03, grad_scale: 8.0 2023-05-16 03:25:58,700 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.990e+02 2.839e+02 3.264e+02 3.961e+02 6.729e+02, threshold=6.527e+02, percent-clipped=1.0 2023-05-16 03:26:28,604 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.2195, 5.0399, 5.0589, 5.2436, 5.0310, 5.1321, 5.0115, 2.7596], device='cuda:0'), covar=tensor([0.0079, 0.0045, 0.0068, 0.0042, 0.0033, 0.0081, 0.0063, 0.0604], device='cuda:0'), in_proj_covar=tensor([0.0062, 0.0071, 0.0075, 0.0068, 0.0056, 0.0085, 0.0073, 0.0088], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-16 03:26:34,000 INFO [finetune.py:992] (0/2) Epoch 5, batch 10050, loss[loss=0.1901, simple_loss=0.2886, pruned_loss=0.0458, over 12146.00 frames. ], tot_loss[loss=0.1752, simple_loss=0.263, pruned_loss=0.04369, over 2373904.31 frames. ], batch size: 34, lr: 4.64e-03, grad_scale: 8.0 2023-05-16 03:27:10,569 INFO [finetune.py:992] (0/2) Epoch 5, batch 10100, loss[loss=0.2277, simple_loss=0.3064, pruned_loss=0.07446, over 7968.00 frames. ], tot_loss[loss=0.1767, simple_loss=0.2641, pruned_loss=0.04467, over 2365708.15 frames. ], batch size: 98, lr: 4.64e-03, grad_scale: 8.0 2023-05-16 03:27:11,287 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.826e+02 3.065e+02 3.689e+02 4.338e+02 1.683e+03, threshold=7.377e+02, percent-clipped=4.0 2023-05-16 03:27:14,209 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=159040.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 03:27:35,322 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=159069.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 03:27:46,475 INFO [finetune.py:992] (0/2) Epoch 5, batch 10150, loss[loss=0.1539, simple_loss=0.2348, pruned_loss=0.03652, over 12182.00 frames. ], tot_loss[loss=0.1753, simple_loss=0.2628, pruned_loss=0.04386, over 2374353.13 frames. ], batch size: 31, lr: 4.64e-03, grad_scale: 8.0 2023-05-16 03:27:48,660 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=159088.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 03:28:08,489 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.8668, 3.7413, 3.4421, 3.2984, 2.9724, 2.7886, 3.7422, 2.3070], device='cuda:0'), covar=tensor([0.0251, 0.0094, 0.0136, 0.0135, 0.0321, 0.0292, 0.0088, 0.0400], device='cuda:0'), in_proj_covar=tensor([0.0185, 0.0155, 0.0149, 0.0177, 0.0199, 0.0190, 0.0157, 0.0197], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 03:28:19,007 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=159130.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 03:28:21,082 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=159133.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 03:28:22,431 INFO [finetune.py:992] (0/2) Epoch 5, batch 10200, loss[loss=0.1523, simple_loss=0.2361, pruned_loss=0.03429, over 12339.00 frames. ], tot_loss[loss=0.1754, simple_loss=0.2628, pruned_loss=0.04397, over 2375061.73 frames. ], batch size: 30, lr: 4.64e-03, grad_scale: 8.0 2023-05-16 03:28:23,161 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.159e+02 2.738e+02 3.327e+02 4.064e+02 6.892e+02, threshold=6.654e+02, percent-clipped=0.0 2023-05-16 03:28:25,376 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=159139.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 03:28:28,273 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.0588, 5.8762, 5.5786, 5.4132, 6.0021, 5.3915, 5.5112, 5.4840], device='cuda:0'), covar=tensor([0.1424, 0.0834, 0.0920, 0.2114, 0.0867, 0.1818, 0.1528, 0.0974], device='cuda:0'), in_proj_covar=tensor([0.0342, 0.0463, 0.0364, 0.0417, 0.0444, 0.0424, 0.0379, 0.0360], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 03:28:40,283 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=159159.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 03:28:44,733 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.89 vs. limit=2.0 2023-05-16 03:28:52,336 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.50 vs. limit=2.0 2023-05-16 03:28:58,530 INFO [finetune.py:992] (0/2) Epoch 5, batch 10250, loss[loss=0.1803, simple_loss=0.2658, pruned_loss=0.0474, over 12084.00 frames. ], tot_loss[loss=0.1744, simple_loss=0.2622, pruned_loss=0.04333, over 2375679.69 frames. ], batch size: 32, lr: 4.64e-03, grad_scale: 8.0 2023-05-16 03:29:02,753 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.3352, 6.0662, 5.6778, 5.6461, 6.1894, 5.3811, 5.6718, 5.7682], device='cuda:0'), covar=tensor([0.1243, 0.0838, 0.1032, 0.1652, 0.0848, 0.1794, 0.1737, 0.0925], device='cuda:0'), in_proj_covar=tensor([0.0341, 0.0461, 0.0363, 0.0416, 0.0442, 0.0422, 0.0377, 0.0358], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 03:29:02,763 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=159191.0, num_to_drop=1, layers_to_drop={3} 2023-05-16 03:29:23,482 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=159220.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 03:29:33,794 INFO [finetune.py:992] (0/2) Epoch 5, batch 10300, loss[loss=0.214, simple_loss=0.2863, pruned_loss=0.0709, over 8204.00 frames. ], tot_loss[loss=0.1751, simple_loss=0.263, pruned_loss=0.0436, over 2369364.21 frames. ], batch size: 97, lr: 4.64e-03, grad_scale: 8.0 2023-05-16 03:29:34,473 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.694e+02 2.968e+02 3.473e+02 4.181e+02 8.488e+02, threshold=6.946e+02, percent-clipped=6.0 2023-05-16 03:29:35,359 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.5059, 4.9756, 5.4424, 4.7681, 5.0262, 4.8395, 5.5244, 5.1166], device='cuda:0'), covar=tensor([0.0231, 0.0343, 0.0249, 0.0231, 0.0342, 0.0334, 0.0198, 0.0231], device='cuda:0'), in_proj_covar=tensor([0.0237, 0.0249, 0.0266, 0.0241, 0.0237, 0.0241, 0.0217, 0.0206], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 03:30:09,727 INFO [finetune.py:992] (0/2) Epoch 5, batch 10350, loss[loss=0.197, simple_loss=0.2852, pruned_loss=0.05441, over 12022.00 frames. ], tot_loss[loss=0.1747, simple_loss=0.2624, pruned_loss=0.04346, over 2377180.63 frames. ], batch size: 40, lr: 4.64e-03, grad_scale: 8.0 2023-05-16 03:30:14,209 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=159290.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 03:30:26,279 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.5615, 2.3494, 3.2679, 4.4525, 2.3139, 4.4203, 4.5261, 4.6185], device='cuda:0'), covar=tensor([0.0111, 0.1175, 0.0405, 0.0136, 0.1262, 0.0195, 0.0124, 0.0088], device='cuda:0'), in_proj_covar=tensor([0.0117, 0.0198, 0.0183, 0.0112, 0.0186, 0.0174, 0.0167, 0.0122], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 03:30:37,760 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=4.67 vs. limit=5.0 2023-05-16 03:30:45,685 INFO [finetune.py:992] (0/2) Epoch 5, batch 10400, loss[loss=0.1707, simple_loss=0.2561, pruned_loss=0.04263, over 11798.00 frames. ], tot_loss[loss=0.1752, simple_loss=0.2629, pruned_loss=0.04371, over 2374420.56 frames. ], batch size: 26, lr: 4.63e-03, grad_scale: 8.0 2023-05-16 03:30:46,381 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.159e+02 2.730e+02 3.340e+02 3.902e+02 7.144e+02, threshold=6.680e+02, percent-clipped=1.0 2023-05-16 03:30:57,224 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=159351.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 03:31:20,668 INFO [finetune.py:992] (0/2) Epoch 5, batch 10450, loss[loss=0.131, simple_loss=0.2175, pruned_loss=0.02223, over 12130.00 frames. ], tot_loss[loss=0.1745, simple_loss=0.2622, pruned_loss=0.04342, over 2382941.12 frames. ], batch size: 30, lr: 4.63e-03, grad_scale: 8.0 2023-05-16 03:31:49,895 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=159425.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 03:31:55,855 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=159433.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 03:31:57,738 INFO [finetune.py:992] (0/2) Epoch 5, batch 10500, loss[loss=0.2207, simple_loss=0.3056, pruned_loss=0.06788, over 12031.00 frames. ], tot_loss[loss=0.1748, simple_loss=0.2625, pruned_loss=0.0436, over 2384288.01 frames. ], batch size: 40, lr: 4.63e-03, grad_scale: 8.0 2023-05-16 03:31:58,428 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.673e+02 2.917e+02 3.389e+02 4.133e+02 8.328e+02, threshold=6.779e+02, percent-clipped=3.0 2023-05-16 03:31:59,423 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.3423, 2.5333, 3.8417, 3.3281, 3.6674, 3.4005, 2.7137, 3.6902], device='cuda:0'), covar=tensor([0.0101, 0.0326, 0.0159, 0.0179, 0.0118, 0.0160, 0.0288, 0.0093], device='cuda:0'), in_proj_covar=tensor([0.0165, 0.0194, 0.0177, 0.0175, 0.0200, 0.0152, 0.0187, 0.0169], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 03:32:00,777 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=159439.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 03:32:30,262 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=159481.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 03:32:33,163 INFO [finetune.py:992] (0/2) Epoch 5, batch 10550, loss[loss=0.141, simple_loss=0.2263, pruned_loss=0.02785, over 12156.00 frames. ], tot_loss[loss=0.1749, simple_loss=0.2629, pruned_loss=0.04348, over 2385327.71 frames. ], batch size: 29, lr: 4.63e-03, grad_scale: 8.0 2023-05-16 03:32:34,632 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=159487.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 03:32:37,535 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=159491.0, num_to_drop=1, layers_to_drop={0} 2023-05-16 03:32:54,619 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=159515.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 03:33:03,958 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-05-16 03:33:08,619 INFO [finetune.py:992] (0/2) Epoch 5, batch 10600, loss[loss=0.1627, simple_loss=0.2546, pruned_loss=0.03543, over 12157.00 frames. ], tot_loss[loss=0.1741, simple_loss=0.2621, pruned_loss=0.04299, over 2390335.30 frames. ], batch size: 34, lr: 4.63e-03, grad_scale: 8.0 2023-05-16 03:33:09,307 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.825e+02 2.829e+02 3.235e+02 4.043e+02 7.652e+02, threshold=6.471e+02, percent-clipped=2.0 2023-05-16 03:33:11,568 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=159539.0, num_to_drop=1, layers_to_drop={1} 2023-05-16 03:33:44,196 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.3644, 4.7378, 2.8507, 2.7610, 3.9587, 2.3657, 4.0615, 3.3079], device='cuda:0'), covar=tensor([0.0684, 0.0547, 0.1048, 0.1429, 0.0313, 0.1538, 0.0490, 0.0858], device='cuda:0'), in_proj_covar=tensor([0.0185, 0.0250, 0.0175, 0.0195, 0.0139, 0.0180, 0.0195, 0.0172], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 03:33:45,386 INFO [finetune.py:992] (0/2) Epoch 5, batch 10650, loss[loss=0.1669, simple_loss=0.2452, pruned_loss=0.04434, over 12341.00 frames. ], tot_loss[loss=0.1746, simple_loss=0.2625, pruned_loss=0.0434, over 2380647.98 frames. ], batch size: 30, lr: 4.63e-03, grad_scale: 8.0 2023-05-16 03:33:45,685 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.6517, 2.7703, 4.4355, 4.5765, 2.8668, 2.5430, 2.8357, 2.0338], device='cuda:0'), covar=tensor([0.1466, 0.2823, 0.0436, 0.0381, 0.1193, 0.2202, 0.2587, 0.3899], device='cuda:0'), in_proj_covar=tensor([0.0274, 0.0370, 0.0265, 0.0288, 0.0253, 0.0282, 0.0354, 0.0355], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-16 03:34:09,872 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.2944, 4.7559, 2.7727, 2.8319, 3.9798, 2.5887, 4.0003, 3.4832], device='cuda:0'), covar=tensor([0.0671, 0.0449, 0.1104, 0.1352, 0.0255, 0.1350, 0.0469, 0.0668], device='cuda:0'), in_proj_covar=tensor([0.0184, 0.0249, 0.0174, 0.0194, 0.0139, 0.0179, 0.0194, 0.0171], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 03:34:18,286 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=159631.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 03:34:20,980 INFO [finetune.py:992] (0/2) Epoch 5, batch 10700, loss[loss=0.1653, simple_loss=0.2565, pruned_loss=0.03701, over 12161.00 frames. ], tot_loss[loss=0.1741, simple_loss=0.2622, pruned_loss=0.04299, over 2390088.62 frames. ], batch size: 34, lr: 4.63e-03, grad_scale: 8.0 2023-05-16 03:34:21,646 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.020e+02 2.820e+02 3.467e+02 4.069e+02 7.983e+02, threshold=6.934e+02, percent-clipped=3.0 2023-05-16 03:34:28,985 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=159646.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 03:34:39,889 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.2419, 4.1560, 4.2607, 4.4776, 3.0890, 4.0664, 2.6015, 4.0434], device='cuda:0'), covar=tensor([0.1574, 0.0621, 0.0799, 0.0538, 0.1079, 0.0511, 0.1745, 0.1146], device='cuda:0'), in_proj_covar=tensor([0.0218, 0.0254, 0.0289, 0.0343, 0.0230, 0.0229, 0.0249, 0.0371], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 03:34:56,620 INFO [finetune.py:992] (0/2) Epoch 5, batch 10750, loss[loss=0.1463, simple_loss=0.2295, pruned_loss=0.03152, over 12349.00 frames. ], tot_loss[loss=0.1741, simple_loss=0.2623, pruned_loss=0.04297, over 2381992.13 frames. ], batch size: 30, lr: 4.63e-03, grad_scale: 8.0 2023-05-16 03:35:01,564 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=159692.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 03:35:26,214 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=159725.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 03:35:33,089 INFO [finetune.py:992] (0/2) Epoch 5, batch 10800, loss[loss=0.1665, simple_loss=0.2662, pruned_loss=0.03343, over 12357.00 frames. ], tot_loss[loss=0.1754, simple_loss=0.2634, pruned_loss=0.04365, over 2374206.84 frames. ], batch size: 35, lr: 4.63e-03, grad_scale: 8.0 2023-05-16 03:35:33,847 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.865e+02 2.840e+02 3.248e+02 3.728e+02 9.603e+02, threshold=6.496e+02, percent-clipped=2.0 2023-05-16 03:35:43,763 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.1101, 5.0875, 4.9848, 5.0072, 4.5988, 5.0891, 5.0730, 5.2879], device='cuda:0'), covar=tensor([0.0241, 0.0121, 0.0147, 0.0228, 0.0754, 0.0273, 0.0143, 0.0141], device='cuda:0'), in_proj_covar=tensor([0.0181, 0.0183, 0.0181, 0.0229, 0.0232, 0.0199, 0.0165, 0.0214], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0005, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:0') 2023-05-16 03:36:00,160 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=159773.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 03:36:05,411 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.38 vs. limit=2.0 2023-05-16 03:36:08,585 INFO [finetune.py:992] (0/2) Epoch 5, batch 10850, loss[loss=0.1621, simple_loss=0.2491, pruned_loss=0.03757, over 12042.00 frames. ], tot_loss[loss=0.1755, simple_loss=0.2634, pruned_loss=0.04384, over 2374069.18 frames. ], batch size: 31, lr: 4.63e-03, grad_scale: 8.0 2023-05-16 03:36:31,078 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=159815.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 03:36:45,688 INFO [finetune.py:992] (0/2) Epoch 5, batch 10900, loss[loss=0.1811, simple_loss=0.278, pruned_loss=0.04204, over 11490.00 frames. ], tot_loss[loss=0.1755, simple_loss=0.2636, pruned_loss=0.04375, over 2377006.97 frames. ], batch size: 48, lr: 4.63e-03, grad_scale: 8.0 2023-05-16 03:36:46,402 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.011e+02 2.789e+02 3.278e+02 4.088e+02 6.595e+02, threshold=6.557e+02, percent-clipped=3.0 2023-05-16 03:36:50,915 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.88 vs. limit=2.0 2023-05-16 03:37:00,777 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=159855.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 03:37:05,113 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.0923, 2.0930, 2.6413, 3.0970, 2.1328, 3.1336, 3.0419, 3.2422], device='cuda:0'), covar=tensor([0.0184, 0.0985, 0.0447, 0.0174, 0.1001, 0.0354, 0.0367, 0.0148], device='cuda:0'), in_proj_covar=tensor([0.0120, 0.0202, 0.0188, 0.0115, 0.0190, 0.0179, 0.0172, 0.0125], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 03:37:06,412 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=159863.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 03:37:07,343 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.5935, 2.8583, 4.4098, 4.7032, 2.9495, 2.7596, 3.0090, 2.0339], device='cuda:0'), covar=tensor([0.1429, 0.2866, 0.0506, 0.0334, 0.1088, 0.1841, 0.2377, 0.3880], device='cuda:0'), in_proj_covar=tensor([0.0274, 0.0369, 0.0265, 0.0288, 0.0253, 0.0282, 0.0353, 0.0354], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-16 03:37:11,559 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.5208, 2.1977, 3.6518, 4.4491, 3.9875, 4.3809, 3.8924, 2.9197], device='cuda:0'), covar=tensor([0.0029, 0.0452, 0.0112, 0.0036, 0.0092, 0.0069, 0.0112, 0.0374], device='cuda:0'), in_proj_covar=tensor([0.0082, 0.0120, 0.0102, 0.0075, 0.0097, 0.0113, 0.0089, 0.0134], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 03:37:22,634 INFO [finetune.py:992] (0/2) Epoch 5, batch 10950, loss[loss=0.1746, simple_loss=0.2711, pruned_loss=0.039, over 12136.00 frames. ], tot_loss[loss=0.1768, simple_loss=0.2645, pruned_loss=0.04459, over 2365033.63 frames. ], batch size: 34, lr: 4.63e-03, grad_scale: 8.0 2023-05-16 03:37:40,238 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=159910.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 03:37:44,452 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=159916.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 03:37:57,645 INFO [finetune.py:992] (0/2) Epoch 5, batch 11000, loss[loss=0.2524, simple_loss=0.3295, pruned_loss=0.0876, over 7550.00 frames. ], tot_loss[loss=0.1788, simple_loss=0.2664, pruned_loss=0.04559, over 2351288.22 frames. ], batch size: 98, lr: 4.63e-03, grad_scale: 8.0 2023-05-16 03:37:58,346 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.944e+02 3.202e+02 3.873e+02 4.673e+02 7.110e+02, threshold=7.746e+02, percent-clipped=2.0 2023-05-16 03:38:05,596 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=159946.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 03:38:23,968 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=159971.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 03:38:34,240 INFO [finetune.py:992] (0/2) Epoch 5, batch 11050, loss[loss=0.2721, simple_loss=0.3495, pruned_loss=0.09739, over 10400.00 frames. ], tot_loss[loss=0.1828, simple_loss=0.2701, pruned_loss=0.04776, over 2314886.75 frames. ], batch size: 68, lr: 4.63e-03, grad_scale: 8.0 2023-05-16 03:38:35,718 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=159987.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 03:38:39,599 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.34 vs. limit=2.0 2023-05-16 03:38:40,681 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=159994.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 03:38:45,096 INFO [checkpoint.py:75] (0/2) Saving checkpoint to pruned_transducer_stateless7/exp_giga_finetune/checkpoint-60000.pt 2023-05-16 03:39:09,555 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.8500, 3.7171, 3.7284, 3.8678, 3.5296, 3.8843, 3.8615, 3.9809], device='cuda:0'), covar=tensor([0.0209, 0.0175, 0.0182, 0.0302, 0.0586, 0.0356, 0.0182, 0.0208], device='cuda:0'), in_proj_covar=tensor([0.0180, 0.0182, 0.0181, 0.0227, 0.0230, 0.0200, 0.0164, 0.0212], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0005, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:0') 2023-05-16 03:39:12,045 INFO [finetune.py:992] (0/2) Epoch 5, batch 11100, loss[loss=0.2514, simple_loss=0.3345, pruned_loss=0.08411, over 11186.00 frames. ], tot_loss[loss=0.188, simple_loss=0.2746, pruned_loss=0.0507, over 2268391.70 frames. ], batch size: 55, lr: 4.63e-03, grad_scale: 8.0 2023-05-16 03:39:12,758 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.336e+02 3.352e+02 3.764e+02 4.543e+02 1.015e+03, threshold=7.528e+02, percent-clipped=3.0 2023-05-16 03:39:48,065 INFO [finetune.py:992] (0/2) Epoch 5, batch 11150, loss[loss=0.286, simple_loss=0.3487, pruned_loss=0.1116, over 6853.00 frames. ], tot_loss[loss=0.1958, simple_loss=0.2812, pruned_loss=0.05525, over 2203854.31 frames. ], batch size: 101, lr: 4.63e-03, grad_scale: 8.0 2023-05-16 03:39:56,062 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.7933, 4.0441, 3.7337, 4.5911, 3.9806, 2.7965, 3.9010, 2.9387], device='cuda:0'), covar=tensor([0.0865, 0.0900, 0.1267, 0.0342, 0.1290, 0.1504, 0.0987, 0.2925], device='cuda:0'), in_proj_covar=tensor([0.0289, 0.0361, 0.0337, 0.0260, 0.0345, 0.0255, 0.0322, 0.0349], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 03:40:06,145 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.0782, 3.9647, 2.5305, 2.3072, 3.4125, 2.3991, 3.6544, 2.8334], device='cuda:0'), covar=tensor([0.0557, 0.0361, 0.0983, 0.1462, 0.0255, 0.1226, 0.0410, 0.0824], device='cuda:0'), in_proj_covar=tensor([0.0181, 0.0246, 0.0172, 0.0192, 0.0136, 0.0177, 0.0191, 0.0170], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 03:40:17,966 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.1445, 2.1338, 3.1241, 4.1563, 2.3144, 4.2221, 4.0934, 4.2704], device='cuda:0'), covar=tensor([0.0132, 0.1349, 0.0426, 0.0098, 0.1310, 0.0187, 0.0235, 0.0082], device='cuda:0'), in_proj_covar=tensor([0.0117, 0.0199, 0.0183, 0.0111, 0.0185, 0.0173, 0.0167, 0.0122], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 03:40:21,727 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=2.02 vs. limit=2.0 2023-05-16 03:40:22,757 INFO [finetune.py:992] (0/2) Epoch 5, batch 11200, loss[loss=0.2016, simple_loss=0.3013, pruned_loss=0.05096, over 10225.00 frames. ], tot_loss[loss=0.2025, simple_loss=0.2871, pruned_loss=0.05896, over 2159534.41 frames. ], batch size: 68, lr: 4.63e-03, grad_scale: 8.0 2023-05-16 03:40:23,513 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.431e+02 3.491e+02 4.351e+02 5.198e+02 9.612e+02, threshold=8.701e+02, percent-clipped=4.0 2023-05-16 03:40:58,854 INFO [finetune.py:992] (0/2) Epoch 5, batch 11250, loss[loss=0.3005, simple_loss=0.355, pruned_loss=0.1229, over 6695.00 frames. ], tot_loss[loss=0.2111, simple_loss=0.2942, pruned_loss=0.06401, over 2086511.53 frames. ], batch size: 98, lr: 4.63e-03, grad_scale: 8.0 2023-05-16 03:41:01,687 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.5988, 2.6841, 3.8937, 4.1275, 2.9218, 2.6508, 2.6926, 1.9779], device='cuda:0'), covar=tensor([0.1362, 0.2639, 0.0516, 0.0381, 0.1020, 0.2052, 0.2523, 0.4076], device='cuda:0'), in_proj_covar=tensor([0.0269, 0.0363, 0.0259, 0.0283, 0.0248, 0.0277, 0.0347, 0.0348], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-16 03:41:13,401 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.8119, 3.7123, 3.8090, 3.5883, 3.7272, 3.5906, 3.7939, 3.5505], device='cuda:0'), covar=tensor([0.0338, 0.0306, 0.0311, 0.0217, 0.0286, 0.0272, 0.0292, 0.0875], device='cuda:0'), in_proj_covar=tensor([0.0224, 0.0235, 0.0250, 0.0228, 0.0224, 0.0225, 0.0205, 0.0192], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 03:41:16,643 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=160211.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 03:41:34,347 INFO [finetune.py:992] (0/2) Epoch 5, batch 11300, loss[loss=0.2885, simple_loss=0.3479, pruned_loss=0.1145, over 7146.00 frames. ], tot_loss[loss=0.2185, simple_loss=0.3001, pruned_loss=0.06842, over 2028790.52 frames. ], batch size: 101, lr: 4.63e-03, grad_scale: 8.0 2023-05-16 03:41:35,057 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.451e+02 3.864e+02 4.459e+02 5.256e+02 1.641e+03, threshold=8.918e+02, percent-clipped=3.0 2023-05-16 03:41:55,463 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=160266.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 03:42:08,360 INFO [finetune.py:992] (0/2) Epoch 5, batch 11350, loss[loss=0.2361, simple_loss=0.3186, pruned_loss=0.07681, over 10397.00 frames. ], tot_loss[loss=0.2225, simple_loss=0.3038, pruned_loss=0.07062, over 2003849.12 frames. ], batch size: 68, lr: 4.63e-03, grad_scale: 8.0 2023-05-16 03:42:10,660 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=160287.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 03:42:10,745 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.3848, 2.8674, 3.7088, 2.3449, 2.5503, 3.1530, 2.9399, 3.2299], device='cuda:0'), covar=tensor([0.0469, 0.1012, 0.0238, 0.1164, 0.1620, 0.0978, 0.1039, 0.0804], device='cuda:0'), in_proj_covar=tensor([0.0224, 0.0222, 0.0224, 0.0173, 0.0227, 0.0272, 0.0215, 0.0239], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:0') 2023-05-16 03:42:24,898 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=160307.0, num_to_drop=1, layers_to_drop={1} 2023-05-16 03:42:35,013 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([1.9168, 2.2575, 2.2961, 2.1964, 2.0221, 1.9535, 2.0496, 1.7078], device='cuda:0'), covar=tensor([0.0275, 0.0125, 0.0146, 0.0161, 0.0244, 0.0186, 0.0142, 0.0336], device='cuda:0'), in_proj_covar=tensor([0.0185, 0.0154, 0.0150, 0.0177, 0.0198, 0.0191, 0.0158, 0.0195], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 03:42:43,627 INFO [finetune.py:992] (0/2) Epoch 5, batch 11400, loss[loss=0.2607, simple_loss=0.3319, pruned_loss=0.09471, over 7241.00 frames. ], tot_loss[loss=0.2296, simple_loss=0.3093, pruned_loss=0.07495, over 1934297.11 frames. ], batch size: 98, lr: 4.62e-03, grad_scale: 8.0 2023-05-16 03:42:43,724 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=160335.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 03:42:44,359 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.326e+02 3.579e+02 4.391e+02 4.990e+02 9.272e+02, threshold=8.781e+02, percent-clipped=0.0 2023-05-16 03:43:07,363 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=160368.0, num_to_drop=1, layers_to_drop={0} 2023-05-16 03:43:19,019 INFO [finetune.py:992] (0/2) Epoch 5, batch 11450, loss[loss=0.3066, simple_loss=0.3608, pruned_loss=0.1262, over 7335.00 frames. ], tot_loss[loss=0.2358, simple_loss=0.3137, pruned_loss=0.07895, over 1868691.56 frames. ], batch size: 99, lr: 4.62e-03, grad_scale: 8.0 2023-05-16 03:43:30,467 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.6639, 2.6614, 3.9624, 4.1375, 3.0335, 2.6284, 2.7202, 1.9922], device='cuda:0'), covar=tensor([0.1431, 0.2741, 0.0514, 0.0432, 0.1013, 0.2121, 0.2563, 0.4300], device='cuda:0'), in_proj_covar=tensor([0.0267, 0.0359, 0.0256, 0.0280, 0.0247, 0.0275, 0.0346, 0.0347], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-16 03:43:32,022 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=4.58 vs. limit=5.0 2023-05-16 03:43:36,721 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.4281, 4.3625, 4.3273, 4.0249, 4.0942, 4.4128, 4.1161, 3.9907], device='cuda:0'), covar=tensor([0.0760, 0.0978, 0.0673, 0.1429, 0.1732, 0.0792, 0.1455, 0.1176], device='cuda:0'), in_proj_covar=tensor([0.0555, 0.0495, 0.0473, 0.0577, 0.0377, 0.0645, 0.0699, 0.0517], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-05-16 03:43:54,189 INFO [finetune.py:992] (0/2) Epoch 5, batch 11500, loss[loss=0.2736, simple_loss=0.3367, pruned_loss=0.1052, over 6832.00 frames. ], tot_loss[loss=0.2388, simple_loss=0.316, pruned_loss=0.08079, over 1848993.35 frames. ], batch size: 99, lr: 4.62e-03, grad_scale: 16.0 2023-05-16 03:43:54,824 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.655e+02 3.566e+02 4.414e+02 5.138e+02 9.853e+02, threshold=8.829e+02, percent-clipped=2.0 2023-05-16 03:44:30,038 INFO [finetune.py:992] (0/2) Epoch 5, batch 11550, loss[loss=0.2281, simple_loss=0.3127, pruned_loss=0.07168, over 11304.00 frames. ], tot_loss[loss=0.2418, simple_loss=0.3179, pruned_loss=0.08283, over 1808207.49 frames. ], batch size: 56, lr: 4.62e-03, grad_scale: 16.0 2023-05-16 03:44:41,848 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.2533, 3.4430, 3.2427, 3.6038, 3.4270, 2.5034, 3.2015, 2.8586], device='cuda:0'), covar=tensor([0.0766, 0.0853, 0.1137, 0.0512, 0.0999, 0.1523, 0.0956, 0.2437], device='cuda:0'), in_proj_covar=tensor([0.0282, 0.0350, 0.0327, 0.0251, 0.0338, 0.0250, 0.0316, 0.0342], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-16 03:44:47,743 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=160511.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 03:44:53,137 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.0190, 4.5560, 4.2010, 4.3317, 4.6000, 4.0709, 4.2704, 4.2038], device='cuda:0'), covar=tensor([0.1251, 0.0949, 0.1275, 0.1706, 0.0930, 0.1899, 0.1396, 0.1165], device='cuda:0'), in_proj_covar=tensor([0.0333, 0.0437, 0.0354, 0.0401, 0.0428, 0.0408, 0.0364, 0.0349], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002], device='cuda:0') 2023-05-16 03:45:02,156 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.8279, 2.4680, 3.4482, 3.5767, 2.8537, 2.7201, 2.6376, 2.3382], device='cuda:0'), covar=tensor([0.1078, 0.2482, 0.0620, 0.0429, 0.0836, 0.1763, 0.2255, 0.3854], device='cuda:0'), in_proj_covar=tensor([0.0268, 0.0360, 0.0256, 0.0279, 0.0247, 0.0276, 0.0346, 0.0349], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-16 03:45:03,388 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.1378, 4.3622, 2.5282, 2.3047, 3.8637, 2.3622, 3.9521, 2.9066], device='cuda:0'), covar=tensor([0.0693, 0.0399, 0.1259, 0.1773, 0.0192, 0.1589, 0.0334, 0.0927], device='cuda:0'), in_proj_covar=tensor([0.0179, 0.0239, 0.0171, 0.0191, 0.0134, 0.0177, 0.0186, 0.0168], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 03:45:03,837 INFO [finetune.py:992] (0/2) Epoch 5, batch 11600, loss[loss=0.2266, simple_loss=0.3121, pruned_loss=0.07054, over 11618.00 frames. ], tot_loss[loss=0.244, simple_loss=0.3192, pruned_loss=0.08439, over 1779492.51 frames. ], batch size: 48, lr: 4.62e-03, grad_scale: 16.0 2023-05-16 03:45:04,461 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.282e+02 3.465e+02 4.046e+02 4.621e+02 7.818e+02, threshold=8.092e+02, percent-clipped=0.0 2023-05-16 03:45:21,380 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=160559.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 03:45:26,654 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=160566.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 03:45:40,418 INFO [finetune.py:992] (0/2) Epoch 5, batch 11650, loss[loss=0.3117, simple_loss=0.367, pruned_loss=0.1282, over 6776.00 frames. ], tot_loss[loss=0.2443, simple_loss=0.3189, pruned_loss=0.08486, over 1761400.56 frames. ], batch size: 98, lr: 4.62e-03, grad_scale: 16.0 2023-05-16 03:46:02,218 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=160614.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 03:46:03,213 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.56 vs. limit=2.0 2023-05-16 03:46:16,226 INFO [finetune.py:992] (0/2) Epoch 5, batch 11700, loss[loss=0.272, simple_loss=0.3368, pruned_loss=0.1036, over 6463.00 frames. ], tot_loss[loss=0.2443, simple_loss=0.3186, pruned_loss=0.08503, over 1744016.34 frames. ], batch size: 98, lr: 4.62e-03, grad_scale: 16.0 2023-05-16 03:46:16,861 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.596e+02 3.486e+02 4.193e+02 4.820e+02 1.126e+03, threshold=8.386e+02, percent-clipped=4.0 2023-05-16 03:46:17,022 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.5018, 4.4576, 4.3567, 4.0076, 4.0720, 4.4739, 4.2325, 4.0674], device='cuda:0'), covar=tensor([0.0733, 0.0923, 0.0681, 0.1389, 0.2487, 0.0814, 0.1288, 0.1091], device='cuda:0'), in_proj_covar=tensor([0.0548, 0.0490, 0.0466, 0.0568, 0.0373, 0.0638, 0.0689, 0.0508], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-05-16 03:46:35,709 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=160663.0, num_to_drop=1, layers_to_drop={2} 2023-05-16 03:46:51,082 INFO [finetune.py:992] (0/2) Epoch 5, batch 11750, loss[loss=0.2636, simple_loss=0.3273, pruned_loss=0.09992, over 7153.00 frames. ], tot_loss[loss=0.2452, simple_loss=0.3191, pruned_loss=0.08565, over 1740793.42 frames. ], batch size: 97, lr: 4.62e-03, grad_scale: 16.0 2023-05-16 03:47:20,815 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.48 vs. limit=2.0 2023-05-16 03:47:25,816 INFO [finetune.py:992] (0/2) Epoch 5, batch 11800, loss[loss=0.2408, simple_loss=0.329, pruned_loss=0.0763, over 10352.00 frames. ], tot_loss[loss=0.2486, simple_loss=0.3215, pruned_loss=0.08785, over 1718431.16 frames. ], batch size: 69, lr: 4.62e-03, grad_scale: 16.0 2023-05-16 03:47:26,498 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.148e+02 3.837e+02 4.388e+02 5.262e+02 1.354e+03, threshold=8.776e+02, percent-clipped=6.0 2023-05-16 03:47:47,424 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=4.52 vs. limit=5.0 2023-05-16 03:47:56,831 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.38 vs. limit=2.0 2023-05-16 03:48:00,546 INFO [finetune.py:992] (0/2) Epoch 5, batch 11850, loss[loss=0.291, simple_loss=0.3464, pruned_loss=0.1178, over 6724.00 frames. ], tot_loss[loss=0.2501, simple_loss=0.323, pruned_loss=0.08858, over 1694777.52 frames. ], batch size: 98, lr: 4.62e-03, grad_scale: 16.0 2023-05-16 03:48:03,532 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.4535, 2.8655, 3.7355, 2.3922, 2.6027, 3.0370, 2.8670, 3.0873], device='cuda:0'), covar=tensor([0.0566, 0.1193, 0.0259, 0.1244, 0.1706, 0.1274, 0.1347, 0.1107], device='cuda:0'), in_proj_covar=tensor([0.0219, 0.0218, 0.0214, 0.0171, 0.0223, 0.0265, 0.0212, 0.0232], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:0') 2023-05-16 03:48:16,251 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.3804, 3.1477, 3.1797, 3.3751, 2.6862, 3.1524, 2.7060, 2.9003], device='cuda:0'), covar=tensor([0.1490, 0.0819, 0.0917, 0.0663, 0.0961, 0.0710, 0.1530, 0.0830], device='cuda:0'), in_proj_covar=tensor([0.0222, 0.0257, 0.0290, 0.0340, 0.0232, 0.0234, 0.0253, 0.0370], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 03:48:34,798 INFO [finetune.py:992] (0/2) Epoch 5, batch 11900, loss[loss=0.2558, simple_loss=0.3252, pruned_loss=0.09315, over 7200.00 frames. ], tot_loss[loss=0.2496, simple_loss=0.3232, pruned_loss=0.08797, over 1686263.38 frames. ], batch size: 98, lr: 4.62e-03, grad_scale: 16.0 2023-05-16 03:48:35,449 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.711e+02 3.579e+02 4.133e+02 4.663e+02 8.316e+02, threshold=8.266e+02, percent-clipped=0.0 2023-05-16 03:49:09,817 INFO [finetune.py:992] (0/2) Epoch 5, batch 11950, loss[loss=0.2152, simple_loss=0.2906, pruned_loss=0.06991, over 6629.00 frames. ], tot_loss[loss=0.2446, simple_loss=0.3194, pruned_loss=0.08488, over 1669809.13 frames. ], batch size: 99, lr: 4.62e-03, grad_scale: 16.0 2023-05-16 03:49:18,163 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.9594, 2.1728, 2.6577, 3.0734, 2.2198, 3.1112, 3.0303, 3.1660], device='cuda:0'), covar=tensor([0.0183, 0.1077, 0.0461, 0.0172, 0.1113, 0.0259, 0.0263, 0.0136], device='cuda:0'), in_proj_covar=tensor([0.0112, 0.0191, 0.0174, 0.0105, 0.0178, 0.0162, 0.0157, 0.0115], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 03:49:21,536 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=160901.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 03:49:31,823 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.7879, 3.6434, 3.7173, 3.8344, 3.4970, 3.8363, 3.8180, 3.9279], device='cuda:0'), covar=tensor([0.0190, 0.0174, 0.0196, 0.0243, 0.0570, 0.0291, 0.0172, 0.0230], device='cuda:0'), in_proj_covar=tensor([0.0157, 0.0159, 0.0158, 0.0199, 0.0201, 0.0176, 0.0144, 0.0189], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:0') 2023-05-16 03:49:45,479 INFO [finetune.py:992] (0/2) Epoch 5, batch 12000, loss[loss=0.209, simple_loss=0.2914, pruned_loss=0.06329, over 7423.00 frames. ], tot_loss[loss=0.238, simple_loss=0.3145, pruned_loss=0.0808, over 1671951.42 frames. ], batch size: 99, lr: 4.62e-03, grad_scale: 16.0 2023-05-16 03:49:45,480 INFO [finetune.py:1017] (0/2) Computing validation loss 2023-05-16 03:49:54,891 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.2889, 2.4734, 3.1791, 4.2931, 2.4569, 4.4507, 4.3155, 4.3553], device='cuda:0'), covar=tensor([0.0119, 0.1322, 0.0459, 0.0117, 0.1438, 0.0132, 0.0161, 0.0079], device='cuda:0'), in_proj_covar=tensor([0.0111, 0.0190, 0.0174, 0.0105, 0.0177, 0.0161, 0.0156, 0.0114], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 03:49:58,096 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.6790, 5.5324, 5.5299, 4.8888, 4.9703, 5.6141, 5.2261, 5.2315], device='cuda:0'), covar=tensor([0.0581, 0.0990, 0.0532, 0.1567, 0.0464, 0.0496, 0.1171, 0.0706], device='cuda:0'), in_proj_covar=tensor([0.0534, 0.0476, 0.0451, 0.0549, 0.0361, 0.0616, 0.0664, 0.0491], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002], device='cuda:0') 2023-05-16 03:50:03,324 INFO [finetune.py:1026] (0/2) Epoch 5, validation: loss=0.2936, simple_loss=0.368, pruned_loss=0.1096, over 1020973.00 frames. 2023-05-16 03:50:03,325 INFO [finetune.py:1027] (0/2) Maximum memory allocated so far is 12745MB 2023-05-16 03:50:03,982 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.975e+02 3.113e+02 3.392e+02 4.101e+02 1.232e+03, threshold=6.784e+02, percent-clipped=2.0 2023-05-16 03:50:06,461 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=4.60 vs. limit=5.0 2023-05-16 03:50:15,706 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.5769, 4.4266, 4.5394, 4.5605, 4.4242, 4.6424, 4.4211, 2.3395], device='cuda:0'), covar=tensor([0.0101, 0.0060, 0.0075, 0.0057, 0.0043, 0.0086, 0.0072, 0.0867], device='cuda:0'), in_proj_covar=tensor([0.0059, 0.0067, 0.0070, 0.0064, 0.0053, 0.0079, 0.0069, 0.0085], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-16 03:50:22,320 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.6423, 4.5433, 4.5940, 4.6536, 4.3451, 4.3906, 4.3184, 4.5175], device='cuda:0'), covar=tensor([0.0619, 0.0617, 0.0771, 0.0629, 0.1813, 0.1212, 0.0542, 0.1044], device='cuda:0'), in_proj_covar=tensor([0.0449, 0.0578, 0.0500, 0.0530, 0.0698, 0.0628, 0.0468, 0.0435], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0004, 0.0004, 0.0005, 0.0004, 0.0003, 0.0003], device='cuda:0') 2023-05-16 03:50:23,012 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=160962.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 03:50:23,660 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=160963.0, num_to_drop=1, layers_to_drop={2} 2023-05-16 03:50:24,988 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=160965.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 03:50:38,408 INFO [finetune.py:992] (0/2) Epoch 5, batch 12050, loss[loss=0.181, simple_loss=0.2664, pruned_loss=0.04782, over 6835.00 frames. ], tot_loss[loss=0.2321, simple_loss=0.3096, pruned_loss=0.07733, over 1668637.61 frames. ], batch size: 99, lr: 4.62e-03, grad_scale: 16.0 2023-05-16 03:50:39,941 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.0578, 4.8924, 4.9459, 5.0227, 4.8130, 5.1000, 4.9180, 2.2743], device='cuda:0'), covar=tensor([0.0100, 0.0054, 0.0072, 0.0054, 0.0043, 0.0074, 0.0071, 0.0927], device='cuda:0'), in_proj_covar=tensor([0.0059, 0.0066, 0.0070, 0.0064, 0.0053, 0.0079, 0.0069, 0.0085], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-16 03:50:42,140 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=2.10 vs. limit=2.0 2023-05-16 03:50:56,234 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=161011.0, num_to_drop=1, layers_to_drop={1} 2023-05-16 03:51:01,985 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.8983, 3.7478, 3.9127, 3.6171, 3.7428, 3.6509, 3.8734, 3.5787], device='cuda:0'), covar=tensor([0.0333, 0.0388, 0.0335, 0.0277, 0.0362, 0.0300, 0.0339, 0.0962], device='cuda:0'), in_proj_covar=tensor([0.0212, 0.0222, 0.0235, 0.0219, 0.0212, 0.0213, 0.0194, 0.0180], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 03:51:05,937 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=161026.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 03:51:12,206 INFO [finetune.py:992] (0/2) Epoch 5, batch 12100, loss[loss=0.2738, simple_loss=0.3352, pruned_loss=0.1062, over 6926.00 frames. ], tot_loss[loss=0.2302, simple_loss=0.3082, pruned_loss=0.07611, over 1658202.96 frames. ], batch size: 98, lr: 4.62e-03, grad_scale: 16.0 2023-05-16 03:51:12,852 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.086e+02 2.989e+02 3.607e+02 4.106e+02 1.013e+03, threshold=7.215e+02, percent-clipped=2.0 2023-05-16 03:51:14,872 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.8265, 2.1091, 3.2694, 3.6914, 3.5163, 3.6690, 3.4375, 2.6912], device='cuda:0'), covar=tensor([0.0039, 0.0428, 0.0113, 0.0035, 0.0098, 0.0072, 0.0093, 0.0346], device='cuda:0'), in_proj_covar=tensor([0.0076, 0.0116, 0.0094, 0.0070, 0.0090, 0.0106, 0.0084, 0.0129], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 03:51:21,138 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.6915, 4.5947, 4.8266, 4.7215, 4.1526, 4.1908, 4.4297, 4.5316], device='cuda:0'), covar=tensor([0.1179, 0.0991, 0.0878, 0.0908, 0.3012, 0.2257, 0.0623, 0.1653], device='cuda:0'), in_proj_covar=tensor([0.0448, 0.0577, 0.0500, 0.0529, 0.0696, 0.0626, 0.0468, 0.0435], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0004, 0.0004, 0.0005, 0.0004, 0.0003, 0.0003], device='cuda:0') 2023-05-16 03:51:23,734 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.0567, 2.0915, 2.7456, 2.9327, 2.9352, 2.9886, 2.8213, 2.4316], device='cuda:0'), covar=tensor([0.0059, 0.0379, 0.0155, 0.0052, 0.0106, 0.0098, 0.0110, 0.0302], device='cuda:0'), in_proj_covar=tensor([0.0076, 0.0116, 0.0095, 0.0070, 0.0091, 0.0106, 0.0084, 0.0130], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 03:51:43,684 INFO [finetune.py:992] (0/2) Epoch 5, batch 12150, loss[loss=0.2366, simple_loss=0.3054, pruned_loss=0.08393, over 7411.00 frames. ], tot_loss[loss=0.2314, simple_loss=0.3093, pruned_loss=0.0767, over 1659143.61 frames. ], batch size: 99, lr: 4.62e-03, grad_scale: 16.0 2023-05-16 03:52:14,798 INFO [finetune.py:992] (0/2) Epoch 5, batch 12200, loss[loss=0.2359, simple_loss=0.3051, pruned_loss=0.08335, over 6961.00 frames. ], tot_loss[loss=0.2322, simple_loss=0.31, pruned_loss=0.0772, over 1653388.31 frames. ], batch size: 100, lr: 4.62e-03, grad_scale: 16.0 2023-05-16 03:52:15,381 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.269e+02 3.485e+02 4.182e+02 5.020e+02 1.258e+03, threshold=8.364e+02, percent-clipped=6.0 2023-05-16 03:52:34,419 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.77 vs. limit=2.0 2023-05-16 03:52:36,329 INFO [checkpoint.py:75] (0/2) Saving checkpoint to pruned_transducer_stateless7/exp_giga_finetune/epoch-5.pt 2023-05-16 03:53:00,874 INFO [finetune.py:992] (0/2) Epoch 6, batch 0, loss[loss=0.2163, simple_loss=0.3021, pruned_loss=0.06522, over 12088.00 frames. ], tot_loss[loss=0.2163, simple_loss=0.3021, pruned_loss=0.06522, over 12088.00 frames. ], batch size: 32, lr: 4.62e-03, grad_scale: 16.0 2023-05-16 03:53:00,875 INFO [finetune.py:1017] (0/2) Computing validation loss 2023-05-16 03:53:08,569 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.6573, 1.9185, 3.2057, 3.6049, 3.4214, 3.5337, 3.3056, 2.4406], device='cuda:0'), covar=tensor([0.0043, 0.0431, 0.0136, 0.0039, 0.0079, 0.0080, 0.0115, 0.0384], device='cuda:0'), in_proj_covar=tensor([0.0076, 0.0115, 0.0095, 0.0070, 0.0090, 0.0105, 0.0083, 0.0129], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 03:53:18,207 INFO [finetune.py:1026] (0/2) Epoch 6, validation: loss=0.2892, simple_loss=0.365, pruned_loss=0.1067, over 1020973.00 frames. 2023-05-16 03:53:18,208 INFO [finetune.py:1027] (0/2) Maximum memory allocated so far is 12745MB 2023-05-16 03:53:40,730 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.2156, 4.2709, 4.1643, 4.5481, 3.2680, 4.0256, 2.4785, 4.2393], device='cuda:0'), covar=tensor([0.1891, 0.0699, 0.0971, 0.0673, 0.1058, 0.0663, 0.2210, 0.1448], device='cuda:0'), in_proj_covar=tensor([0.0218, 0.0251, 0.0281, 0.0329, 0.0225, 0.0229, 0.0245, 0.0359], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 03:53:46,704 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.42 vs. limit=2.0 2023-05-16 03:53:52,194 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=161216.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 03:53:54,142 INFO [finetune.py:992] (0/2) Epoch 6, batch 50, loss[loss=0.1741, simple_loss=0.2707, pruned_loss=0.03871, over 12142.00 frames. ], tot_loss[loss=0.1882, simple_loss=0.2745, pruned_loss=0.05097, over 532022.98 frames. ], batch size: 34, lr: 4.61e-03, grad_scale: 16.0 2023-05-16 03:54:02,584 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=161231.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 03:54:06,010 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.197e+02 3.170e+02 3.764e+02 4.622e+02 1.120e+03, threshold=7.528e+02, percent-clipped=1.0 2023-05-16 03:54:16,972 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.5813, 3.1235, 5.0074, 2.4824, 2.6851, 3.8379, 3.0498, 3.9122], device='cuda:0'), covar=tensor([0.0469, 0.1364, 0.0315, 0.1365, 0.2037, 0.1371, 0.1589, 0.1115], device='cuda:0'), in_proj_covar=tensor([0.0218, 0.0218, 0.0213, 0.0173, 0.0222, 0.0263, 0.0211, 0.0232], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:0') 2023-05-16 03:54:21,709 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=161257.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 03:54:30,230 INFO [finetune.py:992] (0/2) Epoch 6, batch 100, loss[loss=0.1825, simple_loss=0.2611, pruned_loss=0.05191, over 12350.00 frames. ], tot_loss[loss=0.1861, simple_loss=0.2738, pruned_loss=0.04914, over 942253.23 frames. ], batch size: 30, lr: 4.61e-03, grad_scale: 16.0 2023-05-16 03:54:36,724 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=161277.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 03:54:41,669 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.1611, 5.0911, 4.8761, 5.0013, 4.7522, 5.0759, 5.0431, 5.3883], device='cuda:0'), covar=tensor([0.0270, 0.0130, 0.0221, 0.0321, 0.0680, 0.0315, 0.0163, 0.0163], device='cuda:0'), in_proj_covar=tensor([0.0162, 0.0164, 0.0163, 0.0204, 0.0206, 0.0179, 0.0148, 0.0195], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:0') 2023-05-16 03:54:47,480 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=161292.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 03:54:59,154 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.97 vs. limit=2.0 2023-05-16 03:55:05,928 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.0561, 4.9670, 4.7664, 4.8533, 4.5878, 4.9501, 4.9653, 5.2494], device='cuda:0'), covar=tensor([0.0174, 0.0129, 0.0206, 0.0312, 0.0723, 0.0227, 0.0141, 0.0155], device='cuda:0'), in_proj_covar=tensor([0.0163, 0.0166, 0.0165, 0.0207, 0.0209, 0.0181, 0.0149, 0.0197], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:0') 2023-05-16 03:55:06,490 INFO [finetune.py:992] (0/2) Epoch 6, batch 150, loss[loss=0.1805, simple_loss=0.2726, pruned_loss=0.04418, over 12292.00 frames. ], tot_loss[loss=0.1829, simple_loss=0.2708, pruned_loss=0.04747, over 1269973.86 frames. ], batch size: 34, lr: 4.61e-03, grad_scale: 16.0 2023-05-16 03:55:08,011 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=161321.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 03:55:08,849 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.5491, 2.5633, 3.5678, 4.4986, 3.8613, 4.4960, 3.7833, 3.2423], device='cuda:0'), covar=tensor([0.0030, 0.0372, 0.0126, 0.0030, 0.0118, 0.0053, 0.0121, 0.0304], device='cuda:0'), in_proj_covar=tensor([0.0078, 0.0118, 0.0097, 0.0072, 0.0093, 0.0109, 0.0086, 0.0133], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 03:55:18,797 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.883e+02 2.699e+02 3.068e+02 3.800e+02 6.191e+02, threshold=6.136e+02, percent-clipped=0.0 2023-05-16 03:55:39,914 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=4.62 vs. limit=5.0 2023-05-16 03:55:42,305 INFO [finetune.py:992] (0/2) Epoch 6, batch 200, loss[loss=0.1893, simple_loss=0.2813, pruned_loss=0.04869, over 11671.00 frames. ], tot_loss[loss=0.182, simple_loss=0.2704, pruned_loss=0.04676, over 1519377.13 frames. ], batch size: 48, lr: 4.61e-03, grad_scale: 16.0 2023-05-16 03:56:05,469 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=161401.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 03:56:18,786 INFO [finetune.py:992] (0/2) Epoch 6, batch 250, loss[loss=0.1924, simple_loss=0.2749, pruned_loss=0.05497, over 12123.00 frames. ], tot_loss[loss=0.1814, simple_loss=0.2703, pruned_loss=0.04626, over 1719537.09 frames. ], batch size: 39, lr: 4.61e-03, grad_scale: 16.0 2023-05-16 03:56:30,381 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.9977, 2.2729, 3.4744, 3.1122, 3.4327, 3.1128, 2.3445, 3.4946], device='cuda:0'), covar=tensor([0.0118, 0.0338, 0.0163, 0.0194, 0.0124, 0.0167, 0.0367, 0.0108], device='cuda:0'), in_proj_covar=tensor([0.0157, 0.0187, 0.0164, 0.0165, 0.0189, 0.0145, 0.0180, 0.0158], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 03:56:32,101 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.147e+02 3.211e+02 3.699e+02 4.349e+02 1.199e+03, threshold=7.397e+02, percent-clipped=9.0 2023-05-16 03:56:48,263 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=161460.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 03:56:49,724 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=161462.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 03:56:54,328 INFO [finetune.py:992] (0/2) Epoch 6, batch 300, loss[loss=0.1879, simple_loss=0.2808, pruned_loss=0.04755, over 12156.00 frames. ], tot_loss[loss=0.1801, simple_loss=0.2688, pruned_loss=0.04573, over 1872040.68 frames. ], batch size: 34, lr: 4.61e-03, grad_scale: 8.0 2023-05-16 03:57:18,000 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.4605, 3.5766, 3.3400, 3.7653, 3.5731, 2.6171, 3.3721, 2.8397], device='cuda:0'), covar=tensor([0.0746, 0.1047, 0.1374, 0.0590, 0.1159, 0.1477, 0.0999, 0.2797], device='cuda:0'), in_proj_covar=tensor([0.0290, 0.0358, 0.0334, 0.0252, 0.0344, 0.0257, 0.0322, 0.0351], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 03:57:29,865 INFO [finetune.py:992] (0/2) Epoch 6, batch 350, loss[loss=0.179, simple_loss=0.2667, pruned_loss=0.04565, over 12044.00 frames. ], tot_loss[loss=0.1805, simple_loss=0.269, pruned_loss=0.04595, over 1981462.96 frames. ], batch size: 37, lr: 4.61e-03, grad_scale: 8.0 2023-05-16 03:57:31,491 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=161521.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 03:57:42,286 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.107e+02 2.879e+02 3.397e+02 4.006e+02 7.575e+02, threshold=6.795e+02, percent-clipped=1.0 2023-05-16 03:57:57,088 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=161557.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 03:58:05,462 INFO [finetune.py:992] (0/2) Epoch 6, batch 400, loss[loss=0.2423, simple_loss=0.3128, pruned_loss=0.08596, over 8329.00 frames. ], tot_loss[loss=0.1803, simple_loss=0.2686, pruned_loss=0.04594, over 2063972.36 frames. ], batch size: 98, lr: 4.61e-03, grad_scale: 8.0 2023-05-16 03:58:08,307 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=161572.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 03:58:18,888 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=161587.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 03:58:31,851 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=161605.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 03:58:38,919 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.3421, 4.8181, 5.3000, 4.5636, 4.9663, 4.7074, 5.3666, 5.1350], device='cuda:0'), covar=tensor([0.0218, 0.0387, 0.0256, 0.0274, 0.0358, 0.0269, 0.0195, 0.0216], device='cuda:0'), in_proj_covar=tensor([0.0223, 0.0232, 0.0247, 0.0227, 0.0223, 0.0224, 0.0204, 0.0190], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 03:58:40,553 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 2023-05-16 03:58:41,628 INFO [finetune.py:992] (0/2) Epoch 6, batch 450, loss[loss=0.1858, simple_loss=0.2725, pruned_loss=0.04955, over 10437.00 frames. ], tot_loss[loss=0.18, simple_loss=0.2683, pruned_loss=0.04584, over 2136441.99 frames. ], batch size: 68, lr: 4.61e-03, grad_scale: 8.0 2023-05-16 03:58:43,098 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=161621.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 03:58:50,913 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.6921, 3.6732, 3.3892, 3.2237, 2.9727, 2.8948, 3.5405, 2.5372], device='cuda:0'), covar=tensor([0.0314, 0.0106, 0.0152, 0.0172, 0.0341, 0.0291, 0.0132, 0.0351], device='cuda:0'), in_proj_covar=tensor([0.0182, 0.0150, 0.0146, 0.0174, 0.0193, 0.0185, 0.0153, 0.0190], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], device='cuda:0') 2023-05-16 03:58:54,247 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.271e+02 2.709e+02 3.206e+02 3.955e+02 8.153e+02, threshold=6.411e+02, percent-clipped=2.0 2023-05-16 03:59:13,524 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.9138, 5.9381, 5.7055, 5.2281, 5.0434, 5.8168, 5.4350, 5.1524], device='cuda:0'), covar=tensor([0.0738, 0.0776, 0.0626, 0.1700, 0.0677, 0.0758, 0.1537, 0.1110], device='cuda:0'), in_proj_covar=tensor([0.0553, 0.0490, 0.0463, 0.0575, 0.0372, 0.0644, 0.0695, 0.0515], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-05-16 03:59:17,020 INFO [finetune.py:992] (0/2) Epoch 6, batch 500, loss[loss=0.1868, simple_loss=0.2775, pruned_loss=0.04807, over 12108.00 frames. ], tot_loss[loss=0.1791, simple_loss=0.2673, pruned_loss=0.04543, over 2190172.79 frames. ], batch size: 38, lr: 4.61e-03, grad_scale: 8.0 2023-05-16 03:59:17,091 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=161669.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 03:59:32,720 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.36 vs. limit=2.0 2023-05-16 03:59:33,123 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.3012, 4.8776, 5.2619, 4.5753, 4.9595, 4.6671, 5.3215, 4.9944], device='cuda:0'), covar=tensor([0.0261, 0.0377, 0.0288, 0.0285, 0.0373, 0.0280, 0.0228, 0.0325], device='cuda:0'), in_proj_covar=tensor([0.0224, 0.0233, 0.0248, 0.0228, 0.0224, 0.0225, 0.0205, 0.0192], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 03:59:54,167 INFO [finetune.py:992] (0/2) Epoch 6, batch 550, loss[loss=0.1514, simple_loss=0.2304, pruned_loss=0.03617, over 12166.00 frames. ], tot_loss[loss=0.1782, simple_loss=0.2661, pruned_loss=0.04518, over 2225804.72 frames. ], batch size: 29, lr: 4.61e-03, grad_scale: 8.0 2023-05-16 03:59:54,348 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.7164, 4.3421, 4.5878, 4.6305, 4.3419, 4.6773, 4.4976, 2.3702], device='cuda:0'), covar=tensor([0.0089, 0.0078, 0.0082, 0.0062, 0.0058, 0.0085, 0.0083, 0.0800], device='cuda:0'), in_proj_covar=tensor([0.0061, 0.0070, 0.0073, 0.0067, 0.0056, 0.0084, 0.0072, 0.0089], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-16 03:59:59,951 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.4212, 5.2144, 5.2959, 5.3767, 5.0151, 5.0200, 4.8186, 5.2785], device='cuda:0'), covar=tensor([0.0634, 0.0569, 0.0714, 0.0543, 0.1836, 0.1244, 0.0563, 0.1004], device='cuda:0'), in_proj_covar=tensor([0.0473, 0.0611, 0.0529, 0.0553, 0.0740, 0.0660, 0.0493, 0.0449], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0005, 0.0004, 0.0003, 0.0003], device='cuda:0') 2023-05-16 04:00:06,815 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.943e+02 2.917e+02 3.411e+02 3.994e+02 7.722e+02, threshold=6.823e+02, percent-clipped=2.0 2023-05-16 04:00:20,669 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=161757.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 04:00:29,161 INFO [finetune.py:992] (0/2) Epoch 6, batch 600, loss[loss=0.1655, simple_loss=0.2449, pruned_loss=0.04305, over 12089.00 frames. ], tot_loss[loss=0.1775, simple_loss=0.2654, pruned_loss=0.04483, over 2260207.91 frames. ], batch size: 32, lr: 4.61e-03, grad_scale: 8.0 2023-05-16 04:00:33,834 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.98 vs. limit=2.0 2023-05-16 04:00:41,599 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2023-05-16 04:00:53,175 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.8469, 3.7657, 3.4641, 3.3485, 3.0132, 3.0007, 3.6728, 2.5516], device='cuda:0'), covar=tensor([0.0286, 0.0109, 0.0161, 0.0169, 0.0357, 0.0298, 0.0126, 0.0384], device='cuda:0'), in_proj_covar=tensor([0.0183, 0.0151, 0.0147, 0.0175, 0.0195, 0.0186, 0.0155, 0.0192], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], device='cuda:0') 2023-05-16 04:00:58,002 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.0577, 4.8286, 5.0659, 5.1534, 4.7835, 5.1024, 4.8777, 2.8406], device='cuda:0'), covar=tensor([0.0096, 0.0067, 0.0062, 0.0051, 0.0049, 0.0079, 0.0095, 0.0624], device='cuda:0'), in_proj_covar=tensor([0.0061, 0.0070, 0.0073, 0.0067, 0.0056, 0.0084, 0.0072, 0.0089], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-16 04:01:02,208 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=161815.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 04:01:02,842 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=161816.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 04:01:03,105 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-05-16 04:01:04,945 INFO [finetune.py:992] (0/2) Epoch 6, batch 650, loss[loss=0.1725, simple_loss=0.2645, pruned_loss=0.04021, over 12348.00 frames. ], tot_loss[loss=0.1773, simple_loss=0.2651, pruned_loss=0.0447, over 2287402.34 frames. ], batch size: 36, lr: 4.61e-03, grad_scale: 8.0 2023-05-16 04:01:17,791 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.948e+02 2.853e+02 3.256e+02 3.784e+02 6.036e+02, threshold=6.512e+02, percent-clipped=0.0 2023-05-16 04:01:41,725 INFO [finetune.py:992] (0/2) Epoch 6, batch 700, loss[loss=0.1606, simple_loss=0.2494, pruned_loss=0.03592, over 12344.00 frames. ], tot_loss[loss=0.1767, simple_loss=0.2642, pruned_loss=0.04457, over 2305623.74 frames. ], batch size: 31, lr: 4.61e-03, grad_scale: 8.0 2023-05-16 04:01:43,899 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=161872.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 04:01:46,840 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=161876.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 04:01:54,777 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=161887.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 04:02:17,230 INFO [finetune.py:992] (0/2) Epoch 6, batch 750, loss[loss=0.1807, simple_loss=0.2738, pruned_loss=0.04383, over 11581.00 frames. ], tot_loss[loss=0.1764, simple_loss=0.2643, pruned_loss=0.04424, over 2319465.09 frames. ], batch size: 48, lr: 4.61e-03, grad_scale: 8.0 2023-05-16 04:02:18,019 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=161920.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 04:02:21,722 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.5047, 3.1905, 4.8310, 2.5572, 2.6211, 3.7196, 3.1119, 3.7862], device='cuda:0'), covar=tensor([0.0472, 0.1219, 0.0381, 0.1224, 0.2090, 0.1391, 0.1447, 0.1141], device='cuda:0'), in_proj_covar=tensor([0.0229, 0.0227, 0.0227, 0.0179, 0.0232, 0.0276, 0.0222, 0.0244], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:0') 2023-05-16 04:02:28,396 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=161935.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 04:02:29,707 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.059e+02 2.912e+02 3.415e+02 4.088e+02 6.399e+02, threshold=6.830e+02, percent-clipped=0.0 2023-05-16 04:02:52,584 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.84 vs. limit=2.0 2023-05-16 04:02:52,643 INFO [finetune.py:992] (0/2) Epoch 6, batch 800, loss[loss=0.1658, simple_loss=0.2554, pruned_loss=0.03805, over 12265.00 frames. ], tot_loss[loss=0.1762, simple_loss=0.264, pruned_loss=0.04415, over 2317387.54 frames. ], batch size: 32, lr: 4.61e-03, grad_scale: 8.0 2023-05-16 04:03:16,166 INFO [checkpoint.py:75] (0/2) Saving checkpoint to pruned_transducer_stateless7/exp_giga_finetune/checkpoint-62000.pt 2023-05-16 04:03:32,654 INFO [finetune.py:992] (0/2) Epoch 6, batch 850, loss[loss=0.2345, simple_loss=0.3026, pruned_loss=0.08318, over 8272.00 frames. ], tot_loss[loss=0.1759, simple_loss=0.2638, pruned_loss=0.04402, over 2325110.05 frames. ], batch size: 97, lr: 4.61e-03, grad_scale: 8.0 2023-05-16 04:03:37,775 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.3678, 4.9536, 5.3510, 4.7352, 5.0931, 4.7998, 5.3853, 4.9879], device='cuda:0'), covar=tensor([0.0240, 0.0347, 0.0274, 0.0247, 0.0304, 0.0267, 0.0208, 0.0313], device='cuda:0'), in_proj_covar=tensor([0.0231, 0.0240, 0.0255, 0.0236, 0.0230, 0.0232, 0.0211, 0.0198], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 04:03:45,567 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.961e+02 2.955e+02 3.400e+02 4.002e+02 7.780e+02, threshold=6.800e+02, percent-clipped=2.0 2023-05-16 04:03:59,807 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=162057.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 04:04:08,149 INFO [finetune.py:992] (0/2) Epoch 6, batch 900, loss[loss=0.1921, simple_loss=0.2793, pruned_loss=0.05244, over 12103.00 frames. ], tot_loss[loss=0.1764, simple_loss=0.2644, pruned_loss=0.04423, over 2341317.85 frames. ], batch size: 32, lr: 4.61e-03, grad_scale: 8.0 2023-05-16 04:04:14,188 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=162077.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 04:04:22,005 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.2886, 6.0782, 5.7325, 5.6231, 6.1721, 5.4310, 5.6511, 5.7529], device='cuda:0'), covar=tensor([0.1474, 0.0763, 0.0963, 0.1872, 0.0795, 0.2281, 0.1569, 0.0920], device='cuda:0'), in_proj_covar=tensor([0.0341, 0.0454, 0.0362, 0.0413, 0.0437, 0.0418, 0.0371, 0.0356], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 04:04:28,756 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-05-16 04:04:33,885 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=162105.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 04:04:35,361 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([6.1524, 6.0955, 5.8440, 5.3475, 5.2126, 5.9817, 5.6171, 5.3606], device='cuda:0'), covar=tensor([0.0695, 0.0827, 0.0640, 0.1767, 0.0619, 0.0774, 0.1738, 0.1068], device='cuda:0'), in_proj_covar=tensor([0.0559, 0.0498, 0.0470, 0.0584, 0.0377, 0.0654, 0.0710, 0.0524], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-05-16 04:04:41,634 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=162116.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 04:04:43,601 INFO [finetune.py:992] (0/2) Epoch 6, batch 950, loss[loss=0.156, simple_loss=0.2455, pruned_loss=0.03324, over 12284.00 frames. ], tot_loss[loss=0.1763, simple_loss=0.2646, pruned_loss=0.04403, over 2351295.22 frames. ], batch size: 33, lr: 4.61e-03, grad_scale: 8.0 2023-05-16 04:04:45,315 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.1539, 2.7013, 3.9173, 3.3216, 3.6927, 3.4092, 2.7578, 3.7463], device='cuda:0'), covar=tensor([0.0137, 0.0289, 0.0117, 0.0202, 0.0114, 0.0149, 0.0307, 0.0091], device='cuda:0'), in_proj_covar=tensor([0.0163, 0.0193, 0.0172, 0.0172, 0.0197, 0.0150, 0.0185, 0.0165], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 04:04:56,378 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.005e+02 2.818e+02 3.634e+02 4.529e+02 2.030e+03, threshold=7.268e+02, percent-clipped=4.0 2023-05-16 04:04:58,018 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=162138.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 04:05:17,090 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=162164.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 04:05:20,603 INFO [finetune.py:992] (0/2) Epoch 6, batch 1000, loss[loss=0.1711, simple_loss=0.2614, pruned_loss=0.04041, over 12079.00 frames. ], tot_loss[loss=0.1756, simple_loss=0.2639, pruned_loss=0.04365, over 2366008.27 frames. ], batch size: 32, lr: 4.60e-03, grad_scale: 8.0 2023-05-16 04:05:22,129 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=162171.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 04:05:56,046 INFO [finetune.py:992] (0/2) Epoch 6, batch 1050, loss[loss=0.1656, simple_loss=0.2609, pruned_loss=0.03517, over 12154.00 frames. ], tot_loss[loss=0.1749, simple_loss=0.2632, pruned_loss=0.0433, over 2373599.86 frames. ], batch size: 34, lr: 4.60e-03, grad_scale: 8.0 2023-05-16 04:06:04,629 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=162231.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 04:06:08,881 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.037e+02 2.802e+02 3.110e+02 3.560e+02 6.634e+02, threshold=6.220e+02, percent-clipped=0.0 2023-05-16 04:06:31,902 INFO [finetune.py:992] (0/2) Epoch 6, batch 1100, loss[loss=0.1447, simple_loss=0.2257, pruned_loss=0.03185, over 12259.00 frames. ], tot_loss[loss=0.1745, simple_loss=0.2626, pruned_loss=0.04325, over 2372306.07 frames. ], batch size: 28, lr: 4.60e-03, grad_scale: 8.0 2023-05-16 04:06:49,890 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=162292.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 04:07:08,834 INFO [finetune.py:992] (0/2) Epoch 6, batch 1150, loss[loss=0.1634, simple_loss=0.2462, pruned_loss=0.04032, over 12289.00 frames. ], tot_loss[loss=0.1743, simple_loss=0.2622, pruned_loss=0.04323, over 2368596.22 frames. ], batch size: 28, lr: 4.60e-03, grad_scale: 8.0 2023-05-16 04:07:22,449 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.089e+02 3.128e+02 3.626e+02 4.185e+02 7.642e+02, threshold=7.251e+02, percent-clipped=5.0 2023-05-16 04:07:36,968 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.55 vs. limit=2.0 2023-05-16 04:07:44,452 INFO [finetune.py:992] (0/2) Epoch 6, batch 1200, loss[loss=0.1628, simple_loss=0.2415, pruned_loss=0.04207, over 12276.00 frames. ], tot_loss[loss=0.1737, simple_loss=0.2615, pruned_loss=0.04295, over 2374155.28 frames. ], batch size: 28, lr: 4.60e-03, grad_scale: 8.0 2023-05-16 04:07:52,468 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.5237, 2.7785, 3.5162, 4.4854, 3.7805, 4.3095, 3.6550, 3.1975], device='cuda:0'), covar=tensor([0.0030, 0.0338, 0.0147, 0.0030, 0.0130, 0.0077, 0.0129, 0.0306], device='cuda:0'), in_proj_covar=tensor([0.0081, 0.0122, 0.0102, 0.0075, 0.0097, 0.0113, 0.0091, 0.0137], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 04:08:09,777 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=2.87 vs. limit=5.0 2023-05-16 04:08:20,141 INFO [finetune.py:992] (0/2) Epoch 6, batch 1250, loss[loss=0.1471, simple_loss=0.2355, pruned_loss=0.02934, over 12033.00 frames. ], tot_loss[loss=0.174, simple_loss=0.2617, pruned_loss=0.04318, over 2381087.24 frames. ], batch size: 31, lr: 4.60e-03, grad_scale: 8.0 2023-05-16 04:08:30,613 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=162433.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 04:08:34,121 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.032e+02 2.846e+02 3.349e+02 4.161e+02 7.273e+02, threshold=6.698e+02, percent-clipped=1.0 2023-05-16 04:08:37,145 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.1305, 4.7437, 5.1242, 4.4563, 4.8342, 4.5183, 5.1326, 4.7842], device='cuda:0'), covar=tensor([0.0248, 0.0339, 0.0250, 0.0258, 0.0302, 0.0321, 0.0203, 0.0349], device='cuda:0'), in_proj_covar=tensor([0.0230, 0.0241, 0.0255, 0.0236, 0.0231, 0.0233, 0.0210, 0.0198], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 04:08:51,097 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.2791, 4.2113, 4.2565, 4.5104, 3.2488, 4.0361, 2.6941, 4.2275], device='cuda:0'), covar=tensor([0.1622, 0.0688, 0.0856, 0.0644, 0.1004, 0.0591, 0.1765, 0.1309], device='cuda:0'), in_proj_covar=tensor([0.0227, 0.0262, 0.0293, 0.0349, 0.0233, 0.0237, 0.0254, 0.0379], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 04:08:56,529 INFO [finetune.py:992] (0/2) Epoch 6, batch 1300, loss[loss=0.1807, simple_loss=0.2707, pruned_loss=0.04528, over 12283.00 frames. ], tot_loss[loss=0.1741, simple_loss=0.2618, pruned_loss=0.04318, over 2379574.00 frames. ], batch size: 37, lr: 4.60e-03, grad_scale: 8.0 2023-05-16 04:08:58,119 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=162471.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 04:09:18,904 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.8435, 4.7780, 4.6715, 4.7687, 4.3697, 4.9241, 4.8302, 5.1336], device='cuda:0'), covar=tensor([0.0321, 0.0164, 0.0229, 0.0309, 0.0834, 0.0246, 0.0148, 0.0171], device='cuda:0'), in_proj_covar=tensor([0.0180, 0.0184, 0.0182, 0.0229, 0.0230, 0.0201, 0.0164, 0.0216], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0005, 0.0004, 0.0004, 0.0003, 0.0005], device='cuda:0') 2023-05-16 04:09:26,791 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=162511.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 04:09:27,129 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=4.93 vs. limit=5.0 2023-05-16 04:09:32,393 INFO [finetune.py:992] (0/2) Epoch 6, batch 1350, loss[loss=0.1654, simple_loss=0.2479, pruned_loss=0.0415, over 11995.00 frames. ], tot_loss[loss=0.1733, simple_loss=0.261, pruned_loss=0.04276, over 2370780.47 frames. ], batch size: 28, lr: 4.60e-03, grad_scale: 8.0 2023-05-16 04:09:32,461 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=162519.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 04:09:46,044 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.979e+02 2.606e+02 2.970e+02 3.970e+02 1.108e+03, threshold=5.939e+02, percent-clipped=2.0 2023-05-16 04:10:08,428 INFO [finetune.py:992] (0/2) Epoch 6, batch 1400, loss[loss=0.1465, simple_loss=0.2359, pruned_loss=0.02855, over 12358.00 frames. ], tot_loss[loss=0.1733, simple_loss=0.2613, pruned_loss=0.04269, over 2374633.28 frames. ], batch size: 30, lr: 4.60e-03, grad_scale: 8.0 2023-05-16 04:10:10,788 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=162572.0, num_to_drop=1, layers_to_drop={1} 2023-05-16 04:10:22,055 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=162587.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 04:10:45,564 INFO [finetune.py:992] (0/2) Epoch 6, batch 1450, loss[loss=0.2297, simple_loss=0.2994, pruned_loss=0.08002, over 8069.00 frames. ], tot_loss[loss=0.1727, simple_loss=0.2606, pruned_loss=0.04237, over 2377402.17 frames. ], batch size: 97, lr: 4.60e-03, grad_scale: 8.0 2023-05-16 04:10:59,151 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.869e+02 2.825e+02 3.375e+02 3.791e+02 8.443e+02, threshold=6.750e+02, percent-clipped=3.0 2023-05-16 04:11:21,162 INFO [finetune.py:992] (0/2) Epoch 6, batch 1500, loss[loss=0.1527, simple_loss=0.2297, pruned_loss=0.03791, over 12254.00 frames. ], tot_loss[loss=0.1733, simple_loss=0.2614, pruned_loss=0.04259, over 2374431.84 frames. ], batch size: 28, lr: 4.60e-03, grad_scale: 8.0 2023-05-16 04:11:55,833 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.1325, 2.4433, 3.6392, 3.0369, 3.4283, 3.1901, 2.4267, 3.5334], device='cuda:0'), covar=tensor([0.0099, 0.0295, 0.0128, 0.0204, 0.0135, 0.0157, 0.0309, 0.0115], device='cuda:0'), in_proj_covar=tensor([0.0162, 0.0193, 0.0172, 0.0173, 0.0197, 0.0150, 0.0185, 0.0165], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 04:11:56,327 INFO [finetune.py:992] (0/2) Epoch 6, batch 1550, loss[loss=0.1688, simple_loss=0.2592, pruned_loss=0.03917, over 12375.00 frames. ], tot_loss[loss=0.1738, simple_loss=0.2622, pruned_loss=0.04277, over 2372791.74 frames. ], batch size: 38, lr: 4.60e-03, grad_scale: 8.0 2023-05-16 04:12:07,102 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=162733.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 04:12:10,540 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.849e+02 2.853e+02 3.212e+02 3.842e+02 6.930e+02, threshold=6.425e+02, percent-clipped=1.0 2023-05-16 04:12:33,554 INFO [finetune.py:992] (0/2) Epoch 6, batch 1600, loss[loss=0.1638, simple_loss=0.2541, pruned_loss=0.03674, over 12186.00 frames. ], tot_loss[loss=0.1746, simple_loss=0.2633, pruned_loss=0.04299, over 2381271.27 frames. ], batch size: 35, lr: 4.60e-03, grad_scale: 8.0 2023-05-16 04:12:41,595 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.3159, 3.1757, 4.6649, 2.4651, 2.6936, 3.5585, 3.1025, 3.7319], device='cuda:0'), covar=tensor([0.0476, 0.1082, 0.0334, 0.1148, 0.1740, 0.1215, 0.1229, 0.0988], device='cuda:0'), in_proj_covar=tensor([0.0228, 0.0226, 0.0227, 0.0178, 0.0231, 0.0277, 0.0220, 0.0246], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:0') 2023-05-16 04:12:42,046 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=162781.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 04:13:09,126 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.4611, 4.9430, 4.2643, 5.1142, 4.5651, 3.0785, 4.3300, 3.2169], device='cuda:0'), covar=tensor([0.0797, 0.0708, 0.1334, 0.0424, 0.1150, 0.1576, 0.1085, 0.3120], device='cuda:0'), in_proj_covar=tensor([0.0291, 0.0364, 0.0339, 0.0262, 0.0348, 0.0260, 0.0327, 0.0354], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 04:13:09,569 INFO [finetune.py:992] (0/2) Epoch 6, batch 1650, loss[loss=0.1387, simple_loss=0.2188, pruned_loss=0.02928, over 12286.00 frames. ], tot_loss[loss=0.1741, simple_loss=0.2623, pruned_loss=0.04297, over 2383926.49 frames. ], batch size: 28, lr: 4.60e-03, grad_scale: 8.0 2023-05-16 04:13:15,390 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=162827.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 04:13:22,620 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.5642, 2.7045, 4.4184, 4.7018, 2.9602, 2.6627, 2.8115, 2.1421], device='cuda:0'), covar=tensor([0.1502, 0.3152, 0.0485, 0.0361, 0.1176, 0.2078, 0.2594, 0.3830], device='cuda:0'), in_proj_covar=tensor([0.0275, 0.0373, 0.0261, 0.0287, 0.0254, 0.0286, 0.0355, 0.0358], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-16 04:13:23,019 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.944e+02 2.994e+02 3.423e+02 4.213e+02 1.099e+03, threshold=6.846e+02, percent-clipped=1.0 2023-05-16 04:13:35,940 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.38 vs. limit=2.0 2023-05-16 04:13:44,196 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=162867.0, num_to_drop=1, layers_to_drop={1} 2023-05-16 04:13:45,498 INFO [finetune.py:992] (0/2) Epoch 6, batch 1700, loss[loss=0.1569, simple_loss=0.2477, pruned_loss=0.03304, over 12123.00 frames. ], tot_loss[loss=0.1734, simple_loss=0.2613, pruned_loss=0.04269, over 2392760.42 frames. ], batch size: 33, lr: 4.60e-03, grad_scale: 8.0 2023-05-16 04:13:59,072 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=162887.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 04:13:59,909 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=162888.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 04:14:20,324 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.83 vs. limit=2.0 2023-05-16 04:14:22,116 INFO [finetune.py:992] (0/2) Epoch 6, batch 1750, loss[loss=0.1913, simple_loss=0.2763, pruned_loss=0.05312, over 12127.00 frames. ], tot_loss[loss=0.1733, simple_loss=0.261, pruned_loss=0.04277, over 2390846.05 frames. ], batch size: 38, lr: 4.60e-03, grad_scale: 8.0 2023-05-16 04:14:33,499 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=162935.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 04:14:35,534 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.888e+02 2.908e+02 3.413e+02 4.052e+02 1.694e+03, threshold=6.827e+02, percent-clipped=2.0 2023-05-16 04:14:44,851 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.5990, 4.5525, 4.4559, 4.0629, 4.2529, 4.5515, 4.2084, 4.0892], device='cuda:0'), covar=tensor([0.0881, 0.1000, 0.0734, 0.1491, 0.1915, 0.0940, 0.1664, 0.1141], device='cuda:0'), in_proj_covar=tensor([0.0572, 0.0512, 0.0479, 0.0593, 0.0386, 0.0672, 0.0727, 0.0531], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0004, 0.0004, 0.0003], device='cuda:0') 2023-05-16 04:14:49,365 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=4.80 vs. limit=5.0 2023-05-16 04:14:57,146 INFO [finetune.py:992] (0/2) Epoch 6, batch 1800, loss[loss=0.1887, simple_loss=0.2751, pruned_loss=0.0512, over 10697.00 frames. ], tot_loss[loss=0.1744, simple_loss=0.262, pruned_loss=0.04342, over 2374765.59 frames. ], batch size: 68, lr: 4.60e-03, grad_scale: 8.0 2023-05-16 04:15:02,170 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.4063, 4.9073, 3.0083, 2.8614, 4.2191, 2.5769, 4.1081, 3.3133], device='cuda:0'), covar=tensor([0.0629, 0.0483, 0.1106, 0.1322, 0.0211, 0.1269, 0.0463, 0.0788], device='cuda:0'), in_proj_covar=tensor([0.0182, 0.0244, 0.0172, 0.0193, 0.0133, 0.0178, 0.0189, 0.0170], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 04:15:32,559 INFO [finetune.py:992] (0/2) Epoch 6, batch 1850, loss[loss=0.1702, simple_loss=0.2508, pruned_loss=0.04473, over 12023.00 frames. ], tot_loss[loss=0.1755, simple_loss=0.263, pruned_loss=0.04399, over 2380266.72 frames. ], batch size: 31, lr: 4.60e-03, grad_scale: 8.0 2023-05-16 04:15:36,487 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.8373, 2.9714, 4.7453, 5.0291, 3.1890, 2.8217, 3.0341, 2.2807], device='cuda:0'), covar=tensor([0.1403, 0.2739, 0.0380, 0.0312, 0.1074, 0.2016, 0.2559, 0.3727], device='cuda:0'), in_proj_covar=tensor([0.0273, 0.0370, 0.0259, 0.0285, 0.0253, 0.0283, 0.0353, 0.0357], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-16 04:15:39,343 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.1082, 4.4826, 3.8625, 4.7678, 4.3670, 2.7221, 4.0612, 3.0634], device='cuda:0'), covar=tensor([0.0862, 0.0748, 0.1499, 0.0512, 0.1161, 0.1672, 0.1055, 0.3007], device='cuda:0'), in_proj_covar=tensor([0.0291, 0.0365, 0.0340, 0.0263, 0.0350, 0.0260, 0.0327, 0.0354], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 04:15:46,652 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.915e+02 2.870e+02 3.284e+02 4.212e+02 1.107e+03, threshold=6.567e+02, percent-clipped=4.0 2023-05-16 04:15:59,065 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.6225, 4.8787, 4.2010, 5.0877, 4.7882, 2.9134, 4.2995, 3.3052], device='cuda:0'), covar=tensor([0.0557, 0.0621, 0.1191, 0.0439, 0.0746, 0.1511, 0.0856, 0.2815], device='cuda:0'), in_proj_covar=tensor([0.0291, 0.0366, 0.0341, 0.0264, 0.0351, 0.0260, 0.0328, 0.0355], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 04:16:04,124 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 2023-05-16 04:16:09,553 INFO [finetune.py:992] (0/2) Epoch 6, batch 1900, loss[loss=0.1604, simple_loss=0.243, pruned_loss=0.03896, over 11909.00 frames. ], tot_loss[loss=0.1747, simple_loss=0.2619, pruned_loss=0.04376, over 2377797.81 frames. ], batch size: 26, lr: 4.60e-03, grad_scale: 8.0 2023-05-16 04:16:24,741 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=163090.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 04:16:41,081 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.3487, 4.5174, 4.1806, 4.5313, 3.1318, 4.1668, 2.9596, 4.3007], device='cuda:0'), covar=tensor([0.1497, 0.0474, 0.0852, 0.0598, 0.1003, 0.0528, 0.1428, 0.0974], device='cuda:0'), in_proj_covar=tensor([0.0226, 0.0259, 0.0292, 0.0347, 0.0232, 0.0236, 0.0253, 0.0377], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 04:16:45,144 INFO [finetune.py:992] (0/2) Epoch 6, batch 1950, loss[loss=0.1605, simple_loss=0.2513, pruned_loss=0.0348, over 12183.00 frames. ], tot_loss[loss=0.1746, simple_loss=0.2618, pruned_loss=0.04366, over 2376835.63 frames. ], batch size: 31, lr: 4.60e-03, grad_scale: 8.0 2023-05-16 04:16:58,927 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.301e+02 2.853e+02 3.305e+02 3.872e+02 8.317e+02, threshold=6.610e+02, percent-clipped=2.0 2023-05-16 04:17:08,551 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=163151.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 04:17:19,927 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=163167.0, num_to_drop=1, layers_to_drop={0} 2023-05-16 04:17:20,714 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.0461, 3.9424, 3.9813, 4.2874, 2.9065, 3.8239, 2.6109, 3.9475], device='cuda:0'), covar=tensor([0.1738, 0.0767, 0.0892, 0.0705, 0.1151, 0.0640, 0.1759, 0.1167], device='cuda:0'), in_proj_covar=tensor([0.0226, 0.0260, 0.0293, 0.0348, 0.0232, 0.0236, 0.0253, 0.0378], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 04:17:21,169 INFO [finetune.py:992] (0/2) Epoch 6, batch 2000, loss[loss=0.2093, simple_loss=0.2982, pruned_loss=0.06025, over 12032.00 frames. ], tot_loss[loss=0.1744, simple_loss=0.262, pruned_loss=0.04341, over 2386431.71 frames. ], batch size: 40, lr: 4.59e-03, grad_scale: 8.0 2023-05-16 04:17:31,881 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=163183.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 04:17:43,990 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.4523, 5.2863, 5.3871, 5.4511, 5.0691, 5.1178, 4.9081, 5.3639], device='cuda:0'), covar=tensor([0.0648, 0.0551, 0.0687, 0.0562, 0.1851, 0.1206, 0.0541, 0.1017], device='cuda:0'), in_proj_covar=tensor([0.0487, 0.0634, 0.0548, 0.0578, 0.0777, 0.0688, 0.0516, 0.0467], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0005, 0.0005, 0.0004, 0.0004], device='cuda:0') 2023-05-16 04:17:55,677 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=163215.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 04:17:58,417 INFO [finetune.py:992] (0/2) Epoch 6, batch 2050, loss[loss=0.1614, simple_loss=0.2476, pruned_loss=0.03758, over 12352.00 frames. ], tot_loss[loss=0.1744, simple_loss=0.2622, pruned_loss=0.04327, over 2386865.69 frames. ], batch size: 31, lr: 4.59e-03, grad_scale: 8.0 2023-05-16 04:18:11,935 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.203e+02 2.798e+02 3.197e+02 3.643e+02 5.178e+02, threshold=6.394e+02, percent-clipped=0.0 2023-05-16 04:18:13,630 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=163240.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 04:18:33,523 INFO [finetune.py:992] (0/2) Epoch 6, batch 2100, loss[loss=0.1796, simple_loss=0.2806, pruned_loss=0.03932, over 12312.00 frames. ], tot_loss[loss=0.175, simple_loss=0.2629, pruned_loss=0.0436, over 2382936.10 frames. ], batch size: 34, lr: 4.59e-03, grad_scale: 8.0 2023-05-16 04:18:41,024 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.79 vs. limit=2.0 2023-05-16 04:18:51,186 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=4.85 vs. limit=5.0 2023-05-16 04:18:56,420 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=163301.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 04:19:08,983 INFO [finetune.py:992] (0/2) Epoch 6, batch 2150, loss[loss=0.1707, simple_loss=0.2688, pruned_loss=0.03635, over 12143.00 frames. ], tot_loss[loss=0.1739, simple_loss=0.2617, pruned_loss=0.04305, over 2384285.90 frames. ], batch size: 34, lr: 4.59e-03, grad_scale: 8.0 2023-05-16 04:19:14,926 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.0317, 2.3712, 3.7086, 2.9999, 3.5264, 3.2497, 2.2450, 3.4440], device='cuda:0'), covar=tensor([0.0150, 0.0421, 0.0113, 0.0258, 0.0150, 0.0162, 0.0452, 0.0142], device='cuda:0'), in_proj_covar=tensor([0.0166, 0.0196, 0.0177, 0.0177, 0.0201, 0.0153, 0.0188, 0.0169], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 04:19:17,942 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 2023-05-16 04:19:23,938 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.121e+02 2.741e+02 3.315e+02 4.107e+02 8.163e+02, threshold=6.630e+02, percent-clipped=2.0 2023-05-16 04:19:26,442 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.2981, 2.7484, 3.7065, 3.2408, 3.5803, 3.4757, 2.7048, 3.7131], device='cuda:0'), covar=tensor([0.0107, 0.0301, 0.0149, 0.0222, 0.0145, 0.0132, 0.0315, 0.0121], device='cuda:0'), in_proj_covar=tensor([0.0166, 0.0196, 0.0177, 0.0177, 0.0201, 0.0153, 0.0188, 0.0169], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 04:19:46,071 INFO [finetune.py:992] (0/2) Epoch 6, batch 2200, loss[loss=0.1525, simple_loss=0.2434, pruned_loss=0.03086, over 12119.00 frames. ], tot_loss[loss=0.1737, simple_loss=0.2615, pruned_loss=0.04297, over 2387915.94 frames. ], batch size: 33, lr: 4.59e-03, grad_scale: 8.0 2023-05-16 04:20:21,679 INFO [finetune.py:992] (0/2) Epoch 6, batch 2250, loss[loss=0.1852, simple_loss=0.2788, pruned_loss=0.04578, over 11809.00 frames. ], tot_loss[loss=0.1736, simple_loss=0.2616, pruned_loss=0.04278, over 2384822.42 frames. ], batch size: 44, lr: 4.59e-03, grad_scale: 8.0 2023-05-16 04:20:35,268 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.566e+02 2.927e+02 3.259e+02 3.882e+02 7.382e+02, threshold=6.518e+02, percent-clipped=2.0 2023-05-16 04:20:41,101 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=163446.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 04:20:42,435 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.97 vs. limit=2.0 2023-05-16 04:20:51,348 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.38 vs. limit=2.0 2023-05-16 04:20:57,246 INFO [finetune.py:992] (0/2) Epoch 6, batch 2300, loss[loss=0.1427, simple_loss=0.2216, pruned_loss=0.03192, over 12022.00 frames. ], tot_loss[loss=0.1729, simple_loss=0.2609, pruned_loss=0.04239, over 2382767.66 frames. ], batch size: 28, lr: 4.59e-03, grad_scale: 8.0 2023-05-16 04:21:08,179 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=163483.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 04:21:33,481 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.89 vs. limit=2.0 2023-05-16 04:21:34,427 INFO [finetune.py:992] (0/2) Epoch 6, batch 2350, loss[loss=0.1904, simple_loss=0.2823, pruned_loss=0.04929, over 12119.00 frames. ], tot_loss[loss=0.1715, simple_loss=0.2596, pruned_loss=0.04168, over 2389190.28 frames. ], batch size: 38, lr: 4.59e-03, grad_scale: 8.0 2023-05-16 04:21:43,123 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=163531.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 04:21:47,866 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.136e+02 2.894e+02 3.393e+02 4.368e+02 1.170e+03, threshold=6.786e+02, percent-clipped=2.0 2023-05-16 04:22:09,901 INFO [finetune.py:992] (0/2) Epoch 6, batch 2400, loss[loss=0.2149, simple_loss=0.2955, pruned_loss=0.06712, over 12030.00 frames. ], tot_loss[loss=0.172, simple_loss=0.2599, pruned_loss=0.04209, over 2381065.78 frames. ], batch size: 42, lr: 4.59e-03, grad_scale: 8.0 2023-05-16 04:22:25,829 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.0632, 2.3641, 2.9985, 3.9145, 2.0374, 4.0511, 4.0324, 4.1806], device='cuda:0'), covar=tensor([0.0149, 0.1135, 0.0503, 0.0146, 0.1406, 0.0219, 0.0196, 0.0080], device='cuda:0'), in_proj_covar=tensor([0.0117, 0.0202, 0.0186, 0.0113, 0.0189, 0.0174, 0.0169, 0.0121], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 04:22:29,245 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=163596.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 04:22:33,016 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.34 vs. limit=2.0 2023-05-16 04:22:46,562 INFO [finetune.py:992] (0/2) Epoch 6, batch 2450, loss[loss=0.1787, simple_loss=0.2726, pruned_loss=0.04239, over 11230.00 frames. ], tot_loss[loss=0.1738, simple_loss=0.2614, pruned_loss=0.04307, over 2364090.71 frames. ], batch size: 55, lr: 4.59e-03, grad_scale: 8.0 2023-05-16 04:22:48,348 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=3.10 vs. limit=5.0 2023-05-16 04:22:49,083 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.33 vs. limit=2.0 2023-05-16 04:23:00,670 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.818e+02 2.696e+02 3.263e+02 3.997e+02 1.017e+03, threshold=6.526e+02, percent-clipped=2.0 2023-05-16 04:23:22,553 INFO [finetune.py:992] (0/2) Epoch 6, batch 2500, loss[loss=0.1927, simple_loss=0.2895, pruned_loss=0.04797, over 11956.00 frames. ], tot_loss[loss=0.1737, simple_loss=0.2609, pruned_loss=0.04327, over 2361436.66 frames. ], batch size: 42, lr: 4.59e-03, grad_scale: 8.0 2023-05-16 04:23:58,062 INFO [finetune.py:992] (0/2) Epoch 6, batch 2550, loss[loss=0.1889, simple_loss=0.2755, pruned_loss=0.05115, over 10487.00 frames. ], tot_loss[loss=0.1734, simple_loss=0.261, pruned_loss=0.04292, over 2368662.70 frames. ], batch size: 68, lr: 4.59e-03, grad_scale: 8.0 2023-05-16 04:24:08,150 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.7630, 2.4558, 3.8031, 4.7648, 4.2126, 4.6250, 3.9359, 3.5488], device='cuda:0'), covar=tensor([0.0027, 0.0419, 0.0113, 0.0030, 0.0084, 0.0061, 0.0108, 0.0263], device='cuda:0'), in_proj_covar=tensor([0.0082, 0.0122, 0.0101, 0.0076, 0.0098, 0.0112, 0.0091, 0.0138], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 04:24:12,124 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.950e+02 2.990e+02 3.429e+02 4.121e+02 9.520e+02, threshold=6.858e+02, percent-clipped=5.0 2023-05-16 04:24:17,297 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=163746.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 04:24:25,096 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.6633, 2.1891, 2.9257, 2.5911, 2.9153, 2.7872, 2.1337, 2.9708], device='cuda:0'), covar=tensor([0.0113, 0.0284, 0.0151, 0.0214, 0.0159, 0.0171, 0.0315, 0.0127], device='cuda:0'), in_proj_covar=tensor([0.0166, 0.0195, 0.0177, 0.0176, 0.0201, 0.0152, 0.0187, 0.0168], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 04:24:33,963 INFO [finetune.py:992] (0/2) Epoch 6, batch 2600, loss[loss=0.1629, simple_loss=0.247, pruned_loss=0.03935, over 12337.00 frames. ], tot_loss[loss=0.1738, simple_loss=0.2618, pruned_loss=0.04291, over 2376290.17 frames. ], batch size: 30, lr: 4.59e-03, grad_scale: 4.0 2023-05-16 04:24:52,474 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=163794.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 04:24:55,470 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([6.1002, 6.0323, 5.9406, 5.3211, 5.2322, 6.0222, 5.6017, 5.3720], device='cuda:0'), covar=tensor([0.0617, 0.0962, 0.0598, 0.1671, 0.0579, 0.0718, 0.1531, 0.1018], device='cuda:0'), in_proj_covar=tensor([0.0574, 0.0511, 0.0476, 0.0595, 0.0388, 0.0672, 0.0731, 0.0536], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0004, 0.0004, 0.0003], device='cuda:0') 2023-05-16 04:25:10,828 INFO [finetune.py:992] (0/2) Epoch 6, batch 2650, loss[loss=0.1879, simple_loss=0.2723, pruned_loss=0.05172, over 12030.00 frames. ], tot_loss[loss=0.1737, simple_loss=0.2619, pruned_loss=0.04274, over 2379220.64 frames. ], batch size: 40, lr: 4.59e-03, grad_scale: 4.0 2023-05-16 04:25:11,801 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.3459, 3.3031, 3.1753, 3.0493, 2.7954, 2.5460, 3.3074, 2.2240], device='cuda:0'), covar=tensor([0.0335, 0.0116, 0.0132, 0.0163, 0.0304, 0.0338, 0.0091, 0.0394], device='cuda:0'), in_proj_covar=tensor([0.0184, 0.0152, 0.0149, 0.0176, 0.0194, 0.0188, 0.0155, 0.0192], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], device='cuda:0') 2023-05-16 04:25:24,956 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.715e+02 2.806e+02 3.186e+02 3.917e+02 6.948e+02, threshold=6.372e+02, percent-clipped=1.0 2023-05-16 04:25:25,188 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.1889, 4.2084, 2.6015, 2.3520, 3.5412, 2.4375, 3.6778, 3.0039], device='cuda:0'), covar=tensor([0.0676, 0.0607, 0.1179, 0.1511, 0.0328, 0.1282, 0.0489, 0.0815], device='cuda:0'), in_proj_covar=tensor([0.0183, 0.0245, 0.0173, 0.0195, 0.0135, 0.0177, 0.0191, 0.0172], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 04:25:46,441 INFO [finetune.py:992] (0/2) Epoch 6, batch 2700, loss[loss=0.2196, simple_loss=0.3072, pruned_loss=0.06602, over 10535.00 frames. ], tot_loss[loss=0.1742, simple_loss=0.2619, pruned_loss=0.04322, over 2374320.31 frames. ], batch size: 68, lr: 4.59e-03, grad_scale: 4.0 2023-05-16 04:25:56,540 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.63 vs. limit=2.0 2023-05-16 04:25:58,396 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([6.2053, 6.1114, 6.0004, 5.4984, 5.2119, 6.1180, 5.6972, 5.3594], device='cuda:0'), covar=tensor([0.0525, 0.0878, 0.0529, 0.1528, 0.0627, 0.0640, 0.1511, 0.1088], device='cuda:0'), in_proj_covar=tensor([0.0570, 0.0507, 0.0473, 0.0591, 0.0386, 0.0667, 0.0725, 0.0532], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0004, 0.0004, 0.0003], device='cuda:0') 2023-05-16 04:26:05,294 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=163896.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 04:26:22,183 INFO [finetune.py:992] (0/2) Epoch 6, batch 2750, loss[loss=0.1466, simple_loss=0.2303, pruned_loss=0.03145, over 12017.00 frames. ], tot_loss[loss=0.1739, simple_loss=0.2615, pruned_loss=0.04314, over 2374702.39 frames. ], batch size: 28, lr: 4.59e-03, grad_scale: 4.0 2023-05-16 04:26:37,204 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.057e+02 2.839e+02 3.207e+02 3.708e+02 7.990e+02, threshold=6.415e+02, percent-clipped=1.0 2023-05-16 04:26:40,779 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=163944.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 04:26:58,613 INFO [finetune.py:992] (0/2) Epoch 6, batch 2800, loss[loss=0.1678, simple_loss=0.2589, pruned_loss=0.03839, over 12160.00 frames. ], tot_loss[loss=0.1732, simple_loss=0.2606, pruned_loss=0.04292, over 2374993.29 frames. ], batch size: 34, lr: 4.59e-03, grad_scale: 8.0 2023-05-16 04:27:06,585 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.7348, 2.9090, 5.3039, 2.4715, 2.3943, 4.0389, 2.9425, 3.9590], device='cuda:0'), covar=tensor([0.0383, 0.1432, 0.0195, 0.1309, 0.2075, 0.1234, 0.1577, 0.1040], device='cuda:0'), in_proj_covar=tensor([0.0232, 0.0230, 0.0232, 0.0180, 0.0234, 0.0280, 0.0224, 0.0250], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:0') 2023-05-16 04:27:20,821 INFO [checkpoint.py:75] (0/2) Saving checkpoint to pruned_transducer_stateless7/exp_giga_finetune/checkpoint-64000.pt 2023-05-16 04:27:25,792 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.9355, 3.8919, 3.8306, 3.9965, 3.7078, 3.7433, 3.7381, 3.9323], device='cuda:0'), covar=tensor([0.1003, 0.0861, 0.1694, 0.0720, 0.1963, 0.1453, 0.0572, 0.0921], device='cuda:0'), in_proj_covar=tensor([0.0495, 0.0645, 0.0557, 0.0587, 0.0794, 0.0702, 0.0523, 0.0472], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0005, 0.0004, 0.0004, 0.0005, 0.0005, 0.0004, 0.0004], device='cuda:0') 2023-05-16 04:27:36,864 INFO [finetune.py:992] (0/2) Epoch 6, batch 2850, loss[loss=0.147, simple_loss=0.2259, pruned_loss=0.03406, over 12363.00 frames. ], tot_loss[loss=0.1736, simple_loss=0.2609, pruned_loss=0.04316, over 2370477.14 frames. ], batch size: 30, lr: 4.59e-03, grad_scale: 8.0 2023-05-16 04:27:50,949 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.814e+02 2.814e+02 3.240e+02 3.538e+02 6.135e+02, threshold=6.480e+02, percent-clipped=0.0 2023-05-16 04:27:54,896 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-05-16 04:28:08,702 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=2.54 vs. limit=5.0 2023-05-16 04:28:12,654 INFO [finetune.py:992] (0/2) Epoch 6, batch 2900, loss[loss=0.175, simple_loss=0.2698, pruned_loss=0.0401, over 12284.00 frames. ], tot_loss[loss=0.1735, simple_loss=0.261, pruned_loss=0.04299, over 2374224.50 frames. ], batch size: 33, lr: 4.59e-03, grad_scale: 8.0 2023-05-16 04:28:28,440 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=164090.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 04:28:39,775 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.4410, 5.2288, 5.3707, 5.4044, 4.9910, 5.0487, 4.8860, 5.3545], device='cuda:0'), covar=tensor([0.0670, 0.0658, 0.0706, 0.0555, 0.2003, 0.1203, 0.0549, 0.1004], device='cuda:0'), in_proj_covar=tensor([0.0493, 0.0646, 0.0557, 0.0587, 0.0796, 0.0702, 0.0522, 0.0471], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0005, 0.0004, 0.0004, 0.0005, 0.0005, 0.0004, 0.0004], device='cuda:0') 2023-05-16 04:28:48,990 INFO [finetune.py:992] (0/2) Epoch 6, batch 2950, loss[loss=0.1738, simple_loss=0.2671, pruned_loss=0.04025, over 12364.00 frames. ], tot_loss[loss=0.1731, simple_loss=0.2608, pruned_loss=0.0427, over 2370287.53 frames. ], batch size: 35, lr: 4.58e-03, grad_scale: 8.0 2023-05-16 04:29:03,529 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.869e+02 2.731e+02 3.279e+02 3.824e+02 1.716e+03, threshold=6.558e+02, percent-clipped=4.0 2023-05-16 04:29:12,380 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=164151.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 04:29:15,151 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=164155.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 04:29:24,938 INFO [finetune.py:992] (0/2) Epoch 6, batch 3000, loss[loss=0.1584, simple_loss=0.2452, pruned_loss=0.03577, over 11816.00 frames. ], tot_loss[loss=0.1733, simple_loss=0.2612, pruned_loss=0.04274, over 2373910.93 frames. ], batch size: 26, lr: 4.58e-03, grad_scale: 8.0 2023-05-16 04:29:24,939 INFO [finetune.py:1017] (0/2) Computing validation loss 2023-05-16 04:29:31,026 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.9508, 4.1352, 3.9446, 4.6508, 4.4341, 2.6947, 4.2241, 3.0068], device='cuda:0'), covar=tensor([0.0895, 0.1087, 0.1290, 0.0419, 0.1054, 0.1785, 0.0973, 0.3517], device='cuda:0'), in_proj_covar=tensor([0.0295, 0.0370, 0.0346, 0.0268, 0.0356, 0.0264, 0.0329, 0.0359], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 04:29:43,358 INFO [finetune.py:1026] (0/2) Epoch 6, validation: loss=0.3222, simple_loss=0.3994, pruned_loss=0.1225, over 1020973.00 frames. 2023-05-16 04:29:43,359 INFO [finetune.py:1027] (0/2) Maximum memory allocated so far is 12745MB 2023-05-16 04:30:17,611 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=164216.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 04:30:19,530 INFO [finetune.py:992] (0/2) Epoch 6, batch 3050, loss[loss=0.2051, simple_loss=0.286, pruned_loss=0.06212, over 11264.00 frames. ], tot_loss[loss=0.1742, simple_loss=0.2621, pruned_loss=0.04318, over 2378005.00 frames. ], batch size: 55, lr: 4.58e-03, grad_scale: 8.0 2023-05-16 04:30:33,826 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.898e+02 2.794e+02 3.284e+02 3.936e+02 7.141e+02, threshold=6.567e+02, percent-clipped=3.0 2023-05-16 04:30:37,494 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.0165, 4.6329, 5.0236, 4.4356, 4.6849, 4.4277, 5.0448, 4.6891], device='cuda:0'), covar=tensor([0.0236, 0.0344, 0.0247, 0.0222, 0.0288, 0.0313, 0.0177, 0.0456], device='cuda:0'), in_proj_covar=tensor([0.0240, 0.0249, 0.0265, 0.0241, 0.0240, 0.0240, 0.0216, 0.0204], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 04:30:55,508 INFO [finetune.py:992] (0/2) Epoch 6, batch 3100, loss[loss=0.2397, simple_loss=0.3105, pruned_loss=0.08443, over 7629.00 frames. ], tot_loss[loss=0.1745, simple_loss=0.2626, pruned_loss=0.04318, over 2380185.33 frames. ], batch size: 98, lr: 4.58e-03, grad_scale: 8.0 2023-05-16 04:31:15,498 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.9473, 4.8185, 4.7931, 4.7677, 4.4132, 4.8248, 4.8177, 5.0790], device='cuda:0'), covar=tensor([0.0241, 0.0146, 0.0168, 0.0320, 0.0722, 0.0331, 0.0162, 0.0174], device='cuda:0'), in_proj_covar=tensor([0.0182, 0.0185, 0.0185, 0.0234, 0.0235, 0.0204, 0.0165, 0.0217], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0005, 0.0005, 0.0005, 0.0003, 0.0005], device='cuda:0') 2023-05-16 04:31:31,736 INFO [finetune.py:992] (0/2) Epoch 6, batch 3150, loss[loss=0.1518, simple_loss=0.2325, pruned_loss=0.0355, over 12282.00 frames. ], tot_loss[loss=0.1746, simple_loss=0.263, pruned_loss=0.04309, over 2378613.57 frames. ], batch size: 28, lr: 4.58e-03, grad_scale: 8.0 2023-05-16 04:31:46,572 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.748e+02 2.672e+02 3.320e+02 3.847e+02 1.508e+03, threshold=6.640e+02, percent-clipped=5.0 2023-05-16 04:32:01,534 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=164360.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 04:32:07,333 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.7775, 2.6191, 3.5362, 3.6522, 2.8808, 2.6576, 2.6547, 2.4647], device='cuda:0'), covar=tensor([0.1142, 0.2243, 0.0622, 0.0489, 0.0891, 0.1827, 0.2219, 0.2964], device='cuda:0'), in_proj_covar=tensor([0.0275, 0.0373, 0.0260, 0.0291, 0.0256, 0.0285, 0.0355, 0.0359], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-16 04:32:07,781 INFO [finetune.py:992] (0/2) Epoch 6, batch 3200, loss[loss=0.2351, simple_loss=0.304, pruned_loss=0.08313, over 7795.00 frames. ], tot_loss[loss=0.1743, simple_loss=0.2625, pruned_loss=0.043, over 2371734.23 frames. ], batch size: 98, lr: 4.58e-03, grad_scale: 8.0 2023-05-16 04:32:43,312 INFO [finetune.py:992] (0/2) Epoch 6, batch 3250, loss[loss=0.1834, simple_loss=0.2699, pruned_loss=0.04844, over 12134.00 frames. ], tot_loss[loss=0.1748, simple_loss=0.263, pruned_loss=0.04332, over 2360293.36 frames. ], batch size: 39, lr: 4.58e-03, grad_scale: 8.0 2023-05-16 04:32:45,027 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=164421.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 04:32:58,254 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.220e+02 2.909e+02 3.393e+02 4.090e+02 1.179e+03, threshold=6.786e+02, percent-clipped=4.0 2023-05-16 04:33:03,371 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=164446.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 04:33:09,723 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=2.53 vs. limit=5.0 2023-05-16 04:33:20,174 INFO [finetune.py:992] (0/2) Epoch 6, batch 3300, loss[loss=0.1626, simple_loss=0.2575, pruned_loss=0.03385, over 12296.00 frames. ], tot_loss[loss=0.1748, simple_loss=0.2624, pruned_loss=0.04362, over 2347068.27 frames. ], batch size: 34, lr: 4.58e-03, grad_scale: 8.0 2023-05-16 04:33:50,128 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=164511.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 04:33:55,504 INFO [finetune.py:992] (0/2) Epoch 6, batch 3350, loss[loss=0.1887, simple_loss=0.2905, pruned_loss=0.04345, over 11788.00 frames. ], tot_loss[loss=0.1744, simple_loss=0.2618, pruned_loss=0.04355, over 2348132.18 frames. ], batch size: 44, lr: 4.58e-03, grad_scale: 8.0 2023-05-16 04:34:09,195 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.926e+02 2.787e+02 3.205e+02 4.035e+02 6.481e+02, threshold=6.410e+02, percent-clipped=0.0 2023-05-16 04:34:20,110 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.5695, 5.1118, 5.5384, 4.8652, 5.1489, 4.9518, 5.5674, 5.1445], device='cuda:0'), covar=tensor([0.0205, 0.0304, 0.0204, 0.0211, 0.0283, 0.0247, 0.0180, 0.0272], device='cuda:0'), in_proj_covar=tensor([0.0242, 0.0251, 0.0268, 0.0244, 0.0243, 0.0242, 0.0219, 0.0207], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 04:34:30,102 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=164568.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 04:34:30,651 INFO [finetune.py:992] (0/2) Epoch 6, batch 3400, loss[loss=0.1502, simple_loss=0.2452, pruned_loss=0.02758, over 12037.00 frames. ], tot_loss[loss=0.1738, simple_loss=0.2615, pruned_loss=0.0431, over 2360891.41 frames. ], batch size: 31, lr: 4.58e-03, grad_scale: 8.0 2023-05-16 04:34:37,173 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=164578.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 04:34:46,935 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([6.1130, 6.1073, 5.9129, 5.4284, 5.1902, 6.0428, 5.6178, 5.4179], device='cuda:0'), covar=tensor([0.0614, 0.0719, 0.0557, 0.1598, 0.0604, 0.0634, 0.1338, 0.0989], device='cuda:0'), in_proj_covar=tensor([0.0576, 0.0511, 0.0480, 0.0596, 0.0387, 0.0672, 0.0728, 0.0535], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0004, 0.0004, 0.0003], device='cuda:0') 2023-05-16 04:34:47,080 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.8221, 3.3058, 5.2273, 2.7565, 2.9060, 3.9358, 3.3997, 3.9641], device='cuda:0'), covar=tensor([0.0441, 0.1111, 0.0251, 0.1099, 0.1754, 0.1300, 0.1199, 0.1074], device='cuda:0'), in_proj_covar=tensor([0.0229, 0.0228, 0.0231, 0.0178, 0.0232, 0.0277, 0.0221, 0.0249], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:0') 2023-05-16 04:35:07,966 INFO [finetune.py:992] (0/2) Epoch 6, batch 3450, loss[loss=0.1375, simple_loss=0.2187, pruned_loss=0.02816, over 12288.00 frames. ], tot_loss[loss=0.1726, simple_loss=0.2603, pruned_loss=0.04241, over 2367436.29 frames. ], batch size: 28, lr: 4.58e-03, grad_scale: 8.0 2023-05-16 04:35:15,288 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=164629.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 04:35:22,007 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.872e+02 2.736e+02 3.192e+02 3.839e+02 9.389e+02, threshold=6.385e+02, percent-clipped=1.0 2023-05-16 04:35:22,275 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=164639.0, num_to_drop=1, layers_to_drop={2} 2023-05-16 04:35:29,630 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 2023-05-16 04:35:35,912 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=164658.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 04:35:41,657 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.1800, 2.4762, 3.6938, 3.1312, 3.5061, 3.2896, 2.7347, 3.6141], device='cuda:0'), covar=tensor([0.0108, 0.0305, 0.0109, 0.0212, 0.0135, 0.0140, 0.0274, 0.0096], device='cuda:0'), in_proj_covar=tensor([0.0165, 0.0195, 0.0177, 0.0174, 0.0200, 0.0152, 0.0186, 0.0166], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 04:35:43,531 INFO [finetune.py:992] (0/2) Epoch 6, batch 3500, loss[loss=0.1665, simple_loss=0.2537, pruned_loss=0.03963, over 12197.00 frames. ], tot_loss[loss=0.1724, simple_loss=0.2601, pruned_loss=0.04231, over 2368406.44 frames. ], batch size: 29, lr: 4.58e-03, grad_scale: 8.0 2023-05-16 04:36:16,872 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=164716.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 04:36:18,930 INFO [finetune.py:992] (0/2) Epoch 6, batch 3550, loss[loss=0.1867, simple_loss=0.2784, pruned_loss=0.04752, over 12158.00 frames. ], tot_loss[loss=0.1728, simple_loss=0.2607, pruned_loss=0.04246, over 2371010.28 frames. ], batch size: 36, lr: 4.58e-03, grad_scale: 8.0 2023-05-16 04:36:19,125 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=164719.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 04:36:33,855 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.067e+02 2.920e+02 3.415e+02 3.980e+02 6.979e+02, threshold=6.831e+02, percent-clipped=1.0 2023-05-16 04:36:38,936 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=164746.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 04:36:38,968 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=164746.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 04:36:45,474 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.6955, 3.2860, 5.1234, 2.6841, 2.8260, 3.9389, 3.1007, 3.8863], device='cuda:0'), covar=tensor([0.0443, 0.1137, 0.0274, 0.1127, 0.1820, 0.1192, 0.1486, 0.0982], device='cuda:0'), in_proj_covar=tensor([0.0230, 0.0227, 0.0232, 0.0179, 0.0232, 0.0277, 0.0221, 0.0249], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:0') 2023-05-16 04:36:56,021 INFO [finetune.py:992] (0/2) Epoch 6, batch 3600, loss[loss=0.1725, simple_loss=0.2659, pruned_loss=0.03956, over 12165.00 frames. ], tot_loss[loss=0.1727, simple_loss=0.2606, pruned_loss=0.04235, over 2373704.74 frames. ], batch size: 31, lr: 4.58e-03, grad_scale: 8.0 2023-05-16 04:37:13,972 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=164794.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 04:37:17,297 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.40 vs. limit=2.0 2023-05-16 04:37:23,574 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=164807.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 04:37:26,362 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=164811.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 04:37:31,878 INFO [finetune.py:992] (0/2) Epoch 6, batch 3650, loss[loss=0.1616, simple_loss=0.2505, pruned_loss=0.03639, over 12104.00 frames. ], tot_loss[loss=0.1718, simple_loss=0.2598, pruned_loss=0.04194, over 2375666.20 frames. ], batch size: 33, lr: 4.58e-03, grad_scale: 8.0 2023-05-16 04:37:43,360 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.4026, 4.1681, 4.3104, 4.4965, 3.3189, 3.9672, 2.6442, 4.1108], device='cuda:0'), covar=tensor([0.1608, 0.0762, 0.1005, 0.0711, 0.1066, 0.0664, 0.1888, 0.1807], device='cuda:0'), in_proj_covar=tensor([0.0226, 0.0264, 0.0295, 0.0352, 0.0235, 0.0237, 0.0255, 0.0380], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 04:37:45,910 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.779e+02 2.801e+02 3.356e+02 4.004e+02 7.977e+02, threshold=6.711e+02, percent-clipped=2.0 2023-05-16 04:38:00,264 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=164859.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 04:38:07,842 INFO [finetune.py:992] (0/2) Epoch 6, batch 3700, loss[loss=0.2016, simple_loss=0.284, pruned_loss=0.05959, over 12354.00 frames. ], tot_loss[loss=0.1727, simple_loss=0.2603, pruned_loss=0.04255, over 2370809.70 frames. ], batch size: 35, lr: 4.58e-03, grad_scale: 8.0 2023-05-16 04:38:44,103 INFO [finetune.py:992] (0/2) Epoch 6, batch 3750, loss[loss=0.1726, simple_loss=0.2623, pruned_loss=0.04144, over 12350.00 frames. ], tot_loss[loss=0.1729, simple_loss=0.2612, pruned_loss=0.04232, over 2375308.80 frames. ], batch size: 35, lr: 4.58e-03, grad_scale: 8.0 2023-05-16 04:38:47,733 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=164924.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 04:38:54,684 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=164934.0, num_to_drop=1, layers_to_drop={2} 2023-05-16 04:38:58,188 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.882e+02 2.737e+02 3.424e+02 3.917e+02 2.058e+03, threshold=6.847e+02, percent-clipped=1.0 2023-05-16 04:39:19,912 INFO [finetune.py:992] (0/2) Epoch 6, batch 3800, loss[loss=0.1653, simple_loss=0.2404, pruned_loss=0.04506, over 12025.00 frames. ], tot_loss[loss=0.1725, simple_loss=0.2608, pruned_loss=0.04207, over 2380913.22 frames. ], batch size: 31, lr: 4.58e-03, grad_scale: 8.0 2023-05-16 04:39:52,507 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=165014.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 04:39:53,793 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=165016.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 04:39:54,609 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.2423, 2.7634, 3.7612, 3.2589, 3.5309, 3.3462, 2.8327, 3.6140], device='cuda:0'), covar=tensor([0.0135, 0.0270, 0.0135, 0.0196, 0.0150, 0.0147, 0.0269, 0.0121], device='cuda:0'), in_proj_covar=tensor([0.0167, 0.0195, 0.0178, 0.0174, 0.0201, 0.0154, 0.0187, 0.0168], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 04:39:55,855 INFO [finetune.py:992] (0/2) Epoch 6, batch 3850, loss[loss=0.141, simple_loss=0.2219, pruned_loss=0.03008, over 12265.00 frames. ], tot_loss[loss=0.1729, simple_loss=0.2606, pruned_loss=0.0426, over 2367877.33 frames. ], batch size: 28, lr: 4.58e-03, grad_scale: 8.0 2023-05-16 04:40:09,946 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.918e+02 2.707e+02 3.274e+02 3.886e+02 1.094e+03, threshold=6.548e+02, percent-clipped=4.0 2023-05-16 04:40:28,679 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=165064.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 04:40:32,081 INFO [finetune.py:992] (0/2) Epoch 6, batch 3900, loss[loss=0.1664, simple_loss=0.2499, pruned_loss=0.04146, over 12130.00 frames. ], tot_loss[loss=0.1727, simple_loss=0.2604, pruned_loss=0.04253, over 2371600.12 frames. ], batch size: 30, lr: 4.57e-03, grad_scale: 8.0 2023-05-16 04:40:45,243 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.2336, 3.1106, 3.0481, 2.9365, 2.6711, 2.4934, 3.1852, 2.0889], device='cuda:0'), covar=tensor([0.0342, 0.0126, 0.0145, 0.0155, 0.0307, 0.0292, 0.0118, 0.0388], device='cuda:0'), in_proj_covar=tensor([0.0186, 0.0153, 0.0151, 0.0179, 0.0198, 0.0191, 0.0159, 0.0194], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 04:40:45,383 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2023-05-16 04:40:55,728 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=165102.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 04:41:07,762 INFO [finetune.py:992] (0/2) Epoch 6, batch 3950, loss[loss=0.1592, simple_loss=0.2325, pruned_loss=0.04292, over 11753.00 frames. ], tot_loss[loss=0.1722, simple_loss=0.2598, pruned_loss=0.04234, over 2378015.71 frames. ], batch size: 26, lr: 4.57e-03, grad_scale: 8.0 2023-05-16 04:41:10,915 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=165123.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 04:41:21,829 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.843e+02 2.809e+02 3.386e+02 4.084e+02 8.406e+02, threshold=6.772e+02, percent-clipped=2.0 2023-05-16 04:41:43,512 INFO [finetune.py:992] (0/2) Epoch 6, batch 4000, loss[loss=0.1568, simple_loss=0.2377, pruned_loss=0.03791, over 11330.00 frames. ], tot_loss[loss=0.1732, simple_loss=0.2609, pruned_loss=0.0428, over 2378395.18 frames. ], batch size: 25, lr: 4.57e-03, grad_scale: 8.0 2023-05-16 04:41:54,504 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=165184.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 04:42:02,332 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.3076, 4.4738, 2.8024, 2.3864, 3.7929, 2.3146, 3.8339, 3.1347], device='cuda:0'), covar=tensor([0.0527, 0.0512, 0.0913, 0.1448, 0.0297, 0.1334, 0.0427, 0.0746], device='cuda:0'), in_proj_covar=tensor([0.0185, 0.0249, 0.0173, 0.0197, 0.0137, 0.0179, 0.0194, 0.0174], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 04:42:20,390 INFO [finetune.py:992] (0/2) Epoch 6, batch 4050, loss[loss=0.1532, simple_loss=0.2402, pruned_loss=0.03309, over 12289.00 frames. ], tot_loss[loss=0.1733, simple_loss=0.2609, pruned_loss=0.04284, over 2380325.47 frames. ], batch size: 28, lr: 4.57e-03, grad_scale: 8.0 2023-05-16 04:42:24,062 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=165224.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 04:42:31,095 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=165234.0, num_to_drop=1, layers_to_drop={1} 2023-05-16 04:42:34,442 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.951e+02 2.876e+02 3.378e+02 4.022e+02 7.072e+02, threshold=6.755e+02, percent-clipped=1.0 2023-05-16 04:42:38,035 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.2756, 6.1264, 5.6074, 5.7323, 6.1459, 5.4917, 5.8431, 5.6900], device='cuda:0'), covar=tensor([0.1355, 0.0749, 0.1056, 0.1595, 0.0896, 0.2313, 0.1327, 0.0944], device='cuda:0'), in_proj_covar=tensor([0.0348, 0.0466, 0.0370, 0.0423, 0.0452, 0.0432, 0.0379, 0.0366], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 04:42:55,896 INFO [finetune.py:992] (0/2) Epoch 6, batch 4100, loss[loss=0.145, simple_loss=0.2247, pruned_loss=0.03267, over 12033.00 frames. ], tot_loss[loss=0.1737, simple_loss=0.2614, pruned_loss=0.04305, over 2376873.58 frames. ], batch size: 28, lr: 4.57e-03, grad_scale: 8.0 2023-05-16 04:42:58,192 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=165272.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 04:43:05,112 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=165282.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 04:43:17,555 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.62 vs. limit=2.0 2023-05-16 04:43:25,122 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.0189, 2.3919, 3.6065, 3.0076, 3.4158, 3.1518, 2.5055, 3.5162], device='cuda:0'), covar=tensor([0.0124, 0.0337, 0.0153, 0.0201, 0.0136, 0.0166, 0.0358, 0.0105], device='cuda:0'), in_proj_covar=tensor([0.0167, 0.0195, 0.0178, 0.0174, 0.0200, 0.0153, 0.0187, 0.0168], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 04:43:28,638 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=165314.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 04:43:32,088 INFO [finetune.py:992] (0/2) Epoch 6, batch 4150, loss[loss=0.1689, simple_loss=0.2572, pruned_loss=0.0403, over 12076.00 frames. ], tot_loss[loss=0.1727, simple_loss=0.2604, pruned_loss=0.04249, over 2380737.56 frames. ], batch size: 32, lr: 4.57e-03, grad_scale: 8.0 2023-05-16 04:43:45,796 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.801e+02 2.822e+02 3.342e+02 4.030e+02 7.682e+02, threshold=6.684e+02, percent-clipped=1.0 2023-05-16 04:43:53,277 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.89 vs. limit=2.0 2023-05-16 04:44:02,646 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=165362.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 04:44:07,627 INFO [finetune.py:992] (0/2) Epoch 6, batch 4200, loss[loss=0.157, simple_loss=0.2398, pruned_loss=0.03707, over 11782.00 frames. ], tot_loss[loss=0.1739, simple_loss=0.2615, pruned_loss=0.04314, over 2370647.91 frames. ], batch size: 26, lr: 4.57e-03, grad_scale: 8.0 2023-05-16 04:44:09,346 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.5774, 3.9763, 3.9016, 4.4222, 3.4799, 4.0166, 2.5561, 4.2945], device='cuda:0'), covar=tensor([0.1266, 0.0704, 0.1185, 0.0770, 0.0771, 0.0502, 0.1661, 0.1351], device='cuda:0'), in_proj_covar=tensor([0.0226, 0.0263, 0.0295, 0.0354, 0.0235, 0.0238, 0.0256, 0.0381], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 04:44:12,842 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.8014, 3.8458, 3.5560, 3.4755, 3.2254, 3.0856, 3.9242, 2.6044], device='cuda:0'), covar=tensor([0.0284, 0.0086, 0.0136, 0.0146, 0.0315, 0.0275, 0.0107, 0.0331], device='cuda:0'), in_proj_covar=tensor([0.0186, 0.0153, 0.0151, 0.0179, 0.0198, 0.0190, 0.0158, 0.0193], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 04:44:31,723 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=165402.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 04:44:43,700 INFO [finetune.py:992] (0/2) Epoch 6, batch 4250, loss[loss=0.2038, simple_loss=0.2795, pruned_loss=0.06407, over 8168.00 frames. ], tot_loss[loss=0.1736, simple_loss=0.2613, pruned_loss=0.04289, over 2366004.54 frames. ], batch size: 98, lr: 4.57e-03, grad_scale: 8.0 2023-05-16 04:44:58,494 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.070e+02 2.871e+02 3.395e+02 3.960e+02 1.167e+03, threshold=6.791e+02, percent-clipped=2.0 2023-05-16 04:45:06,279 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=165450.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 04:45:19,592 INFO [finetune.py:992] (0/2) Epoch 6, batch 4300, loss[loss=0.1832, simple_loss=0.2814, pruned_loss=0.04251, over 12148.00 frames. ], tot_loss[loss=0.174, simple_loss=0.262, pruned_loss=0.043, over 2371467.52 frames. ], batch size: 39, lr: 4.57e-03, grad_scale: 8.0 2023-05-16 04:45:25,363 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.5706, 5.3654, 5.5049, 5.5662, 5.1243, 5.2114, 4.9724, 5.4914], device='cuda:0'), covar=tensor([0.0557, 0.0561, 0.0675, 0.0480, 0.1902, 0.1106, 0.0517, 0.0900], device='cuda:0'), in_proj_covar=tensor([0.0498, 0.0653, 0.0560, 0.0588, 0.0812, 0.0712, 0.0529, 0.0475], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0005, 0.0004, 0.0004, 0.0005, 0.0005, 0.0004, 0.0004], device='cuda:0') 2023-05-16 04:45:26,796 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=165479.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 04:45:49,111 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.0029, 4.3200, 3.7822, 4.5869, 4.2539, 2.8871, 4.0831, 2.9674], device='cuda:0'), covar=tensor([0.0870, 0.0808, 0.1435, 0.0529, 0.1116, 0.1467, 0.0924, 0.3068], device='cuda:0'), in_proj_covar=tensor([0.0297, 0.0369, 0.0349, 0.0270, 0.0358, 0.0265, 0.0331, 0.0358], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 04:45:55,517 INFO [finetune.py:992] (0/2) Epoch 6, batch 4350, loss[loss=0.1863, simple_loss=0.2718, pruned_loss=0.05043, over 10732.00 frames. ], tot_loss[loss=0.1741, simple_loss=0.2621, pruned_loss=0.04307, over 2369115.25 frames. ], batch size: 68, lr: 4.57e-03, grad_scale: 8.0 2023-05-16 04:46:09,536 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.091e+02 2.872e+02 3.335e+02 4.070e+02 1.129e+03, threshold=6.671e+02, percent-clipped=3.0 2023-05-16 04:46:18,329 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=165551.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 04:46:30,026 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.9090, 3.4796, 5.2124, 2.8268, 2.8969, 4.0432, 3.5236, 3.9739], device='cuda:0'), covar=tensor([0.0392, 0.0987, 0.0206, 0.0968, 0.1699, 0.1094, 0.1090, 0.1120], device='cuda:0'), in_proj_covar=tensor([0.0231, 0.0228, 0.0232, 0.0178, 0.0233, 0.0279, 0.0222, 0.0251], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:0') 2023-05-16 04:46:30,892 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 2023-05-16 04:46:31,191 INFO [finetune.py:992] (0/2) Epoch 6, batch 4400, loss[loss=0.1853, simple_loss=0.2764, pruned_loss=0.04705, over 12358.00 frames. ], tot_loss[loss=0.1734, simple_loss=0.2615, pruned_loss=0.04264, over 2375498.81 frames. ], batch size: 35, lr: 4.57e-03, grad_scale: 8.0 2023-05-16 04:46:35,868 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=4.91 vs. limit=5.0 2023-05-16 04:47:02,743 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.2903, 4.8048, 5.2375, 4.6130, 4.8539, 4.6957, 5.2669, 4.9020], device='cuda:0'), covar=tensor([0.0241, 0.0375, 0.0265, 0.0256, 0.0327, 0.0310, 0.0218, 0.0307], device='cuda:0'), in_proj_covar=tensor([0.0243, 0.0250, 0.0269, 0.0245, 0.0243, 0.0242, 0.0219, 0.0206], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 04:47:02,802 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=165612.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 04:47:07,535 INFO [finetune.py:992] (0/2) Epoch 6, batch 4450, loss[loss=0.1803, simple_loss=0.2763, pruned_loss=0.04217, over 12091.00 frames. ], tot_loss[loss=0.1736, simple_loss=0.2617, pruned_loss=0.04282, over 2373078.19 frames. ], batch size: 38, lr: 4.57e-03, grad_scale: 8.0 2023-05-16 04:47:21,523 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.094e+02 2.699e+02 3.249e+02 3.797e+02 5.988e+02, threshold=6.497e+02, percent-clipped=0.0 2023-05-16 04:47:43,180 INFO [finetune.py:992] (0/2) Epoch 6, batch 4500, loss[loss=0.1876, simple_loss=0.267, pruned_loss=0.05412, over 12123.00 frames. ], tot_loss[loss=0.1749, simple_loss=0.263, pruned_loss=0.04338, over 2373614.62 frames. ], batch size: 38, lr: 4.57e-03, grad_scale: 8.0 2023-05-16 04:48:18,246 INFO [finetune.py:992] (0/2) Epoch 6, batch 4550, loss[loss=0.2033, simple_loss=0.2927, pruned_loss=0.05696, over 12367.00 frames. ], tot_loss[loss=0.1758, simple_loss=0.2639, pruned_loss=0.04391, over 2366456.85 frames. ], batch size: 35, lr: 4.57e-03, grad_scale: 8.0 2023-05-16 04:48:33,096 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.258e+02 2.965e+02 3.483e+02 4.253e+02 9.406e+02, threshold=6.967e+02, percent-clipped=5.0 2023-05-16 04:48:34,671 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.3958, 4.2763, 4.2502, 4.3325, 3.8845, 4.3769, 4.3556, 4.5548], device='cuda:0'), covar=tensor([0.0205, 0.0157, 0.0207, 0.0308, 0.0743, 0.0352, 0.0171, 0.0188], device='cuda:0'), in_proj_covar=tensor([0.0182, 0.0184, 0.0186, 0.0233, 0.0235, 0.0205, 0.0165, 0.0215], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0005, 0.0004, 0.0005, 0.0004, 0.0005], device='cuda:0') 2023-05-16 04:48:39,084 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.3572, 4.1623, 4.1594, 4.5126, 3.1824, 3.8993, 2.8169, 4.0110], device='cuda:0'), covar=tensor([0.1586, 0.0703, 0.0937, 0.0604, 0.1094, 0.0623, 0.1664, 0.1602], device='cuda:0'), in_proj_covar=tensor([0.0225, 0.0261, 0.0295, 0.0353, 0.0234, 0.0237, 0.0254, 0.0379], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 04:48:41,186 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=165750.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 04:48:44,071 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.1543, 4.8506, 4.9919, 4.9606, 4.7808, 5.0761, 4.8859, 2.7358], device='cuda:0'), covar=tensor([0.0084, 0.0066, 0.0081, 0.0059, 0.0042, 0.0073, 0.0087, 0.0689], device='cuda:0'), in_proj_covar=tensor([0.0065, 0.0074, 0.0078, 0.0071, 0.0059, 0.0088, 0.0077, 0.0092], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 04:48:54,397 INFO [finetune.py:992] (0/2) Epoch 6, batch 4600, loss[loss=0.1841, simple_loss=0.2711, pruned_loss=0.04858, over 12042.00 frames. ], tot_loss[loss=0.1752, simple_loss=0.263, pruned_loss=0.04371, over 2366694.14 frames. ], batch size: 42, lr: 4.57e-03, grad_scale: 16.0 2023-05-16 04:49:01,524 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=165779.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 04:49:14,908 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.7605, 3.9738, 3.6280, 4.2379, 3.9247, 2.7037, 3.6956, 2.8647], device='cuda:0'), covar=tensor([0.0870, 0.0904, 0.1364, 0.0519, 0.1125, 0.1595, 0.1000, 0.2856], device='cuda:0'), in_proj_covar=tensor([0.0296, 0.0369, 0.0348, 0.0270, 0.0357, 0.0264, 0.0331, 0.0358], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 04:49:19,450 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=165803.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 04:49:22,431 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.1399, 2.8487, 2.7910, 2.7682, 2.4783, 2.3423, 2.9121, 1.9916], device='cuda:0'), covar=tensor([0.0354, 0.0178, 0.0187, 0.0162, 0.0375, 0.0258, 0.0131, 0.0402], device='cuda:0'), in_proj_covar=tensor([0.0188, 0.0154, 0.0153, 0.0180, 0.0199, 0.0191, 0.0159, 0.0196], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 04:49:25,251 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.4771, 2.1873, 3.8173, 4.4869, 4.1307, 4.2341, 4.0352, 2.9618], device='cuda:0'), covar=tensor([0.0028, 0.0471, 0.0106, 0.0031, 0.0070, 0.0081, 0.0078, 0.0358], device='cuda:0'), in_proj_covar=tensor([0.0081, 0.0120, 0.0100, 0.0075, 0.0097, 0.0111, 0.0089, 0.0136], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 04:49:25,275 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=165811.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 04:49:30,744 INFO [finetune.py:992] (0/2) Epoch 6, batch 4650, loss[loss=0.1598, simple_loss=0.256, pruned_loss=0.0318, over 12109.00 frames. ], tot_loss[loss=0.1754, simple_loss=0.2633, pruned_loss=0.04372, over 2371335.28 frames. ], batch size: 33, lr: 4.57e-03, grad_scale: 16.0 2023-05-16 04:49:36,497 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=165827.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 04:49:39,430 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.0010, 2.2644, 3.6072, 3.0003, 3.3840, 3.1999, 2.5492, 3.5209], device='cuda:0'), covar=tensor([0.0120, 0.0361, 0.0134, 0.0227, 0.0153, 0.0156, 0.0328, 0.0108], device='cuda:0'), in_proj_covar=tensor([0.0168, 0.0197, 0.0180, 0.0176, 0.0202, 0.0155, 0.0189, 0.0169], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 04:49:44,552 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.799e+02 2.807e+02 3.354e+02 4.024e+02 1.121e+03, threshold=6.709e+02, percent-clipped=1.0 2023-05-16 04:49:57,101 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.71 vs. limit=2.0 2023-05-16 04:50:02,728 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=165864.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 04:50:04,892 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.4966, 4.9826, 5.4497, 4.7410, 5.0273, 4.7299, 5.4890, 5.0444], device='cuda:0'), covar=tensor([0.0222, 0.0396, 0.0265, 0.0235, 0.0317, 0.0354, 0.0205, 0.0247], device='cuda:0'), in_proj_covar=tensor([0.0243, 0.0251, 0.0268, 0.0246, 0.0243, 0.0244, 0.0220, 0.0204], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 04:50:06,673 INFO [finetune.py:992] (0/2) Epoch 6, batch 4700, loss[loss=0.203, simple_loss=0.2827, pruned_loss=0.06163, over 12048.00 frames. ], tot_loss[loss=0.1756, simple_loss=0.2633, pruned_loss=0.04394, over 2366543.53 frames. ], batch size: 37, lr: 4.57e-03, grad_scale: 16.0 2023-05-16 04:50:33,694 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=165907.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 04:50:40,524 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.96 vs. limit=2.0 2023-05-16 04:50:42,176 INFO [finetune.py:992] (0/2) Epoch 6, batch 4750, loss[loss=0.1683, simple_loss=0.2535, pruned_loss=0.04148, over 12263.00 frames. ], tot_loss[loss=0.1754, simple_loss=0.2629, pruned_loss=0.04393, over 2365367.44 frames. ], batch size: 32, lr: 4.57e-03, grad_scale: 16.0 2023-05-16 04:50:56,969 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.949e+02 2.971e+02 3.330e+02 3.952e+02 9.689e+02, threshold=6.660e+02, percent-clipped=4.0 2023-05-16 04:51:03,845 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=3.00 vs. limit=5.0 2023-05-16 04:51:18,404 INFO [finetune.py:992] (0/2) Epoch 6, batch 4800, loss[loss=0.1478, simple_loss=0.229, pruned_loss=0.03334, over 12179.00 frames. ], tot_loss[loss=0.1746, simple_loss=0.2621, pruned_loss=0.0436, over 2367656.53 frames. ], batch size: 29, lr: 4.57e-03, grad_scale: 16.0 2023-05-16 04:51:40,188 INFO [checkpoint.py:75] (0/2) Saving checkpoint to pruned_transducer_stateless7/exp_giga_finetune/checkpoint-66000.pt 2023-05-16 04:51:48,254 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.76 vs. limit=2.0 2023-05-16 04:51:49,831 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.41 vs. limit=2.0 2023-05-16 04:51:56,981 INFO [finetune.py:992] (0/2) Epoch 6, batch 4850, loss[loss=0.1807, simple_loss=0.2652, pruned_loss=0.04808, over 12104.00 frames. ], tot_loss[loss=0.1744, simple_loss=0.2619, pruned_loss=0.04347, over 2374202.77 frames. ], batch size: 33, lr: 4.56e-03, grad_scale: 16.0 2023-05-16 04:52:11,253 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.665e+02 2.971e+02 3.475e+02 4.113e+02 7.039e+02, threshold=6.950e+02, percent-clipped=1.0 2023-05-16 04:52:22,104 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=166054.0, num_to_drop=1, layers_to_drop={1} 2023-05-16 04:52:32,234 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.7995, 2.9776, 4.4864, 4.9267, 3.1904, 2.9043, 3.0546, 2.1317], device='cuda:0'), covar=tensor([0.1342, 0.2752, 0.0510, 0.0345, 0.1019, 0.1914, 0.2499, 0.4107], device='cuda:0'), in_proj_covar=tensor([0.0273, 0.0372, 0.0262, 0.0290, 0.0255, 0.0285, 0.0355, 0.0358], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-16 04:52:32,634 INFO [finetune.py:992] (0/2) Epoch 6, batch 4900, loss[loss=0.1866, simple_loss=0.2631, pruned_loss=0.05503, over 12292.00 frames. ], tot_loss[loss=0.1737, simple_loss=0.2613, pruned_loss=0.04308, over 2383584.03 frames. ], batch size: 33, lr: 4.56e-03, grad_scale: 16.0 2023-05-16 04:52:48,422 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.2385, 2.3064, 3.4999, 4.1971, 3.7371, 4.0373, 3.6729, 2.8964], device='cuda:0'), covar=tensor([0.0029, 0.0395, 0.0128, 0.0034, 0.0102, 0.0065, 0.0115, 0.0311], device='cuda:0'), in_proj_covar=tensor([0.0080, 0.0118, 0.0098, 0.0074, 0.0096, 0.0109, 0.0088, 0.0134], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 04:52:59,864 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=166106.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 04:53:06,363 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=166115.0, num_to_drop=1, layers_to_drop={3} 2023-05-16 04:53:08,951 INFO [finetune.py:992] (0/2) Epoch 6, batch 4950, loss[loss=0.1594, simple_loss=0.2398, pruned_loss=0.03949, over 12275.00 frames. ], tot_loss[loss=0.1738, simple_loss=0.2613, pruned_loss=0.04309, over 2372971.24 frames. ], batch size: 28, lr: 4.56e-03, grad_scale: 16.0 2023-05-16 04:53:23,234 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.728e+02 2.719e+02 3.145e+02 3.634e+02 5.945e+02, threshold=6.289e+02, percent-clipped=0.0 2023-05-16 04:53:28,860 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=166147.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 04:53:37,396 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=166159.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 04:53:45,075 INFO [finetune.py:992] (0/2) Epoch 6, batch 5000, loss[loss=0.1783, simple_loss=0.2661, pruned_loss=0.04528, over 12253.00 frames. ], tot_loss[loss=0.1736, simple_loss=0.2612, pruned_loss=0.04303, over 2379443.02 frames. ], batch size: 32, lr: 4.56e-03, grad_scale: 16.0 2023-05-16 04:53:57,174 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.77 vs. limit=2.0 2023-05-16 04:54:12,670 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=166207.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 04:54:13,510 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=166208.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 04:54:21,065 INFO [finetune.py:992] (0/2) Epoch 6, batch 5050, loss[loss=0.1665, simple_loss=0.2512, pruned_loss=0.04094, over 12010.00 frames. ], tot_loss[loss=0.1737, simple_loss=0.2614, pruned_loss=0.04294, over 2377727.28 frames. ], batch size: 28, lr: 4.56e-03, grad_scale: 16.0 2023-05-16 04:54:28,543 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=4.51 vs. limit=5.0 2023-05-16 04:54:35,719 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.038e+02 2.948e+02 3.357e+02 4.167e+02 1.413e+03, threshold=6.713e+02, percent-clipped=3.0 2023-05-16 04:54:47,212 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=166255.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 04:54:56,974 INFO [finetune.py:992] (0/2) Epoch 6, batch 5100, loss[loss=0.2371, simple_loss=0.3202, pruned_loss=0.07705, over 10509.00 frames. ], tot_loss[loss=0.1748, simple_loss=0.2624, pruned_loss=0.04357, over 2367067.29 frames. ], batch size: 68, lr: 4.56e-03, grad_scale: 16.0 2023-05-16 04:55:04,768 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.5728, 3.7698, 3.3589, 3.3164, 3.0998, 2.8499, 3.7601, 2.5392], device='cuda:0'), covar=tensor([0.0329, 0.0092, 0.0152, 0.0163, 0.0324, 0.0338, 0.0102, 0.0345], device='cuda:0'), in_proj_covar=tensor([0.0189, 0.0155, 0.0152, 0.0181, 0.0199, 0.0192, 0.0161, 0.0196], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 04:55:23,879 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=166307.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 04:55:32,838 INFO [finetune.py:992] (0/2) Epoch 6, batch 5150, loss[loss=0.1721, simple_loss=0.2619, pruned_loss=0.04119, over 12319.00 frames. ], tot_loss[loss=0.1745, simple_loss=0.2617, pruned_loss=0.04361, over 2367233.47 frames. ], batch size: 34, lr: 4.56e-03, grad_scale: 16.0 2023-05-16 04:55:44,908 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.3608, 4.0976, 4.1537, 4.2488, 4.0940, 4.3012, 4.1512, 2.5695], device='cuda:0'), covar=tensor([0.0085, 0.0074, 0.0107, 0.0078, 0.0057, 0.0106, 0.0108, 0.0688], device='cuda:0'), in_proj_covar=tensor([0.0065, 0.0074, 0.0078, 0.0071, 0.0059, 0.0088, 0.0077, 0.0092], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 04:55:46,811 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.176e+02 2.917e+02 3.425e+02 4.057e+02 6.546e+02, threshold=6.849e+02, percent-clipped=0.0 2023-05-16 04:56:08,131 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=166368.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 04:56:08,617 INFO [finetune.py:992] (0/2) Epoch 6, batch 5200, loss[loss=0.1527, simple_loss=0.2416, pruned_loss=0.03192, over 12289.00 frames. ], tot_loss[loss=0.1748, simple_loss=0.2621, pruned_loss=0.04376, over 2360486.73 frames. ], batch size: 33, lr: 4.56e-03, grad_scale: 8.0 2023-05-16 04:56:24,384 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.1862, 2.0999, 2.7351, 3.1570, 2.1769, 3.2491, 3.1878, 3.2836], device='cuda:0'), covar=tensor([0.0154, 0.1049, 0.0406, 0.0179, 0.1023, 0.0275, 0.0261, 0.0132], device='cuda:0'), in_proj_covar=tensor([0.0116, 0.0201, 0.0183, 0.0115, 0.0187, 0.0175, 0.0170, 0.0121], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 04:56:35,401 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=166406.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 04:56:38,153 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=166410.0, num_to_drop=1, layers_to_drop={0} 2023-05-16 04:56:44,235 INFO [finetune.py:992] (0/2) Epoch 6, batch 5250, loss[loss=0.1778, simple_loss=0.2695, pruned_loss=0.04303, over 12363.00 frames. ], tot_loss[loss=0.1735, simple_loss=0.2606, pruned_loss=0.04325, over 2363945.08 frames. ], batch size: 35, lr: 4.56e-03, grad_scale: 8.0 2023-05-16 04:56:59,458 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.091e+02 2.900e+02 3.310e+02 3.936e+02 6.741e+02, threshold=6.620e+02, percent-clipped=0.0 2023-05-16 04:57:10,147 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=166454.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 04:57:13,877 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=166459.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 04:57:20,096 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.9310, 4.5295, 4.6275, 4.6482, 4.5192, 4.8035, 4.6391, 2.6604], device='cuda:0'), covar=tensor([0.0099, 0.0074, 0.0115, 0.0083, 0.0056, 0.0104, 0.0088, 0.0783], device='cuda:0'), in_proj_covar=tensor([0.0065, 0.0074, 0.0077, 0.0070, 0.0058, 0.0087, 0.0076, 0.0091], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 04:57:20,624 INFO [finetune.py:992] (0/2) Epoch 6, batch 5300, loss[loss=0.1866, simple_loss=0.274, pruned_loss=0.04966, over 12139.00 frames. ], tot_loss[loss=0.1736, simple_loss=0.2608, pruned_loss=0.04314, over 2373509.62 frames. ], batch size: 39, lr: 4.56e-03, grad_scale: 8.0 2023-05-16 04:57:29,426 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=166481.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 04:57:44,821 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=166503.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 04:57:47,428 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=166507.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 04:57:49,063 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=166509.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 04:57:56,709 INFO [finetune.py:992] (0/2) Epoch 6, batch 5350, loss[loss=0.1621, simple_loss=0.2398, pruned_loss=0.04224, over 12356.00 frames. ], tot_loss[loss=0.1735, simple_loss=0.2607, pruned_loss=0.04313, over 2372482.08 frames. ], batch size: 30, lr: 4.56e-03, grad_scale: 8.0 2023-05-16 04:58:05,513 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=166531.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 04:58:11,567 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.966e+02 2.816e+02 3.217e+02 3.761e+02 6.495e+02, threshold=6.434e+02, percent-clipped=0.0 2023-05-16 04:58:13,170 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=166542.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 04:58:32,287 INFO [finetune.py:992] (0/2) Epoch 6, batch 5400, loss[loss=0.1829, simple_loss=0.2693, pruned_loss=0.04827, over 12366.00 frames. ], tot_loss[loss=0.1734, simple_loss=0.2607, pruned_loss=0.04311, over 2367492.41 frames. ], batch size: 38, lr: 4.56e-03, grad_scale: 8.0 2023-05-16 04:58:32,513 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.8478, 2.3636, 3.3011, 2.8388, 3.2036, 2.9814, 2.1814, 3.2592], device='cuda:0'), covar=tensor([0.0131, 0.0314, 0.0156, 0.0214, 0.0135, 0.0177, 0.0357, 0.0123], device='cuda:0'), in_proj_covar=tensor([0.0169, 0.0198, 0.0181, 0.0177, 0.0203, 0.0155, 0.0189, 0.0171], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 04:58:33,210 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=166570.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 04:58:49,413 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=166592.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 04:59:08,572 INFO [finetune.py:992] (0/2) Epoch 6, batch 5450, loss[loss=0.1481, simple_loss=0.2337, pruned_loss=0.03129, over 11990.00 frames. ], tot_loss[loss=0.174, simple_loss=0.2614, pruned_loss=0.04336, over 2366374.09 frames. ], batch size: 28, lr: 4.56e-03, grad_scale: 8.0 2023-05-16 04:59:23,339 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.010e+02 2.821e+02 3.503e+02 4.549e+02 8.622e+02, threshold=7.005e+02, percent-clipped=5.0 2023-05-16 04:59:40,643 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=166663.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 04:59:44,769 INFO [finetune.py:992] (0/2) Epoch 6, batch 5500, loss[loss=0.1714, simple_loss=0.2628, pruned_loss=0.04002, over 12378.00 frames. ], tot_loss[loss=0.1744, simple_loss=0.262, pruned_loss=0.04341, over 2367191.54 frames. ], batch size: 35, lr: 4.56e-03, grad_scale: 8.0 2023-05-16 04:59:45,647 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.7800, 3.0940, 3.4518, 4.5408, 2.5356, 4.6958, 4.7030, 4.8709], device='cuda:0'), covar=tensor([0.0093, 0.0877, 0.0373, 0.0121, 0.1168, 0.0165, 0.0119, 0.0067], device='cuda:0'), in_proj_covar=tensor([0.0117, 0.0201, 0.0184, 0.0116, 0.0187, 0.0175, 0.0171, 0.0121], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 05:00:13,873 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=166710.0, num_to_drop=1, layers_to_drop={2} 2023-05-16 05:00:20,103 INFO [finetune.py:992] (0/2) Epoch 6, batch 5550, loss[loss=0.169, simple_loss=0.2615, pruned_loss=0.03822, over 12275.00 frames. ], tot_loss[loss=0.1744, simple_loss=0.2624, pruned_loss=0.04326, over 2365885.51 frames. ], batch size: 37, lr: 4.56e-03, grad_scale: 8.0 2023-05-16 05:00:25,504 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=3.19 vs. limit=5.0 2023-05-16 05:00:32,417 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=166735.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 05:00:35,570 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.829e+02 2.867e+02 3.413e+02 3.972e+02 6.921e+02, threshold=6.827e+02, percent-clipped=0.0 2023-05-16 05:00:48,530 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=166758.0, num_to_drop=1, layers_to_drop={0} 2023-05-16 05:00:52,838 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=166764.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 05:00:56,212 INFO [finetune.py:992] (0/2) Epoch 6, batch 5600, loss[loss=0.1786, simple_loss=0.2731, pruned_loss=0.04203, over 12021.00 frames. ], tot_loss[loss=0.1751, simple_loss=0.2632, pruned_loss=0.04349, over 2365807.66 frames. ], batch size: 40, lr: 4.56e-03, grad_scale: 8.0 2023-05-16 05:01:15,823 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=166796.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 05:01:16,473 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.9859, 5.9794, 5.8070, 5.2776, 5.1876, 5.9545, 5.4928, 5.3185], device='cuda:0'), covar=tensor([0.0664, 0.0845, 0.0651, 0.1571, 0.0595, 0.0637, 0.1392, 0.1048], device='cuda:0'), in_proj_covar=tensor([0.0588, 0.0516, 0.0493, 0.0608, 0.0395, 0.0691, 0.0744, 0.0543], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0004, 0.0004, 0.0003], device='cuda:0') 2023-05-16 05:01:21,647 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=166803.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 05:01:22,356 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([6.1735, 6.0851, 5.9663, 5.4383, 5.2305, 6.0923, 5.6921, 5.4254], device='cuda:0'), covar=tensor([0.0594, 0.0968, 0.0611, 0.1589, 0.0609, 0.0628, 0.1236, 0.1072], device='cuda:0'), in_proj_covar=tensor([0.0589, 0.0515, 0.0493, 0.0608, 0.0395, 0.0691, 0.0744, 0.0543], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0004, 0.0004, 0.0003], device='cuda:0') 2023-05-16 05:01:32,913 INFO [finetune.py:992] (0/2) Epoch 6, batch 5650, loss[loss=0.1439, simple_loss=0.2283, pruned_loss=0.02978, over 11997.00 frames. ], tot_loss[loss=0.1741, simple_loss=0.2623, pruned_loss=0.04298, over 2371077.92 frames. ], batch size: 28, lr: 4.56e-03, grad_scale: 8.0 2023-05-16 05:01:37,508 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=166825.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 05:01:45,919 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=166837.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 05:01:47,693 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.51 vs. limit=2.0 2023-05-16 05:01:48,042 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.930e+02 2.719e+02 3.399e+02 4.144e+02 8.049e+02, threshold=6.798e+02, percent-clipped=1.0 2023-05-16 05:01:55,883 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=166851.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 05:01:59,137 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.41 vs. limit=2.0 2023-05-16 05:02:05,824 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=166865.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 05:02:08,596 INFO [finetune.py:992] (0/2) Epoch 6, batch 5700, loss[loss=0.1676, simple_loss=0.2612, pruned_loss=0.037, over 12356.00 frames. ], tot_loss[loss=0.1737, simple_loss=0.2616, pruned_loss=0.04284, over 2373298.00 frames. ], batch size: 36, lr: 4.56e-03, grad_scale: 8.0 2023-05-16 05:02:21,923 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=166887.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 05:02:33,599 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.4945, 4.6859, 4.3055, 5.0772, 4.6885, 3.0900, 4.4563, 3.0309], device='cuda:0'), covar=tensor([0.0646, 0.0781, 0.1204, 0.0381, 0.0901, 0.1383, 0.0838, 0.3195], device='cuda:0'), in_proj_covar=tensor([0.0292, 0.0365, 0.0343, 0.0267, 0.0354, 0.0261, 0.0328, 0.0353], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 05:02:39,321 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.1305, 4.6304, 2.8262, 2.4653, 3.8941, 2.3850, 3.8871, 3.1911], device='cuda:0'), covar=tensor([0.0666, 0.0412, 0.0988, 0.1426, 0.0248, 0.1248, 0.0434, 0.0701], device='cuda:0'), in_proj_covar=tensor([0.0183, 0.0245, 0.0171, 0.0193, 0.0135, 0.0177, 0.0190, 0.0171], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 05:02:44,683 INFO [finetune.py:992] (0/2) Epoch 6, batch 5750, loss[loss=0.1777, simple_loss=0.2691, pruned_loss=0.04315, over 12150.00 frames. ], tot_loss[loss=0.1747, simple_loss=0.2628, pruned_loss=0.04329, over 2374739.50 frames. ], batch size: 36, lr: 4.56e-03, grad_scale: 8.0 2023-05-16 05:02:47,709 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([6.1378, 6.0843, 5.9284, 5.4214, 5.2751, 6.0254, 5.6441, 5.4367], device='cuda:0'), covar=tensor([0.0677, 0.0848, 0.0639, 0.1538, 0.0657, 0.0709, 0.1480, 0.1069], device='cuda:0'), in_proj_covar=tensor([0.0596, 0.0517, 0.0496, 0.0613, 0.0399, 0.0695, 0.0749, 0.0548], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0004, 0.0004, 0.0003], device='cuda:0') 2023-05-16 05:02:51,830 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=166929.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 05:02:54,410 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=4.84 vs. limit=5.0 2023-05-16 05:02:59,395 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.867e+02 2.780e+02 3.132e+02 3.989e+02 8.117e+02, threshold=6.265e+02, percent-clipped=1.0 2023-05-16 05:03:09,930 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 2023-05-16 05:03:13,229 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.9471, 3.4825, 5.1421, 2.5163, 2.9000, 3.8638, 3.4823, 3.8885], device='cuda:0'), covar=tensor([0.0287, 0.1023, 0.0285, 0.1241, 0.1843, 0.1306, 0.1160, 0.1118], device='cuda:0'), in_proj_covar=tensor([0.0229, 0.0229, 0.0229, 0.0176, 0.0231, 0.0277, 0.0219, 0.0248], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:0') 2023-05-16 05:03:16,740 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=166963.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 05:03:20,774 INFO [finetune.py:992] (0/2) Epoch 6, batch 5800, loss[loss=0.2071, simple_loss=0.296, pruned_loss=0.05904, over 11885.00 frames. ], tot_loss[loss=0.1741, simple_loss=0.2618, pruned_loss=0.04314, over 2374879.83 frames. ], batch size: 44, lr: 4.55e-03, grad_scale: 8.0 2023-05-16 05:03:35,741 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=166990.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 05:03:41,549 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.0484, 4.6362, 5.0097, 4.3635, 4.6178, 4.4596, 5.0213, 4.6450], device='cuda:0'), covar=tensor([0.0231, 0.0344, 0.0243, 0.0245, 0.0345, 0.0299, 0.0212, 0.0363], device='cuda:0'), in_proj_covar=tensor([0.0243, 0.0252, 0.0269, 0.0244, 0.0243, 0.0243, 0.0221, 0.0205], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 05:03:50,909 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=167011.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 05:03:56,460 INFO [finetune.py:992] (0/2) Epoch 6, batch 5850, loss[loss=0.1782, simple_loss=0.2784, pruned_loss=0.03902, over 12149.00 frames. ], tot_loss[loss=0.1744, simple_loss=0.2622, pruned_loss=0.0433, over 2373245.04 frames. ], batch size: 34, lr: 4.55e-03, grad_scale: 8.0 2023-05-16 05:04:08,628 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.0025, 2.2994, 2.3623, 2.2987, 2.1476, 1.9016, 2.2526, 1.7943], device='cuda:0'), covar=tensor([0.0258, 0.0145, 0.0162, 0.0184, 0.0284, 0.0215, 0.0163, 0.0348], device='cuda:0'), in_proj_covar=tensor([0.0187, 0.0154, 0.0151, 0.0182, 0.0197, 0.0192, 0.0159, 0.0196], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 05:04:11,948 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.879e+02 2.923e+02 3.385e+02 4.055e+02 7.861e+02, threshold=6.771e+02, percent-clipped=2.0 2023-05-16 05:04:32,515 INFO [finetune.py:992] (0/2) Epoch 6, batch 5900, loss[loss=0.1427, simple_loss=0.2257, pruned_loss=0.02985, over 12290.00 frames. ], tot_loss[loss=0.1744, simple_loss=0.2622, pruned_loss=0.04331, over 2373978.82 frames. ], batch size: 28, lr: 4.55e-03, grad_scale: 8.0 2023-05-16 05:04:33,036 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.87 vs. limit=2.0 2023-05-16 05:04:48,331 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=167091.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 05:04:55,448 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=167100.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 05:04:58,239 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=167104.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 05:05:08,542 INFO [finetune.py:992] (0/2) Epoch 6, batch 5950, loss[loss=0.1691, simple_loss=0.2559, pruned_loss=0.04119, over 12260.00 frames. ], tot_loss[loss=0.1765, simple_loss=0.2641, pruned_loss=0.04446, over 2362370.85 frames. ], batch size: 32, lr: 4.55e-03, grad_scale: 8.0 2023-05-16 05:05:09,284 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=167120.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 05:05:13,134 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.3822, 4.7744, 4.1938, 5.0516, 4.6349, 2.7732, 4.2663, 3.0539], device='cuda:0'), covar=tensor([0.0633, 0.0640, 0.1170, 0.0386, 0.0889, 0.1635, 0.0935, 0.2960], device='cuda:0'), in_proj_covar=tensor([0.0293, 0.0366, 0.0345, 0.0268, 0.0356, 0.0263, 0.0330, 0.0355], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 05:05:21,662 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=167137.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 05:05:23,600 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.785e+02 2.888e+02 3.544e+02 4.208e+02 1.110e+03, threshold=7.088e+02, percent-clipped=2.0 2023-05-16 05:05:38,538 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=167161.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 05:05:42,004 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=167165.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 05:05:42,048 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=167165.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 05:05:44,723 INFO [finetune.py:992] (0/2) Epoch 6, batch 6000, loss[loss=0.1744, simple_loss=0.2721, pruned_loss=0.03832, over 12358.00 frames. ], tot_loss[loss=0.1755, simple_loss=0.263, pruned_loss=0.04402, over 2368762.60 frames. ], batch size: 35, lr: 4.55e-03, grad_scale: 8.0 2023-05-16 05:05:44,723 INFO [finetune.py:1017] (0/2) Computing validation loss 2023-05-16 05:06:03,175 INFO [finetune.py:1026] (0/2) Epoch 6, validation: loss=0.3235, simple_loss=0.4002, pruned_loss=0.1234, over 1020973.00 frames. 2023-05-16 05:06:03,176 INFO [finetune.py:1027] (0/2) Maximum memory allocated so far is 12745MB 2023-05-16 05:06:13,787 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.0397, 4.6629, 4.8332, 4.8363, 4.5730, 4.8328, 4.7689, 2.7338], device='cuda:0'), covar=tensor([0.0093, 0.0067, 0.0074, 0.0062, 0.0060, 0.0088, 0.0094, 0.0657], device='cuda:0'), in_proj_covar=tensor([0.0066, 0.0075, 0.0078, 0.0071, 0.0059, 0.0089, 0.0078, 0.0092], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 05:06:14,360 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=167185.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 05:06:15,784 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=167187.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 05:06:35,290 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=167213.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 05:06:39,498 INFO [finetune.py:992] (0/2) Epoch 6, batch 6050, loss[loss=0.1425, simple_loss=0.2296, pruned_loss=0.02769, over 12344.00 frames. ], tot_loss[loss=0.1751, simple_loss=0.2627, pruned_loss=0.04376, over 2370647.33 frames. ], batch size: 31, lr: 4.55e-03, grad_scale: 8.0 2023-05-16 05:06:50,749 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=167235.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 05:06:54,243 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.800e+02 2.751e+02 3.329e+02 3.847e+02 8.710e+02, threshold=6.657e+02, percent-clipped=1.0 2023-05-16 05:07:15,452 INFO [finetune.py:992] (0/2) Epoch 6, batch 6100, loss[loss=0.1559, simple_loss=0.2442, pruned_loss=0.0338, over 12338.00 frames. ], tot_loss[loss=0.1743, simple_loss=0.2622, pruned_loss=0.04322, over 2375414.07 frames. ], batch size: 31, lr: 4.55e-03, grad_scale: 8.0 2023-05-16 05:07:26,570 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=167285.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 05:07:34,671 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=2.05 vs. limit=2.0 2023-05-16 05:07:50,740 INFO [finetune.py:992] (0/2) Epoch 6, batch 6150, loss[loss=0.1687, simple_loss=0.2569, pruned_loss=0.04031, over 12120.00 frames. ], tot_loss[loss=0.1752, simple_loss=0.2628, pruned_loss=0.04378, over 2370107.63 frames. ], batch size: 39, lr: 4.55e-03, grad_scale: 8.0 2023-05-16 05:08:02,498 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2023-05-16 05:08:06,376 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.990e+02 2.796e+02 3.234e+02 3.972e+02 7.631e+02, threshold=6.469e+02, percent-clipped=1.0 2023-05-16 05:08:26,729 INFO [finetune.py:992] (0/2) Epoch 6, batch 6200, loss[loss=0.1467, simple_loss=0.2308, pruned_loss=0.03127, over 12168.00 frames. ], tot_loss[loss=0.1756, simple_loss=0.2628, pruned_loss=0.04415, over 2363554.23 frames. ], batch size: 29, lr: 4.55e-03, grad_scale: 8.0 2023-05-16 05:08:42,486 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=167391.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 05:09:03,327 INFO [finetune.py:992] (0/2) Epoch 6, batch 6250, loss[loss=0.1693, simple_loss=0.2568, pruned_loss=0.04087, over 12352.00 frames. ], tot_loss[loss=0.1747, simple_loss=0.2619, pruned_loss=0.04373, over 2367120.73 frames. ], batch size: 36, lr: 4.55e-03, grad_scale: 8.0 2023-05-16 05:09:04,216 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=167420.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 05:09:12,829 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=167432.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 05:09:17,731 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=167439.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 05:09:18,325 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.096e+02 2.911e+02 3.395e+02 4.123e+02 7.089e+02, threshold=6.790e+02, percent-clipped=2.0 2023-05-16 05:09:29,671 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=167456.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 05:09:32,535 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=167460.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 05:09:38,153 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=167468.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 05:09:38,831 INFO [finetune.py:992] (0/2) Epoch 6, batch 6300, loss[loss=0.1557, simple_loss=0.2506, pruned_loss=0.03039, over 12285.00 frames. ], tot_loss[loss=0.1741, simple_loss=0.262, pruned_loss=0.04312, over 2367331.41 frames. ], batch size: 37, lr: 4.55e-03, grad_scale: 8.0 2023-05-16 05:09:56,778 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=167493.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 05:10:15,220 INFO [finetune.py:992] (0/2) Epoch 6, batch 6350, loss[loss=0.1731, simple_loss=0.2541, pruned_loss=0.04604, over 12138.00 frames. ], tot_loss[loss=0.1735, simple_loss=0.2614, pruned_loss=0.04285, over 2372247.27 frames. ], batch size: 30, lr: 4.55e-03, grad_scale: 8.0 2023-05-16 05:10:21,665 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.9519, 6.0099, 5.6806, 5.2171, 5.1452, 5.8665, 5.4181, 5.1744], device='cuda:0'), covar=tensor([0.0692, 0.0676, 0.0683, 0.1608, 0.0741, 0.0734, 0.1593, 0.1099], device='cuda:0'), in_proj_covar=tensor([0.0583, 0.0508, 0.0490, 0.0604, 0.0393, 0.0685, 0.0740, 0.0540], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0004, 0.0004, 0.0003], device='cuda:0') 2023-05-16 05:10:29,930 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.740e+02 2.792e+02 3.317e+02 4.286e+02 2.773e+03, threshold=6.635e+02, percent-clipped=3.0 2023-05-16 05:10:45,451 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.59 vs. limit=2.0 2023-05-16 05:10:51,373 INFO [finetune.py:992] (0/2) Epoch 6, batch 6400, loss[loss=0.1952, simple_loss=0.2816, pruned_loss=0.05438, over 12373.00 frames. ], tot_loss[loss=0.1732, simple_loss=0.2609, pruned_loss=0.04276, over 2380352.34 frames. ], batch size: 38, lr: 4.55e-03, grad_scale: 8.0 2023-05-16 05:11:03,048 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=167585.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 05:11:27,442 INFO [finetune.py:992] (0/2) Epoch 6, batch 6450, loss[loss=0.173, simple_loss=0.2661, pruned_loss=0.03998, over 12146.00 frames. ], tot_loss[loss=0.1737, simple_loss=0.2615, pruned_loss=0.04294, over 2377383.42 frames. ], batch size: 39, lr: 4.55e-03, grad_scale: 8.0 2023-05-16 05:11:37,441 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=167633.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 05:11:37,801 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.41 vs. limit=2.0 2023-05-16 05:11:43,111 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.863e+02 2.692e+02 3.176e+02 3.788e+02 6.386e+02, threshold=6.352e+02, percent-clipped=0.0 2023-05-16 05:12:03,562 INFO [finetune.py:992] (0/2) Epoch 6, batch 6500, loss[loss=0.1308, simple_loss=0.2198, pruned_loss=0.02095, over 12154.00 frames. ], tot_loss[loss=0.1734, simple_loss=0.2612, pruned_loss=0.04277, over 2371934.13 frames. ], batch size: 29, lr: 4.55e-03, grad_scale: 8.0 2023-05-16 05:12:11,416 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.0898, 4.1817, 4.0503, 4.5085, 2.8063, 3.8454, 2.6455, 4.0901], device='cuda:0'), covar=tensor([0.1630, 0.0575, 0.0835, 0.0526, 0.1108, 0.0596, 0.1680, 0.1095], device='cuda:0'), in_proj_covar=tensor([0.0223, 0.0259, 0.0290, 0.0348, 0.0232, 0.0234, 0.0252, 0.0376], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 05:12:25,063 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.8961, 3.4365, 5.1980, 2.8191, 2.8209, 3.9213, 3.4594, 3.8655], device='cuda:0'), covar=tensor([0.0418, 0.1100, 0.0254, 0.1145, 0.1888, 0.1366, 0.1187, 0.1094], device='cuda:0'), in_proj_covar=tensor([0.0229, 0.0226, 0.0229, 0.0176, 0.0230, 0.0276, 0.0218, 0.0247], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:0') 2023-05-16 05:12:39,433 INFO [finetune.py:992] (0/2) Epoch 6, batch 6550, loss[loss=0.1913, simple_loss=0.282, pruned_loss=0.05034, over 12018.00 frames. ], tot_loss[loss=0.1753, simple_loss=0.2629, pruned_loss=0.0439, over 2353647.48 frames. ], batch size: 40, lr: 4.55e-03, grad_scale: 8.0 2023-05-16 05:12:39,661 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=167719.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 05:12:54,559 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.510e+02 2.847e+02 3.210e+02 3.816e+02 6.751e+02, threshold=6.421e+02, percent-clipped=3.0 2023-05-16 05:13:05,949 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=167756.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 05:13:08,724 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=167760.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 05:13:10,956 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.1920, 4.3431, 3.9839, 4.8054, 4.3996, 2.7469, 4.1212, 3.0021], device='cuda:0'), covar=tensor([0.0797, 0.0899, 0.1302, 0.0386, 0.1100, 0.1628, 0.1043, 0.3222], device='cuda:0'), in_proj_covar=tensor([0.0296, 0.0372, 0.0349, 0.0271, 0.0359, 0.0264, 0.0333, 0.0357], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 05:13:13,812 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=167767.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 05:13:15,072 INFO [finetune.py:992] (0/2) Epoch 6, batch 6600, loss[loss=0.1666, simple_loss=0.2623, pruned_loss=0.0354, over 12364.00 frames. ], tot_loss[loss=0.1751, simple_loss=0.2629, pruned_loss=0.04363, over 2363693.14 frames. ], batch size: 35, lr: 4.55e-03, grad_scale: 8.0 2023-05-16 05:13:23,917 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=167780.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 05:13:29,533 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=167788.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 05:13:41,098 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=167804.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 05:13:43,962 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=167808.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 05:13:51,670 INFO [finetune.py:992] (0/2) Epoch 6, batch 6650, loss[loss=0.1856, simple_loss=0.2786, pruned_loss=0.04628, over 11514.00 frames. ], tot_loss[loss=0.1745, simple_loss=0.2624, pruned_loss=0.04332, over 2368386.72 frames. ], batch size: 48, lr: 4.55e-03, grad_scale: 8.0 2023-05-16 05:13:58,716 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.5383, 5.0098, 5.4661, 4.7871, 5.0498, 4.9188, 5.5006, 5.1155], device='cuda:0'), covar=tensor([0.0215, 0.0330, 0.0242, 0.0240, 0.0319, 0.0269, 0.0203, 0.0218], device='cuda:0'), in_proj_covar=tensor([0.0245, 0.0253, 0.0272, 0.0247, 0.0244, 0.0244, 0.0222, 0.0206], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 05:13:58,779 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=167828.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 05:14:07,255 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.971e+02 2.793e+02 3.244e+02 3.782e+02 8.020e+02, threshold=6.487e+02, percent-clipped=1.0 2023-05-16 05:14:27,892 INFO [finetune.py:992] (0/2) Epoch 6, batch 6700, loss[loss=0.1632, simple_loss=0.2445, pruned_loss=0.04095, over 12185.00 frames. ], tot_loss[loss=0.1743, simple_loss=0.2621, pruned_loss=0.04321, over 2367047.57 frames. ], batch size: 29, lr: 4.55e-03, grad_scale: 8.0 2023-05-16 05:14:50,797 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.4268, 4.7151, 4.1124, 5.0635, 4.5627, 2.7802, 4.1925, 3.1689], device='cuda:0'), covar=tensor([0.0668, 0.0731, 0.1281, 0.0333, 0.0902, 0.1635, 0.1055, 0.2772], device='cuda:0'), in_proj_covar=tensor([0.0296, 0.0371, 0.0348, 0.0271, 0.0359, 0.0263, 0.0332, 0.0356], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 05:15:03,834 INFO [finetune.py:992] (0/2) Epoch 6, batch 6750, loss[loss=0.1844, simple_loss=0.2742, pruned_loss=0.0473, over 11664.00 frames. ], tot_loss[loss=0.1744, simple_loss=0.2617, pruned_loss=0.04358, over 2367828.68 frames. ], batch size: 48, lr: 4.54e-03, grad_scale: 8.0 2023-05-16 05:15:18,834 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.022e+02 2.940e+02 3.475e+02 4.006e+02 8.248e+02, threshold=6.951e+02, percent-clipped=1.0 2023-05-16 05:15:39,986 INFO [finetune.py:992] (0/2) Epoch 6, batch 6800, loss[loss=0.1722, simple_loss=0.27, pruned_loss=0.03721, over 12355.00 frames. ], tot_loss[loss=0.1751, simple_loss=0.2625, pruned_loss=0.04385, over 2367619.45 frames. ], batch size: 36, lr: 4.54e-03, grad_scale: 8.0 2023-05-16 05:15:57,212 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.9078, 5.9158, 5.6465, 5.2534, 5.1439, 5.8218, 5.3945, 5.1931], device='cuda:0'), covar=tensor([0.0701, 0.0858, 0.0698, 0.1505, 0.0710, 0.0710, 0.1529, 0.1026], device='cuda:0'), in_proj_covar=tensor([0.0584, 0.0507, 0.0490, 0.0600, 0.0390, 0.0683, 0.0734, 0.0540], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0004, 0.0004, 0.0003], device='cuda:0') 2023-05-16 05:15:59,277 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.4388, 4.9959, 5.4151, 4.6668, 5.0287, 4.8165, 5.4607, 5.1078], device='cuda:0'), covar=tensor([0.0228, 0.0290, 0.0223, 0.0252, 0.0300, 0.0271, 0.0170, 0.0223], device='cuda:0'), in_proj_covar=tensor([0.0243, 0.0251, 0.0269, 0.0246, 0.0243, 0.0244, 0.0220, 0.0204], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 05:16:02,256 INFO [checkpoint.py:75] (0/2) Saving checkpoint to pruned_transducer_stateless7/exp_giga_finetune/checkpoint-68000.pt 2023-05-16 05:16:18,599 INFO [finetune.py:992] (0/2) Epoch 6, batch 6850, loss[loss=0.209, simple_loss=0.3039, pruned_loss=0.05704, over 11584.00 frames. ], tot_loss[loss=0.1756, simple_loss=0.2629, pruned_loss=0.04415, over 2362954.18 frames. ], batch size: 48, lr: 4.54e-03, grad_scale: 8.0 2023-05-16 05:16:30,144 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=168035.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 05:16:33,493 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.737e+02 2.783e+02 3.526e+02 3.993e+02 7.031e+02, threshold=7.052e+02, percent-clipped=1.0 2023-05-16 05:16:45,871 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.8696, 4.8014, 4.7375, 4.7457, 4.4181, 4.8073, 4.8411, 5.1232], device='cuda:0'), covar=tensor([0.0242, 0.0158, 0.0189, 0.0329, 0.0746, 0.0286, 0.0182, 0.0153], device='cuda:0'), in_proj_covar=tensor([0.0183, 0.0184, 0.0184, 0.0232, 0.0233, 0.0205, 0.0167, 0.0214], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0005, 0.0004, 0.0005, 0.0004, 0.0005], device='cuda:0') 2023-05-16 05:16:54,709 INFO [finetune.py:992] (0/2) Epoch 6, batch 6900, loss[loss=0.1651, simple_loss=0.2469, pruned_loss=0.04163, over 12380.00 frames. ], tot_loss[loss=0.1759, simple_loss=0.2636, pruned_loss=0.04406, over 2373741.77 frames. ], batch size: 30, lr: 4.54e-03, grad_scale: 8.0 2023-05-16 05:16:59,107 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=168075.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 05:17:08,330 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=168088.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 05:17:13,982 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=168096.0, num_to_drop=1, layers_to_drop={0} 2023-05-16 05:17:30,856 INFO [finetune.py:992] (0/2) Epoch 6, batch 6950, loss[loss=0.1555, simple_loss=0.2451, pruned_loss=0.03302, over 12170.00 frames. ], tot_loss[loss=0.1755, simple_loss=0.2631, pruned_loss=0.044, over 2369063.16 frames. ], batch size: 31, lr: 4.54e-03, grad_scale: 8.0 2023-05-16 05:17:33,770 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=168123.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 05:17:39,702 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.0427, 3.3787, 5.2470, 2.7093, 2.7523, 3.8310, 3.3667, 3.7536], device='cuda:0'), covar=tensor([0.0317, 0.1086, 0.0232, 0.1257, 0.1983, 0.1502, 0.1266, 0.1210], device='cuda:0'), in_proj_covar=tensor([0.0229, 0.0227, 0.0230, 0.0177, 0.0231, 0.0278, 0.0219, 0.0249], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:0') 2023-05-16 05:17:41,755 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.3239, 6.2569, 5.7967, 5.7979, 6.2254, 5.5085, 5.8397, 5.8073], device='cuda:0'), covar=tensor([0.1372, 0.0804, 0.1021, 0.1594, 0.0889, 0.2040, 0.1678, 0.0903], device='cuda:0'), in_proj_covar=tensor([0.0346, 0.0460, 0.0367, 0.0419, 0.0441, 0.0423, 0.0382, 0.0362], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 05:17:43,188 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=168136.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 05:17:46,045 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.837e+02 2.788e+02 3.289e+02 3.840e+02 6.856e+02, threshold=6.579e+02, percent-clipped=0.0 2023-05-16 05:18:06,438 INFO [finetune.py:992] (0/2) Epoch 6, batch 7000, loss[loss=0.1794, simple_loss=0.2677, pruned_loss=0.04555, over 12358.00 frames. ], tot_loss[loss=0.1739, simple_loss=0.2613, pruned_loss=0.04329, over 2376168.73 frames. ], batch size: 35, lr: 4.54e-03, grad_scale: 8.0 2023-05-16 05:18:42,876 INFO [finetune.py:992] (0/2) Epoch 6, batch 7050, loss[loss=0.1542, simple_loss=0.2416, pruned_loss=0.03337, over 12028.00 frames. ], tot_loss[loss=0.1732, simple_loss=0.2606, pruned_loss=0.04285, over 2381875.75 frames. ], batch size: 31, lr: 4.54e-03, grad_scale: 8.0 2023-05-16 05:18:52,807 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=168233.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 05:18:57,556 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.930e+02 2.650e+02 3.194e+02 3.802e+02 6.150e+02, threshold=6.388e+02, percent-clipped=0.0 2023-05-16 05:19:08,499 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=4.67 vs. limit=5.0 2023-05-16 05:19:12,475 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.9277, 4.8734, 4.7731, 4.8486, 4.3796, 4.9207, 4.8798, 5.1265], device='cuda:0'), covar=tensor([0.0258, 0.0143, 0.0190, 0.0254, 0.0854, 0.0271, 0.0166, 0.0169], device='cuda:0'), in_proj_covar=tensor([0.0184, 0.0187, 0.0186, 0.0236, 0.0238, 0.0207, 0.0169, 0.0216], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0005, 0.0005, 0.0005, 0.0004, 0.0005], device='cuda:0') 2023-05-16 05:19:18,616 INFO [finetune.py:992] (0/2) Epoch 6, batch 7100, loss[loss=0.1545, simple_loss=0.2409, pruned_loss=0.03405, over 12337.00 frames. ], tot_loss[loss=0.1729, simple_loss=0.2602, pruned_loss=0.04281, over 2387139.85 frames. ], batch size: 30, lr: 4.54e-03, grad_scale: 8.0 2023-05-16 05:19:21,796 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.92 vs. limit=2.0 2023-05-16 05:19:36,410 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=168294.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 05:19:39,025 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.2093, 5.9824, 5.5988, 5.5663, 6.0859, 5.4126, 5.6845, 5.6226], device='cuda:0'), covar=tensor([0.1337, 0.0846, 0.0825, 0.1746, 0.0833, 0.1996, 0.1399, 0.0866], device='cuda:0'), in_proj_covar=tensor([0.0344, 0.0459, 0.0365, 0.0418, 0.0440, 0.0424, 0.0379, 0.0361], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 05:19:48,417 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 2023-05-16 05:19:53,531 INFO [finetune.py:992] (0/2) Epoch 6, batch 7150, loss[loss=0.1866, simple_loss=0.2753, pruned_loss=0.04895, over 11704.00 frames. ], tot_loss[loss=0.1735, simple_loss=0.2608, pruned_loss=0.04308, over 2386487.95 frames. ], batch size: 44, lr: 4.54e-03, grad_scale: 8.0 2023-05-16 05:19:55,810 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.9576, 4.8824, 4.8029, 4.7955, 4.4222, 4.9531, 4.9310, 5.1442], device='cuda:0'), covar=tensor([0.0199, 0.0153, 0.0192, 0.0249, 0.0807, 0.0248, 0.0153, 0.0183], device='cuda:0'), in_proj_covar=tensor([0.0185, 0.0189, 0.0188, 0.0238, 0.0240, 0.0209, 0.0170, 0.0218], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0005, 0.0005, 0.0005, 0.0004, 0.0005], device='cuda:0') 2023-05-16 05:19:59,481 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.4269, 2.3007, 3.0813, 4.2947, 2.2319, 4.4530, 4.3382, 4.5463], device='cuda:0'), covar=tensor([0.0159, 0.1310, 0.0532, 0.0153, 0.1343, 0.0188, 0.0151, 0.0111], device='cuda:0'), in_proj_covar=tensor([0.0119, 0.0203, 0.0186, 0.0118, 0.0189, 0.0177, 0.0171, 0.0123], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 05:20:08,623 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.197e+02 2.955e+02 3.468e+02 4.300e+02 9.458e+02, threshold=6.936e+02, percent-clipped=4.0 2023-05-16 05:20:29,780 INFO [finetune.py:992] (0/2) Epoch 6, batch 7200, loss[loss=0.1846, simple_loss=0.2806, pruned_loss=0.04436, over 12272.00 frames. ], tot_loss[loss=0.173, simple_loss=0.2606, pruned_loss=0.04274, over 2385482.27 frames. ], batch size: 37, lr: 4.54e-03, grad_scale: 16.0 2023-05-16 05:20:34,034 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=168375.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 05:20:45,201 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=168391.0, num_to_drop=1, layers_to_drop={3} 2023-05-16 05:20:51,365 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.43 vs. limit=2.0 2023-05-16 05:21:06,129 INFO [finetune.py:992] (0/2) Epoch 6, batch 7250, loss[loss=0.2087, simple_loss=0.295, pruned_loss=0.06119, over 12123.00 frames. ], tot_loss[loss=0.1736, simple_loss=0.2613, pruned_loss=0.04293, over 2387332.42 frames. ], batch size: 39, lr: 4.54e-03, grad_scale: 16.0 2023-05-16 05:21:08,939 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=168423.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 05:21:08,975 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=168423.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 05:21:12,297 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.89 vs. limit=2.0 2023-05-16 05:21:21,085 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.057e+02 2.870e+02 3.424e+02 3.941e+02 5.479e+02, threshold=6.848e+02, percent-clipped=0.0 2023-05-16 05:21:41,640 INFO [finetune.py:992] (0/2) Epoch 6, batch 7300, loss[loss=0.1868, simple_loss=0.2834, pruned_loss=0.04507, over 11761.00 frames. ], tot_loss[loss=0.1747, simple_loss=0.2625, pruned_loss=0.04345, over 2377948.10 frames. ], batch size: 44, lr: 4.54e-03, grad_scale: 16.0 2023-05-16 05:21:43,200 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=168471.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 05:22:17,731 INFO [finetune.py:992] (0/2) Epoch 6, batch 7350, loss[loss=0.1692, simple_loss=0.269, pruned_loss=0.03467, over 12128.00 frames. ], tot_loss[loss=0.1755, simple_loss=0.263, pruned_loss=0.04402, over 2372528.13 frames. ], batch size: 38, lr: 4.54e-03, grad_scale: 16.0 2023-05-16 05:22:32,505 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.835e+02 2.876e+02 3.392e+02 4.190e+02 6.520e+02, threshold=6.783e+02, percent-clipped=0.0 2023-05-16 05:22:53,779 INFO [finetune.py:992] (0/2) Epoch 6, batch 7400, loss[loss=0.1552, simple_loss=0.2403, pruned_loss=0.03503, over 12341.00 frames. ], tot_loss[loss=0.1738, simple_loss=0.2615, pruned_loss=0.04309, over 2380953.99 frames. ], batch size: 31, lr: 4.54e-03, grad_scale: 16.0 2023-05-16 05:23:07,847 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=168589.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 05:23:11,821 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=2.78 vs. limit=5.0 2023-05-16 05:23:19,981 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.96 vs. limit=2.0 2023-05-16 05:23:29,679 INFO [finetune.py:992] (0/2) Epoch 6, batch 7450, loss[loss=0.1547, simple_loss=0.2278, pruned_loss=0.04083, over 12253.00 frames. ], tot_loss[loss=0.1746, simple_loss=0.2621, pruned_loss=0.04354, over 2375895.34 frames. ], batch size: 28, lr: 4.54e-03, grad_scale: 16.0 2023-05-16 05:23:44,429 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.051e+02 2.861e+02 3.391e+02 3.964e+02 1.281e+03, threshold=6.782e+02, percent-clipped=3.0 2023-05-16 05:24:05,612 INFO [finetune.py:992] (0/2) Epoch 6, batch 7500, loss[loss=0.1691, simple_loss=0.2616, pruned_loss=0.03832, over 12151.00 frames. ], tot_loss[loss=0.1746, simple_loss=0.2622, pruned_loss=0.04351, over 2371848.69 frames. ], batch size: 36, lr: 4.54e-03, grad_scale: 16.0 2023-05-16 05:24:07,348 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.6574, 2.7195, 4.5914, 4.9304, 3.2739, 2.5497, 2.8122, 2.0471], device='cuda:0'), covar=tensor([0.1441, 0.3190, 0.0426, 0.0272, 0.0921, 0.2216, 0.2719, 0.3996], device='cuda:0'), in_proj_covar=tensor([0.0276, 0.0374, 0.0263, 0.0290, 0.0257, 0.0288, 0.0358, 0.0359], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-16 05:24:21,139 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=168691.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 05:24:41,653 INFO [finetune.py:992] (0/2) Epoch 6, batch 7550, loss[loss=0.2041, simple_loss=0.2871, pruned_loss=0.06053, over 10679.00 frames. ], tot_loss[loss=0.1741, simple_loss=0.2614, pruned_loss=0.0434, over 2377372.45 frames. ], batch size: 68, lr: 4.54e-03, grad_scale: 16.0 2023-05-16 05:24:49,530 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.9849, 4.8953, 4.9062, 4.8776, 4.4414, 5.0175, 5.0531, 5.2002], device='cuda:0'), covar=tensor([0.0245, 0.0149, 0.0163, 0.0304, 0.0797, 0.0325, 0.0138, 0.0175], device='cuda:0'), in_proj_covar=tensor([0.0184, 0.0187, 0.0184, 0.0235, 0.0236, 0.0208, 0.0167, 0.0217], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0005, 0.0004, 0.0005, 0.0004, 0.0005], device='cuda:0') 2023-05-16 05:24:55,709 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=168739.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 05:24:56,319 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.766e+02 2.882e+02 3.472e+02 4.339e+02 9.651e+02, threshold=6.943e+02, percent-clipped=3.0 2023-05-16 05:25:10,890 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.6348, 4.5756, 4.4975, 4.1215, 4.2533, 4.6014, 4.2604, 4.1375], device='cuda:0'), covar=tensor([0.0830, 0.1008, 0.0692, 0.1412, 0.1663, 0.0848, 0.1543, 0.1202], device='cuda:0'), in_proj_covar=tensor([0.0584, 0.0511, 0.0493, 0.0604, 0.0395, 0.0684, 0.0740, 0.0543], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0004, 0.0004, 0.0003], device='cuda:0') 2023-05-16 05:25:17,129 INFO [finetune.py:992] (0/2) Epoch 6, batch 7600, loss[loss=0.1612, simple_loss=0.2412, pruned_loss=0.04062, over 11761.00 frames. ], tot_loss[loss=0.1756, simple_loss=0.2628, pruned_loss=0.04417, over 2364565.39 frames. ], batch size: 26, lr: 4.54e-03, grad_scale: 16.0 2023-05-16 05:25:54,188 INFO [finetune.py:992] (0/2) Epoch 6, batch 7650, loss[loss=0.1556, simple_loss=0.2428, pruned_loss=0.03418, over 12351.00 frames. ], tot_loss[loss=0.1754, simple_loss=0.2627, pruned_loss=0.04408, over 2366614.69 frames. ], batch size: 31, lr: 4.54e-03, grad_scale: 8.0 2023-05-16 05:25:55,477 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.37 vs. limit=2.0 2023-05-16 05:26:10,539 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.198e+02 3.049e+02 3.497e+02 4.286e+02 7.262e+02, threshold=6.993e+02, percent-clipped=1.0 2023-05-16 05:26:30,290 INFO [finetune.py:992] (0/2) Epoch 6, batch 7700, loss[loss=0.1644, simple_loss=0.2602, pruned_loss=0.0343, over 12299.00 frames. ], tot_loss[loss=0.1755, simple_loss=0.2629, pruned_loss=0.04406, over 2360226.70 frames. ], batch size: 34, lr: 4.53e-03, grad_scale: 8.0 2023-05-16 05:26:44,537 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=168889.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 05:26:51,091 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.8534, 2.5726, 3.5995, 3.7071, 2.9345, 2.7133, 2.6159, 2.3637], device='cuda:0'), covar=tensor([0.1024, 0.2271, 0.0546, 0.0433, 0.0833, 0.1648, 0.2217, 0.3006], device='cuda:0'), in_proj_covar=tensor([0.0275, 0.0374, 0.0264, 0.0290, 0.0256, 0.0287, 0.0356, 0.0358], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-16 05:26:58,919 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.2018, 4.4421, 2.6577, 2.4343, 3.7488, 2.3384, 3.8734, 3.0123], device='cuda:0'), covar=tensor([0.0700, 0.0482, 0.1125, 0.1534, 0.0323, 0.1328, 0.0456, 0.0808], device='cuda:0'), in_proj_covar=tensor([0.0186, 0.0249, 0.0175, 0.0196, 0.0136, 0.0179, 0.0195, 0.0174], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 05:27:05,793 INFO [finetune.py:992] (0/2) Epoch 6, batch 7750, loss[loss=0.1741, simple_loss=0.2698, pruned_loss=0.03919, over 12108.00 frames. ], tot_loss[loss=0.1761, simple_loss=0.2638, pruned_loss=0.04424, over 2357573.95 frames. ], batch size: 33, lr: 4.53e-03, grad_scale: 8.0 2023-05-16 05:27:19,098 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=168937.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 05:27:21,701 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.879e+02 2.889e+02 3.508e+02 4.779e+02 7.003e+02, threshold=7.017e+02, percent-clipped=1.0 2023-05-16 05:27:40,272 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=168967.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 05:27:41,493 INFO [finetune.py:992] (0/2) Epoch 6, batch 7800, loss[loss=0.1542, simple_loss=0.2457, pruned_loss=0.03129, over 11853.00 frames. ], tot_loss[loss=0.1763, simple_loss=0.2636, pruned_loss=0.04445, over 2355152.36 frames. ], batch size: 44, lr: 4.53e-03, grad_scale: 8.0 2023-05-16 05:27:41,720 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=168969.0, num_to_drop=1, layers_to_drop={0} 2023-05-16 05:28:03,185 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.4660, 4.8254, 4.3550, 5.2356, 4.7120, 2.9501, 4.3912, 3.1126], device='cuda:0'), covar=tensor([0.0679, 0.0661, 0.1118, 0.0386, 0.0991, 0.1582, 0.0898, 0.2979], device='cuda:0'), in_proj_covar=tensor([0.0294, 0.0371, 0.0345, 0.0270, 0.0356, 0.0262, 0.0331, 0.0353], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 05:28:18,039 INFO [finetune.py:992] (0/2) Epoch 6, batch 7850, loss[loss=0.1443, simple_loss=0.2275, pruned_loss=0.03056, over 12022.00 frames. ], tot_loss[loss=0.1767, simple_loss=0.2632, pruned_loss=0.04505, over 2350161.68 frames. ], batch size: 28, lr: 4.53e-03, grad_scale: 8.0 2023-05-16 05:28:24,378 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=169028.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 05:28:25,751 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=169030.0, num_to_drop=1, layers_to_drop={3} 2023-05-16 05:28:33,238 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.052e+02 2.923e+02 3.545e+02 4.610e+02 1.978e+03, threshold=7.090e+02, percent-clipped=5.0 2023-05-16 05:28:40,701 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.8642, 3.0506, 4.8281, 5.0846, 3.0210, 2.9042, 3.1651, 2.2411], device='cuda:0'), covar=tensor([0.1401, 0.3023, 0.0411, 0.0339, 0.1189, 0.1948, 0.2483, 0.3893], device='cuda:0'), in_proj_covar=tensor([0.0274, 0.0372, 0.0264, 0.0291, 0.0255, 0.0286, 0.0355, 0.0356], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-16 05:28:54,140 INFO [finetune.py:992] (0/2) Epoch 6, batch 7900, loss[loss=0.1774, simple_loss=0.2689, pruned_loss=0.04297, over 12148.00 frames. ], tot_loss[loss=0.1762, simple_loss=0.2632, pruned_loss=0.04467, over 2354372.67 frames. ], batch size: 34, lr: 4.53e-03, grad_scale: 8.0 2023-05-16 05:29:03,287 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=169082.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 05:29:26,919 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.8530, 1.9550, 3.3388, 3.7214, 3.4789, 3.6582, 3.2921, 2.6784], device='cuda:0'), covar=tensor([0.0035, 0.0446, 0.0139, 0.0043, 0.0095, 0.0083, 0.0125, 0.0362], device='cuda:0'), in_proj_covar=tensor([0.0082, 0.0121, 0.0102, 0.0076, 0.0100, 0.0110, 0.0091, 0.0137], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 05:29:28,388 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=169116.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 05:29:30,254 INFO [finetune.py:992] (0/2) Epoch 6, batch 7950, loss[loss=0.1766, simple_loss=0.2681, pruned_loss=0.04256, over 12093.00 frames. ], tot_loss[loss=0.1763, simple_loss=0.2635, pruned_loss=0.04449, over 2355537.37 frames. ], batch size: 32, lr: 4.53e-03, grad_scale: 8.0 2023-05-16 05:29:45,230 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.8936, 4.6207, 4.6548, 4.8221, 4.6123, 4.8011, 4.7706, 2.8487], device='cuda:0'), covar=tensor([0.0113, 0.0067, 0.0091, 0.0065, 0.0055, 0.0097, 0.0111, 0.0667], device='cuda:0'), in_proj_covar=tensor([0.0064, 0.0073, 0.0076, 0.0069, 0.0058, 0.0087, 0.0076, 0.0089], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 05:29:45,744 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.956e+02 2.769e+02 3.316e+02 3.894e+02 6.900e+02, threshold=6.632e+02, percent-clipped=0.0 2023-05-16 05:29:47,421 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=169143.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 05:29:58,831 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=169159.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 05:30:05,678 INFO [finetune.py:992] (0/2) Epoch 6, batch 8000, loss[loss=0.166, simple_loss=0.2485, pruned_loss=0.0418, over 12018.00 frames. ], tot_loss[loss=0.1748, simple_loss=0.2619, pruned_loss=0.04383, over 2363863.12 frames. ], batch size: 28, lr: 4.53e-03, grad_scale: 8.0 2023-05-16 05:30:11,609 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.5656, 3.5689, 3.3383, 3.1241, 2.9108, 2.7846, 3.5925, 2.2973], device='cuda:0'), covar=tensor([0.0350, 0.0128, 0.0132, 0.0179, 0.0328, 0.0301, 0.0115, 0.0415], device='cuda:0'), in_proj_covar=tensor([0.0192, 0.0157, 0.0153, 0.0184, 0.0199, 0.0194, 0.0163, 0.0200], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 05:30:11,620 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=169177.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 05:30:42,274 INFO [finetune.py:992] (0/2) Epoch 6, batch 8050, loss[loss=0.1706, simple_loss=0.2516, pruned_loss=0.04477, over 11989.00 frames. ], tot_loss[loss=0.1746, simple_loss=0.2616, pruned_loss=0.04381, over 2369434.97 frames. ], batch size: 28, lr: 4.53e-03, grad_scale: 8.0 2023-05-16 05:30:43,154 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=169220.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 05:30:57,584 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.840e+02 2.811e+02 3.316e+02 3.937e+02 6.960e+02, threshold=6.632e+02, percent-clipped=1.0 2023-05-16 05:31:02,186 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.7596, 3.1984, 5.1044, 2.5687, 2.7015, 3.7833, 3.2775, 3.8699], device='cuda:0'), covar=tensor([0.0359, 0.1173, 0.0249, 0.1206, 0.1909, 0.1249, 0.1246, 0.1008], device='cuda:0'), in_proj_covar=tensor([0.0230, 0.0225, 0.0230, 0.0178, 0.0232, 0.0277, 0.0220, 0.0249], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:0') 2023-05-16 05:31:18,466 INFO [finetune.py:992] (0/2) Epoch 6, batch 8100, loss[loss=0.1696, simple_loss=0.2654, pruned_loss=0.03692, over 12351.00 frames. ], tot_loss[loss=0.1744, simple_loss=0.2615, pruned_loss=0.04361, over 2377101.43 frames. ], batch size: 35, lr: 4.53e-03, grad_scale: 8.0 2023-05-16 05:31:18,654 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.9582, 4.9241, 4.8303, 4.8968, 4.4391, 4.9568, 4.9399, 5.1361], device='cuda:0'), covar=tensor([0.0246, 0.0160, 0.0178, 0.0297, 0.0772, 0.0310, 0.0156, 0.0184], device='cuda:0'), in_proj_covar=tensor([0.0186, 0.0190, 0.0185, 0.0238, 0.0238, 0.0208, 0.0169, 0.0219], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0005, 0.0005, 0.0005, 0.0004, 0.0005], device='cuda:0') 2023-05-16 05:31:21,869 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.92 vs. limit=2.0 2023-05-16 05:31:54,301 INFO [finetune.py:992] (0/2) Epoch 6, batch 8150, loss[loss=0.2878, simple_loss=0.3428, pruned_loss=0.1164, over 7957.00 frames. ], tot_loss[loss=0.1756, simple_loss=0.2625, pruned_loss=0.04437, over 2371289.66 frames. ], batch size: 98, lr: 4.53e-03, grad_scale: 4.0 2023-05-16 05:31:57,269 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=169323.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 05:31:58,737 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=169325.0, num_to_drop=1, layers_to_drop={0} 2023-05-16 05:32:10,273 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.122e+02 2.928e+02 3.402e+02 4.619e+02 3.270e+03, threshold=6.805e+02, percent-clipped=9.0 2023-05-16 05:32:30,108 INFO [finetune.py:992] (0/2) Epoch 6, batch 8200, loss[loss=0.1744, simple_loss=0.2451, pruned_loss=0.05183, over 12000.00 frames. ], tot_loss[loss=0.1758, simple_loss=0.2624, pruned_loss=0.04461, over 2370590.33 frames. ], batch size: 28, lr: 4.53e-03, grad_scale: 4.0 2023-05-16 05:32:51,177 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.82 vs. limit=2.0 2023-05-16 05:33:06,657 INFO [finetune.py:992] (0/2) Epoch 6, batch 8250, loss[loss=0.143, simple_loss=0.2216, pruned_loss=0.03217, over 11805.00 frames. ], tot_loss[loss=0.1764, simple_loss=0.2628, pruned_loss=0.045, over 2368026.11 frames. ], batch size: 26, lr: 4.53e-03, grad_scale: 4.0 2023-05-16 05:33:10,535 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.1696, 4.0603, 4.1219, 4.4654, 2.9454, 3.8820, 2.6489, 4.0451], device='cuda:0'), covar=tensor([0.1583, 0.0678, 0.0778, 0.0589, 0.1080, 0.0633, 0.1719, 0.1182], device='cuda:0'), in_proj_covar=tensor([0.0220, 0.0256, 0.0290, 0.0345, 0.0230, 0.0233, 0.0250, 0.0370], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 05:33:20,416 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=169438.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 05:33:23,183 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.299e+02 2.930e+02 3.366e+02 4.211e+02 1.512e+03, threshold=6.732e+02, percent-clipped=5.0 2023-05-16 05:33:42,697 INFO [finetune.py:992] (0/2) Epoch 6, batch 8300, loss[loss=0.1469, simple_loss=0.2325, pruned_loss=0.03063, over 12358.00 frames. ], tot_loss[loss=0.1753, simple_loss=0.2621, pruned_loss=0.04425, over 2374240.85 frames. ], batch size: 31, lr: 4.53e-03, grad_scale: 4.0 2023-05-16 05:33:44,820 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=169472.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 05:34:09,816 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.8243, 2.5205, 3.6180, 3.7154, 2.9085, 2.6661, 2.6090, 2.2712], device='cuda:0'), covar=tensor([0.1072, 0.2372, 0.0530, 0.0486, 0.0874, 0.1719, 0.2185, 0.3124], device='cuda:0'), in_proj_covar=tensor([0.0275, 0.0373, 0.0264, 0.0292, 0.0256, 0.0286, 0.0356, 0.0358], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-16 05:34:10,642 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.44 vs. limit=2.0 2023-05-16 05:34:16,014 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=169515.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 05:34:18,756 INFO [finetune.py:992] (0/2) Epoch 6, batch 8350, loss[loss=0.1702, simple_loss=0.2609, pruned_loss=0.03975, over 12344.00 frames. ], tot_loss[loss=0.1759, simple_loss=0.2628, pruned_loss=0.04449, over 2368881.44 frames. ], batch size: 35, lr: 4.53e-03, grad_scale: 4.0 2023-05-16 05:34:35,833 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.860e+02 2.672e+02 3.314e+02 4.024e+02 1.331e+03, threshold=6.629e+02, percent-clipped=5.0 2023-05-16 05:34:39,327 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=4.82 vs. limit=5.0 2023-05-16 05:34:42,977 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.87 vs. limit=2.0 2023-05-16 05:34:55,377 INFO [finetune.py:992] (0/2) Epoch 6, batch 8400, loss[loss=0.2099, simple_loss=0.2916, pruned_loss=0.06407, over 12277.00 frames. ], tot_loss[loss=0.1743, simple_loss=0.2618, pruned_loss=0.04343, over 2377886.50 frames. ], batch size: 37, lr: 4.53e-03, grad_scale: 8.0 2023-05-16 05:35:01,799 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.1962, 4.7095, 5.1392, 4.5278, 4.7572, 4.5057, 5.1922, 4.8294], device='cuda:0'), covar=tensor([0.0269, 0.0392, 0.0290, 0.0239, 0.0354, 0.0321, 0.0207, 0.0337], device='cuda:0'), in_proj_covar=tensor([0.0239, 0.0246, 0.0267, 0.0241, 0.0239, 0.0240, 0.0217, 0.0201], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 05:35:29,470 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.2427, 3.9777, 4.0830, 4.3962, 3.0749, 3.9504, 2.6440, 4.0825], device='cuda:0'), covar=tensor([0.1493, 0.0673, 0.0848, 0.0632, 0.1006, 0.0551, 0.1659, 0.0863], device='cuda:0'), in_proj_covar=tensor([0.0220, 0.0256, 0.0290, 0.0347, 0.0230, 0.0233, 0.0250, 0.0372], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 05:35:31,406 INFO [finetune.py:992] (0/2) Epoch 6, batch 8450, loss[loss=0.1774, simple_loss=0.2673, pruned_loss=0.04381, over 12049.00 frames. ], tot_loss[loss=0.1742, simple_loss=0.2616, pruned_loss=0.04344, over 2380436.95 frames. ], batch size: 37, lr: 4.53e-03, grad_scale: 8.0 2023-05-16 05:35:33,010 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=169621.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 05:35:34,435 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=169623.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 05:35:35,865 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=169625.0, num_to_drop=1, layers_to_drop={0} 2023-05-16 05:35:48,309 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.861e+02 2.765e+02 3.281e+02 4.006e+02 6.222e+02, threshold=6.562e+02, percent-clipped=0.0 2023-05-16 05:36:08,327 INFO [finetune.py:992] (0/2) Epoch 6, batch 8500, loss[loss=0.1727, simple_loss=0.2658, pruned_loss=0.03979, over 11578.00 frames. ], tot_loss[loss=0.1749, simple_loss=0.2625, pruned_loss=0.0437, over 2366742.55 frames. ], batch size: 48, lr: 4.53e-03, grad_scale: 8.0 2023-05-16 05:36:09,848 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=169671.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 05:36:11,325 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=169673.0, num_to_drop=1, layers_to_drop={0} 2023-05-16 05:36:12,884 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.83 vs. limit=2.0 2023-05-16 05:36:17,674 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=169682.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 05:36:43,592 INFO [finetune.py:992] (0/2) Epoch 6, batch 8550, loss[loss=0.1787, simple_loss=0.2705, pruned_loss=0.04342, over 12268.00 frames. ], tot_loss[loss=0.1742, simple_loss=0.262, pruned_loss=0.0432, over 2369389.77 frames. ], batch size: 37, lr: 4.53e-03, grad_scale: 8.0 2023-05-16 05:36:51,010 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=169729.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 05:36:57,423 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=169738.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 05:37:00,184 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.052e+02 2.624e+02 3.312e+02 3.872e+02 8.850e+02, threshold=6.624e+02, percent-clipped=2.0 2023-05-16 05:37:02,784 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.72 vs. limit=2.0 2023-05-16 05:37:04,057 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=169747.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 05:37:19,694 INFO [finetune.py:992] (0/2) Epoch 6, batch 8600, loss[loss=0.1551, simple_loss=0.2409, pruned_loss=0.03464, over 12105.00 frames. ], tot_loss[loss=0.1737, simple_loss=0.2618, pruned_loss=0.04285, over 2374791.44 frames. ], batch size: 33, lr: 4.53e-03, grad_scale: 8.0 2023-05-16 05:37:22,079 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=169772.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 05:37:32,576 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=169786.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 05:37:32,743 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=169786.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 05:37:35,491 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=169790.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 05:37:49,436 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=169808.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 05:37:54,405 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=169815.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 05:37:57,103 INFO [finetune.py:992] (0/2) Epoch 6, batch 8650, loss[loss=0.1785, simple_loss=0.2682, pruned_loss=0.04437, over 12266.00 frames. ], tot_loss[loss=0.1726, simple_loss=0.2608, pruned_loss=0.04217, over 2382521.25 frames. ], batch size: 37, lr: 4.52e-03, grad_scale: 8.0 2023-05-16 05:37:57,874 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=169820.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 05:38:13,690 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.759e+02 2.733e+02 3.214e+02 3.902e+02 8.674e+02, threshold=6.428e+02, percent-clipped=2.0 2023-05-16 05:38:17,541 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=169847.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 05:38:28,948 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=169863.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 05:38:33,157 INFO [finetune.py:992] (0/2) Epoch 6, batch 8700, loss[loss=0.158, simple_loss=0.2385, pruned_loss=0.03876, over 11991.00 frames. ], tot_loss[loss=0.174, simple_loss=0.2619, pruned_loss=0.04308, over 2375599.89 frames. ], batch size: 28, lr: 4.52e-03, grad_scale: 8.0 2023-05-16 05:38:33,370 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.5841, 2.4499, 3.6768, 4.5622, 4.0988, 4.4545, 3.9068, 3.4319], device='cuda:0'), covar=tensor([0.0031, 0.0416, 0.0125, 0.0032, 0.0087, 0.0071, 0.0095, 0.0307], device='cuda:0'), in_proj_covar=tensor([0.0084, 0.0122, 0.0103, 0.0076, 0.0100, 0.0112, 0.0093, 0.0139], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 05:38:47,713 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.5837, 4.4859, 4.4176, 4.5050, 4.0313, 4.5867, 4.5590, 4.7643], device='cuda:0'), covar=tensor([0.0312, 0.0174, 0.0228, 0.0335, 0.0873, 0.0352, 0.0173, 0.0216], device='cuda:0'), in_proj_covar=tensor([0.0184, 0.0189, 0.0184, 0.0237, 0.0236, 0.0208, 0.0169, 0.0216], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0005, 0.0005, 0.0005, 0.0004, 0.0005], device='cuda:0') 2023-05-16 05:39:00,597 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.5683, 3.2823, 3.1778, 3.0965, 2.8561, 2.6328, 3.4006, 2.3626], device='cuda:0'), covar=tensor([0.0332, 0.0152, 0.0165, 0.0183, 0.0332, 0.0339, 0.0125, 0.0388], device='cuda:0'), in_proj_covar=tensor([0.0191, 0.0156, 0.0153, 0.0184, 0.0199, 0.0194, 0.0162, 0.0197], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 05:39:08,885 INFO [finetune.py:992] (0/2) Epoch 6, batch 8750, loss[loss=0.1557, simple_loss=0.2345, pruned_loss=0.03848, over 12288.00 frames. ], tot_loss[loss=0.173, simple_loss=0.2608, pruned_loss=0.04262, over 2378923.15 frames. ], batch size: 28, lr: 4.52e-03, grad_scale: 8.0 2023-05-16 05:39:26,002 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.498e+02 2.814e+02 3.203e+02 3.882e+02 8.670e+02, threshold=6.406e+02, percent-clipped=2.0 2023-05-16 05:39:34,838 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=2.76 vs. limit=5.0 2023-05-16 05:39:45,834 INFO [finetune.py:992] (0/2) Epoch 6, batch 8800, loss[loss=0.1814, simple_loss=0.27, pruned_loss=0.04638, over 12168.00 frames. ], tot_loss[loss=0.1733, simple_loss=0.261, pruned_loss=0.0428, over 2377854.41 frames. ], batch size: 35, lr: 4.52e-03, grad_scale: 8.0 2023-05-16 05:39:51,505 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=169977.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 05:40:08,097 INFO [checkpoint.py:75] (0/2) Saving checkpoint to pruned_transducer_stateless7/exp_giga_finetune/checkpoint-70000.pt 2023-05-16 05:40:24,643 INFO [finetune.py:992] (0/2) Epoch 6, batch 8850, loss[loss=0.1726, simple_loss=0.2654, pruned_loss=0.03989, over 12267.00 frames. ], tot_loss[loss=0.1731, simple_loss=0.2609, pruned_loss=0.04265, over 2381196.36 frames. ], batch size: 37, lr: 4.52e-03, grad_scale: 8.0 2023-05-16 05:40:32,119 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.44 vs. limit=2.0 2023-05-16 05:40:40,961 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.125e+02 2.852e+02 3.342e+02 4.079e+02 1.347e+03, threshold=6.684e+02, percent-clipped=1.0 2023-05-16 05:40:43,269 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=170045.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 05:40:51,797 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.3356, 4.7059, 2.8050, 2.7213, 3.9335, 2.5815, 3.9383, 3.0675], device='cuda:0'), covar=tensor([0.0681, 0.0513, 0.1079, 0.1475, 0.0280, 0.1367, 0.0509, 0.0905], device='cuda:0'), in_proj_covar=tensor([0.0187, 0.0251, 0.0175, 0.0196, 0.0138, 0.0180, 0.0196, 0.0175], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 05:41:00,658 INFO [finetune.py:992] (0/2) Epoch 6, batch 8900, loss[loss=0.1598, simple_loss=0.2481, pruned_loss=0.03576, over 12361.00 frames. ], tot_loss[loss=0.1732, simple_loss=0.2611, pruned_loss=0.04264, over 2381169.61 frames. ], batch size: 36, lr: 4.52e-03, grad_scale: 8.0 2023-05-16 05:41:12,246 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=170085.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 05:41:25,964 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=170103.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 05:41:28,210 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=170106.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 05:41:37,258 INFO [finetune.py:992] (0/2) Epoch 6, batch 8950, loss[loss=0.1777, simple_loss=0.2723, pruned_loss=0.04159, over 12146.00 frames. ], tot_loss[loss=0.1731, simple_loss=0.2611, pruned_loss=0.04257, over 2385232.96 frames. ], batch size: 34, lr: 4.52e-03, grad_scale: 8.0 2023-05-16 05:41:44,294 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 2023-05-16 05:41:49,850 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.6760, 2.9658, 4.5318, 4.6756, 2.9669, 2.6612, 2.8889, 1.9652], device='cuda:0'), covar=tensor([0.1374, 0.2522, 0.0429, 0.0374, 0.1136, 0.2052, 0.2589, 0.3877], device='cuda:0'), in_proj_covar=tensor([0.0274, 0.0370, 0.0263, 0.0288, 0.0257, 0.0286, 0.0354, 0.0357], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-16 05:41:53,762 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.831e+02 2.640e+02 3.012e+02 3.913e+02 7.653e+02, threshold=6.023e+02, percent-clipped=3.0 2023-05-16 05:41:53,895 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=170142.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 05:42:13,144 INFO [finetune.py:992] (0/2) Epoch 6, batch 9000, loss[loss=0.188, simple_loss=0.2824, pruned_loss=0.04684, over 12293.00 frames. ], tot_loss[loss=0.1727, simple_loss=0.2609, pruned_loss=0.04223, over 2390452.83 frames. ], batch size: 34, lr: 4.52e-03, grad_scale: 8.0 2023-05-16 05:42:13,145 INFO [finetune.py:1017] (0/2) Computing validation loss 2023-05-16 05:42:31,142 INFO [finetune.py:1026] (0/2) Epoch 6, validation: loss=0.3329, simple_loss=0.4046, pruned_loss=0.1306, over 1020973.00 frames. 2023-05-16 05:42:31,143 INFO [finetune.py:1027] (0/2) Maximum memory allocated so far is 12745MB 2023-05-16 05:42:50,100 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.2807, 2.7459, 3.5215, 4.2730, 3.7346, 4.2375, 3.6144, 2.9876], device='cuda:0'), covar=tensor([0.0039, 0.0326, 0.0143, 0.0037, 0.0121, 0.0061, 0.0132, 0.0348], device='cuda:0'), in_proj_covar=tensor([0.0084, 0.0123, 0.0104, 0.0077, 0.0101, 0.0113, 0.0094, 0.0141], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 05:42:52,999 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2023-05-16 05:43:08,310 INFO [finetune.py:992] (0/2) Epoch 6, batch 9050, loss[loss=0.157, simple_loss=0.238, pruned_loss=0.03804, over 11838.00 frames. ], tot_loss[loss=0.1743, simple_loss=0.2621, pruned_loss=0.04319, over 2381837.52 frames. ], batch size: 26, lr: 4.52e-03, grad_scale: 8.0 2023-05-16 05:43:24,847 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.254e+02 2.871e+02 3.359e+02 3.998e+02 1.069e+03, threshold=6.719e+02, percent-clipped=4.0 2023-05-16 05:43:43,951 INFO [finetune.py:992] (0/2) Epoch 6, batch 9100, loss[loss=0.205, simple_loss=0.2904, pruned_loss=0.05985, over 12156.00 frames. ], tot_loss[loss=0.1747, simple_loss=0.2623, pruned_loss=0.04357, over 2376307.70 frames. ], batch size: 36, lr: 4.52e-03, grad_scale: 8.0 2023-05-16 05:43:49,814 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=170277.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 05:44:12,429 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.3162, 4.4778, 4.2629, 4.9314, 4.6677, 2.7836, 4.2719, 3.1482], device='cuda:0'), covar=tensor([0.0691, 0.0788, 0.1171, 0.0431, 0.0901, 0.1534, 0.0969, 0.2737], device='cuda:0'), in_proj_covar=tensor([0.0295, 0.0371, 0.0347, 0.0271, 0.0355, 0.0259, 0.0332, 0.0353], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 05:44:18,086 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.8586, 3.0552, 4.9087, 5.0337, 3.1859, 2.8437, 3.0993, 2.2289], device='cuda:0'), covar=tensor([0.1407, 0.2612, 0.0370, 0.0335, 0.1086, 0.2077, 0.2430, 0.3795], device='cuda:0'), in_proj_covar=tensor([0.0273, 0.0370, 0.0262, 0.0287, 0.0256, 0.0285, 0.0353, 0.0355], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-16 05:44:19,916 INFO [finetune.py:992] (0/2) Epoch 6, batch 9150, loss[loss=0.1819, simple_loss=0.2696, pruned_loss=0.04714, over 12187.00 frames. ], tot_loss[loss=0.1758, simple_loss=0.2635, pruned_loss=0.04411, over 2364713.32 frames. ], batch size: 35, lr: 4.52e-03, grad_scale: 8.0 2023-05-16 05:44:20,211 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.7033, 2.8995, 4.7150, 4.8604, 3.0702, 2.7332, 2.9957, 2.0955], device='cuda:0'), covar=tensor([0.1437, 0.2899, 0.0434, 0.0368, 0.1151, 0.2025, 0.2548, 0.3883], device='cuda:0'), in_proj_covar=tensor([0.0273, 0.0370, 0.0263, 0.0287, 0.0256, 0.0285, 0.0353, 0.0356], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-16 05:44:24,172 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=170325.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 05:44:36,794 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.002e+02 2.818e+02 3.271e+02 4.074e+02 1.050e+03, threshold=6.541e+02, percent-clipped=5.0 2023-05-16 05:44:56,295 INFO [finetune.py:992] (0/2) Epoch 6, batch 9200, loss[loss=0.1496, simple_loss=0.2366, pruned_loss=0.03129, over 12331.00 frames. ], tot_loss[loss=0.1749, simple_loss=0.2623, pruned_loss=0.04377, over 2373809.26 frames. ], batch size: 31, lr: 4.52e-03, grad_scale: 8.0 2023-05-16 05:45:07,745 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=170385.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 05:45:19,456 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=170401.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 05:45:20,815 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=170403.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 05:45:25,176 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.5531, 3.5891, 3.3669, 3.2493, 3.0516, 2.8397, 3.6623, 2.3504], device='cuda:0'), covar=tensor([0.0360, 0.0172, 0.0167, 0.0178, 0.0326, 0.0330, 0.0140, 0.0401], device='cuda:0'), in_proj_covar=tensor([0.0192, 0.0157, 0.0155, 0.0184, 0.0199, 0.0193, 0.0162, 0.0199], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 05:45:32,036 INFO [finetune.py:992] (0/2) Epoch 6, batch 9250, loss[loss=0.1808, simple_loss=0.2756, pruned_loss=0.043, over 11279.00 frames. ], tot_loss[loss=0.1748, simple_loss=0.2624, pruned_loss=0.04362, over 2370712.60 frames. ], batch size: 55, lr: 4.52e-03, grad_scale: 8.0 2023-05-16 05:45:42,040 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=170433.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 05:45:48,292 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.993e+02 2.833e+02 3.194e+02 3.964e+02 6.931e+02, threshold=6.389e+02, percent-clipped=0.0 2023-05-16 05:45:48,484 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=170442.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 05:45:54,898 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=170451.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 05:46:02,800 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=170461.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 05:46:08,232 INFO [finetune.py:992] (0/2) Epoch 6, batch 9300, loss[loss=0.1953, simple_loss=0.2833, pruned_loss=0.05359, over 11262.00 frames. ], tot_loss[loss=0.1748, simple_loss=0.2628, pruned_loss=0.04337, over 2378236.09 frames. ], batch size: 55, lr: 4.52e-03, grad_scale: 8.0 2023-05-16 05:46:15,591 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.3198, 2.2512, 3.0497, 4.1736, 2.0951, 4.2223, 4.2357, 4.3934], device='cuda:0'), covar=tensor([0.0124, 0.1301, 0.0492, 0.0131, 0.1370, 0.0204, 0.0167, 0.0075], device='cuda:0'), in_proj_covar=tensor([0.0118, 0.0200, 0.0184, 0.0114, 0.0187, 0.0175, 0.0170, 0.0120], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 05:46:21,234 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.54 vs. limit=2.0 2023-05-16 05:46:23,808 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=170490.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 05:46:44,295 INFO [finetune.py:992] (0/2) Epoch 6, batch 9350, loss[loss=0.1984, simple_loss=0.2774, pruned_loss=0.05967, over 11866.00 frames. ], tot_loss[loss=0.1735, simple_loss=0.2614, pruned_loss=0.04284, over 2387919.24 frames. ], batch size: 44, lr: 4.52e-03, grad_scale: 8.0 2023-05-16 05:46:46,711 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=170522.0, num_to_drop=1, layers_to_drop={1} 2023-05-16 05:47:00,813 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.973e+02 2.784e+02 3.192e+02 3.869e+02 5.304e+02, threshold=6.383e+02, percent-clipped=1.0 2023-05-16 05:47:04,627 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.3652, 3.4322, 3.1535, 3.0916, 2.8532, 2.5861, 3.3804, 2.1930], device='cuda:0'), covar=tensor([0.0388, 0.0167, 0.0205, 0.0201, 0.0367, 0.0376, 0.0166, 0.0463], device='cuda:0'), in_proj_covar=tensor([0.0192, 0.0158, 0.0155, 0.0185, 0.0200, 0.0194, 0.0163, 0.0199], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 05:47:19,835 INFO [finetune.py:992] (0/2) Epoch 6, batch 9400, loss[loss=0.1683, simple_loss=0.2614, pruned_loss=0.03759, over 12286.00 frames. ], tot_loss[loss=0.1737, simple_loss=0.2616, pruned_loss=0.04293, over 2381936.86 frames. ], batch size: 37, lr: 4.52e-03, grad_scale: 8.0 2023-05-16 05:47:20,010 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=170569.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 05:47:56,151 INFO [finetune.py:992] (0/2) Epoch 6, batch 9450, loss[loss=0.1595, simple_loss=0.2498, pruned_loss=0.03463, over 12164.00 frames. ], tot_loss[loss=0.174, simple_loss=0.262, pruned_loss=0.043, over 2376964.13 frames. ], batch size: 34, lr: 4.52e-03, grad_scale: 8.0 2023-05-16 05:47:58,454 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.3697, 4.8947, 5.3389, 4.6384, 4.9440, 4.7647, 5.3945, 5.0370], device='cuda:0'), covar=tensor([0.0238, 0.0377, 0.0255, 0.0248, 0.0292, 0.0292, 0.0196, 0.0313], device='cuda:0'), in_proj_covar=tensor([0.0243, 0.0247, 0.0270, 0.0247, 0.0242, 0.0244, 0.0220, 0.0207], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 05:47:59,599 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 2023-05-16 05:48:04,984 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=170630.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 05:48:13,261 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.066e+02 2.989e+02 3.445e+02 4.059e+02 9.553e+02, threshold=6.891e+02, percent-clipped=4.0 2023-05-16 05:48:23,483 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.3402, 3.9418, 3.9558, 4.3650, 2.9136, 3.8946, 2.5267, 3.9346], device='cuda:0'), covar=tensor([0.1510, 0.0705, 0.0907, 0.0556, 0.1093, 0.0620, 0.1773, 0.1195], device='cuda:0'), in_proj_covar=tensor([0.0221, 0.0258, 0.0291, 0.0347, 0.0231, 0.0237, 0.0251, 0.0373], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 05:48:32,555 INFO [finetune.py:992] (0/2) Epoch 6, batch 9500, loss[loss=0.1603, simple_loss=0.2496, pruned_loss=0.03552, over 12155.00 frames. ], tot_loss[loss=0.1744, simple_loss=0.2625, pruned_loss=0.04312, over 2382634.09 frames. ], batch size: 34, lr: 4.52e-03, grad_scale: 8.0 2023-05-16 05:48:55,545 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=170701.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 05:49:00,448 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.2662, 5.0829, 5.1760, 5.2368, 4.8296, 4.8982, 4.6747, 5.1814], device='cuda:0'), covar=tensor([0.0605, 0.0594, 0.0854, 0.0600, 0.1903, 0.1259, 0.0607, 0.0914], device='cuda:0'), in_proj_covar=tensor([0.0488, 0.0650, 0.0556, 0.0589, 0.0798, 0.0705, 0.0526, 0.0468], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0005, 0.0004, 0.0004, 0.0005, 0.0005, 0.0004, 0.0004], device='cuda:0') 2023-05-16 05:49:00,542 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.0515, 2.4790, 3.5730, 2.9439, 3.2957, 3.0965, 2.3802, 3.4322], device='cuda:0'), covar=tensor([0.0118, 0.0339, 0.0139, 0.0247, 0.0155, 0.0187, 0.0372, 0.0124], device='cuda:0'), in_proj_covar=tensor([0.0165, 0.0196, 0.0180, 0.0173, 0.0199, 0.0151, 0.0186, 0.0170], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 05:49:08,207 INFO [finetune.py:992] (0/2) Epoch 6, batch 9550, loss[loss=0.1566, simple_loss=0.2444, pruned_loss=0.03439, over 12122.00 frames. ], tot_loss[loss=0.1743, simple_loss=0.2627, pruned_loss=0.04294, over 2380936.31 frames. ], batch size: 30, lr: 4.52e-03, grad_scale: 8.0 2023-05-16 05:49:24,657 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.815e+02 2.831e+02 3.344e+02 4.153e+02 1.450e+03, threshold=6.688e+02, percent-clipped=3.0 2023-05-16 05:49:29,521 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=170749.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 05:49:45,084 INFO [finetune.py:992] (0/2) Epoch 6, batch 9600, loss[loss=0.1576, simple_loss=0.2434, pruned_loss=0.03588, over 12019.00 frames. ], tot_loss[loss=0.1742, simple_loss=0.2624, pruned_loss=0.04304, over 2386739.11 frames. ], batch size: 31, lr: 4.51e-03, grad_scale: 8.0 2023-05-16 05:50:18,404 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.2765, 4.6111, 2.8653, 2.4788, 4.0214, 2.6608, 4.0196, 3.2542], device='cuda:0'), covar=tensor([0.0724, 0.0621, 0.1094, 0.1603, 0.0222, 0.1235, 0.0417, 0.0762], device='cuda:0'), in_proj_covar=tensor([0.0187, 0.0251, 0.0176, 0.0197, 0.0138, 0.0180, 0.0196, 0.0175], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 05:50:19,698 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=170817.0, num_to_drop=1, layers_to_drop={2} 2023-05-16 05:50:20,995 INFO [finetune.py:992] (0/2) Epoch 6, batch 9650, loss[loss=0.1842, simple_loss=0.267, pruned_loss=0.05068, over 12293.00 frames. ], tot_loss[loss=0.1748, simple_loss=0.2627, pruned_loss=0.04349, over 2375765.54 frames. ], batch size: 34, lr: 4.51e-03, grad_scale: 8.0 2023-05-16 05:50:37,410 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.977e+02 2.759e+02 3.162e+02 3.723e+02 7.931e+02, threshold=6.323e+02, percent-clipped=2.0 2023-05-16 05:50:43,308 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.2152, 2.0266, 2.4500, 2.1930, 2.3883, 2.4193, 1.8277, 2.4150], device='cuda:0'), covar=tensor([0.0105, 0.0267, 0.0165, 0.0187, 0.0142, 0.0146, 0.0259, 0.0136], device='cuda:0'), in_proj_covar=tensor([0.0163, 0.0195, 0.0178, 0.0171, 0.0198, 0.0150, 0.0184, 0.0168], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 05:50:55,638 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.2792, 4.5765, 4.0555, 4.9042, 4.6065, 2.7687, 4.2775, 3.0257], device='cuda:0'), covar=tensor([0.0667, 0.0681, 0.1272, 0.0373, 0.0882, 0.1524, 0.0894, 0.3038], device='cuda:0'), in_proj_covar=tensor([0.0297, 0.0370, 0.0348, 0.0272, 0.0359, 0.0260, 0.0334, 0.0355], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 05:50:56,767 INFO [finetune.py:992] (0/2) Epoch 6, batch 9700, loss[loss=0.1675, simple_loss=0.2569, pruned_loss=0.03907, over 12278.00 frames. ], tot_loss[loss=0.1743, simple_loss=0.2621, pruned_loss=0.04325, over 2374834.69 frames. ], batch size: 37, lr: 4.51e-03, grad_scale: 8.0 2023-05-16 05:51:15,980 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=4.69 vs. limit=5.0 2023-05-16 05:51:33,889 INFO [finetune.py:992] (0/2) Epoch 6, batch 9750, loss[loss=0.1543, simple_loss=0.2394, pruned_loss=0.03456, over 12000.00 frames. ], tot_loss[loss=0.1755, simple_loss=0.2632, pruned_loss=0.0439, over 2365743.72 frames. ], batch size: 28, lr: 4.51e-03, grad_scale: 8.0 2023-05-16 05:51:38,246 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=170925.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 05:51:40,547 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=170928.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 05:51:50,279 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.012e+02 2.889e+02 3.336e+02 3.908e+02 8.110e+02, threshold=6.672e+02, percent-clipped=2.0 2023-05-16 05:51:51,191 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([6.1266, 6.0417, 5.8571, 5.3346, 5.2113, 6.0110, 5.5850, 5.4061], device='cuda:0'), covar=tensor([0.0632, 0.0994, 0.0650, 0.1518, 0.0600, 0.0700, 0.1594, 0.0978], device='cuda:0'), in_proj_covar=tensor([0.0578, 0.0507, 0.0486, 0.0591, 0.0391, 0.0678, 0.0730, 0.0535], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0004, 0.0004, 0.0003], device='cuda:0') 2023-05-16 05:52:09,565 INFO [finetune.py:992] (0/2) Epoch 6, batch 9800, loss[loss=0.1567, simple_loss=0.2422, pruned_loss=0.03554, over 12028.00 frames. ], tot_loss[loss=0.1755, simple_loss=0.2631, pruned_loss=0.04399, over 2368985.75 frames. ], batch size: 31, lr: 4.51e-03, grad_scale: 8.0 2023-05-16 05:52:12,585 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.9467, 4.8944, 4.8204, 4.7727, 4.5372, 4.9293, 4.9258, 5.1397], device='cuda:0'), covar=tensor([0.0222, 0.0167, 0.0173, 0.0343, 0.0667, 0.0245, 0.0151, 0.0160], device='cuda:0'), in_proj_covar=tensor([0.0183, 0.0187, 0.0183, 0.0233, 0.0233, 0.0206, 0.0167, 0.0213], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0005, 0.0004, 0.0005, 0.0004, 0.0005], device='cuda:0') 2023-05-16 05:52:20,552 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.8138, 3.4479, 5.1147, 2.7137, 2.7675, 3.8273, 3.3104, 3.8661], device='cuda:0'), covar=tensor([0.0312, 0.0961, 0.0235, 0.1129, 0.1845, 0.1263, 0.1133, 0.1130], device='cuda:0'), in_proj_covar=tensor([0.0228, 0.0225, 0.0232, 0.0177, 0.0232, 0.0279, 0.0220, 0.0249], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:0') 2023-05-16 05:52:24,074 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=170989.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 05:52:25,582 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.3260, 4.6204, 4.0143, 4.9353, 4.5111, 2.8183, 4.3841, 3.1657], device='cuda:0'), covar=tensor([0.0783, 0.0737, 0.1429, 0.0411, 0.1117, 0.1569, 0.0950, 0.2919], device='cuda:0'), in_proj_covar=tensor([0.0297, 0.0371, 0.0349, 0.0274, 0.0360, 0.0262, 0.0335, 0.0357], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 05:52:25,808 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=4.49 vs. limit=5.0 2023-05-16 05:52:41,985 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.2947, 6.0730, 5.6647, 5.7357, 6.1765, 5.5462, 5.8071, 5.7100], device='cuda:0'), covar=tensor([0.1377, 0.0793, 0.0856, 0.1657, 0.0885, 0.1917, 0.1396, 0.0975], device='cuda:0'), in_proj_covar=tensor([0.0341, 0.0461, 0.0365, 0.0415, 0.0439, 0.0419, 0.0376, 0.0363], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 05:52:45,519 INFO [finetune.py:992] (0/2) Epoch 6, batch 9850, loss[loss=0.1726, simple_loss=0.2622, pruned_loss=0.0415, over 12204.00 frames. ], tot_loss[loss=0.1749, simple_loss=0.2622, pruned_loss=0.04379, over 2366020.07 frames. ], batch size: 35, lr: 4.51e-03, grad_scale: 8.0 2023-05-16 05:52:50,156 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=3.43 vs. limit=5.0 2023-05-16 05:52:51,392 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.7499, 4.4053, 4.5957, 4.6751, 4.5747, 4.5951, 4.5362, 2.7176], device='cuda:0'), covar=tensor([0.0111, 0.0072, 0.0078, 0.0062, 0.0052, 0.0103, 0.0088, 0.0658], device='cuda:0'), in_proj_covar=tensor([0.0065, 0.0074, 0.0078, 0.0071, 0.0059, 0.0089, 0.0077, 0.0091], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 05:53:03,166 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.919e+02 2.930e+02 3.532e+02 4.214e+02 7.457e+02, threshold=7.065e+02, percent-clipped=2.0 2023-05-16 05:53:21,975 INFO [finetune.py:992] (0/2) Epoch 6, batch 9900, loss[loss=0.1719, simple_loss=0.259, pruned_loss=0.04241, over 12082.00 frames. ], tot_loss[loss=0.1749, simple_loss=0.2623, pruned_loss=0.04377, over 2368710.87 frames. ], batch size: 33, lr: 4.51e-03, grad_scale: 8.0 2023-05-16 05:53:38,346 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=171092.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 05:53:56,115 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=171117.0, num_to_drop=1, layers_to_drop={1} 2023-05-16 05:53:57,356 INFO [finetune.py:992] (0/2) Epoch 6, batch 9950, loss[loss=0.1809, simple_loss=0.2749, pruned_loss=0.04349, over 11772.00 frames. ], tot_loss[loss=0.1761, simple_loss=0.2637, pruned_loss=0.04421, over 2356849.28 frames. ], batch size: 44, lr: 4.51e-03, grad_scale: 8.0 2023-05-16 05:54:13,383 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.015e+02 2.846e+02 3.389e+02 4.113e+02 8.851e+02, threshold=6.779e+02, percent-clipped=1.0 2023-05-16 05:54:20,775 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.91 vs. limit=2.0 2023-05-16 05:54:21,282 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=171153.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 05:54:29,801 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=171165.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 05:54:32,655 INFO [finetune.py:992] (0/2) Epoch 6, batch 10000, loss[loss=0.1813, simple_loss=0.2651, pruned_loss=0.04878, over 12128.00 frames. ], tot_loss[loss=0.1758, simple_loss=0.2634, pruned_loss=0.04413, over 2363512.93 frames. ], batch size: 38, lr: 4.51e-03, grad_scale: 8.0 2023-05-16 05:54:53,210 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=171196.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 05:55:09,435 INFO [finetune.py:992] (0/2) Epoch 6, batch 10050, loss[loss=0.1456, simple_loss=0.2284, pruned_loss=0.03142, over 12208.00 frames. ], tot_loss[loss=0.1749, simple_loss=0.2623, pruned_loss=0.04373, over 2365624.97 frames. ], batch size: 29, lr: 4.51e-03, grad_scale: 8.0 2023-05-16 05:55:13,627 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=171225.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 05:55:25,461 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.122e+02 2.682e+02 3.254e+02 3.835e+02 7.863e+02, threshold=6.507e+02, percent-clipped=3.0 2023-05-16 05:55:36,131 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=171257.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 05:55:44,425 INFO [finetune.py:992] (0/2) Epoch 6, batch 10100, loss[loss=0.1599, simple_loss=0.2495, pruned_loss=0.03513, over 12258.00 frames. ], tot_loss[loss=0.1753, simple_loss=0.2628, pruned_loss=0.04392, over 2362290.19 frames. ], batch size: 32, lr: 4.51e-03, grad_scale: 8.0 2023-05-16 05:55:47,300 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=171273.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 05:55:55,317 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=171284.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 05:56:20,084 INFO [finetune.py:992] (0/2) Epoch 6, batch 10150, loss[loss=0.1639, simple_loss=0.2568, pruned_loss=0.0355, over 12077.00 frames. ], tot_loss[loss=0.1752, simple_loss=0.2627, pruned_loss=0.04386, over 2368577.67 frames. ], batch size: 40, lr: 4.51e-03, grad_scale: 16.0 2023-05-16 05:56:37,677 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.872e+02 2.599e+02 3.165e+02 3.723e+02 8.067e+02, threshold=6.331e+02, percent-clipped=2.0 2023-05-16 05:56:40,061 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=171345.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 05:56:56,938 INFO [finetune.py:992] (0/2) Epoch 6, batch 10200, loss[loss=0.1458, simple_loss=0.2306, pruned_loss=0.03051, over 11345.00 frames. ], tot_loss[loss=0.1738, simple_loss=0.2617, pruned_loss=0.043, over 2377244.60 frames. ], batch size: 25, lr: 4.51e-03, grad_scale: 16.0 2023-05-16 05:57:16,035 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.7510, 4.0258, 3.5976, 4.2871, 3.8892, 2.6553, 3.6998, 2.8091], device='cuda:0'), covar=tensor([0.0945, 0.0902, 0.1485, 0.0464, 0.1192, 0.1603, 0.1060, 0.3202], device='cuda:0'), in_proj_covar=tensor([0.0298, 0.0372, 0.0347, 0.0273, 0.0358, 0.0263, 0.0334, 0.0356], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 05:57:19,711 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([6.2213, 6.1944, 6.0290, 5.4039, 5.1663, 6.1455, 5.7521, 5.6326], device='cuda:0'), covar=tensor([0.0612, 0.0794, 0.0533, 0.1500, 0.0619, 0.0693, 0.1458, 0.0927], device='cuda:0'), in_proj_covar=tensor([0.0580, 0.0505, 0.0486, 0.0589, 0.0393, 0.0676, 0.0733, 0.0536], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0004, 0.0004, 0.0003], device='cuda:0') 2023-05-16 05:57:24,030 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=171406.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 05:57:33,087 INFO [finetune.py:992] (0/2) Epoch 6, batch 10250, loss[loss=0.1426, simple_loss=0.2244, pruned_loss=0.03038, over 12350.00 frames. ], tot_loss[loss=0.1739, simple_loss=0.2619, pruned_loss=0.04295, over 2370640.85 frames. ], batch size: 30, lr: 4.51e-03, grad_scale: 16.0 2023-05-16 05:57:49,597 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.761e+02 2.739e+02 3.296e+02 4.036e+02 9.177e+02, threshold=6.592e+02, percent-clipped=3.0 2023-05-16 05:57:53,892 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=171448.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 05:58:04,087 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.1503, 5.0944, 4.9834, 5.0987, 4.7101, 5.1896, 5.1754, 5.3525], device='cuda:0'), covar=tensor([0.0204, 0.0152, 0.0192, 0.0265, 0.0714, 0.0233, 0.0135, 0.0139], device='cuda:0'), in_proj_covar=tensor([0.0184, 0.0189, 0.0185, 0.0236, 0.0235, 0.0207, 0.0168, 0.0215], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0005, 0.0004, 0.0005, 0.0004, 0.0005], device='cuda:0') 2023-05-16 05:58:08,840 INFO [finetune.py:992] (0/2) Epoch 6, batch 10300, loss[loss=0.1901, simple_loss=0.2813, pruned_loss=0.04947, over 12298.00 frames. ], tot_loss[loss=0.1739, simple_loss=0.2619, pruned_loss=0.04295, over 2371897.15 frames. ], batch size: 34, lr: 4.51e-03, grad_scale: 16.0 2023-05-16 05:58:30,777 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=171498.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 05:58:32,152 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.0692, 4.6805, 4.8566, 4.9342, 4.8658, 4.9414, 4.8918, 2.7815], device='cuda:0'), covar=tensor([0.0090, 0.0063, 0.0084, 0.0059, 0.0046, 0.0077, 0.0100, 0.0706], device='cuda:0'), in_proj_covar=tensor([0.0066, 0.0075, 0.0079, 0.0072, 0.0059, 0.0090, 0.0078, 0.0093], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 05:58:45,642 INFO [finetune.py:992] (0/2) Epoch 6, batch 10350, loss[loss=0.1787, simple_loss=0.2718, pruned_loss=0.04284, over 12154.00 frames. ], tot_loss[loss=0.1734, simple_loss=0.2616, pruned_loss=0.04263, over 2376825.42 frames. ], batch size: 36, lr: 4.51e-03, grad_scale: 16.0 2023-05-16 05:59:01,570 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.868e+02 2.760e+02 3.243e+02 4.133e+02 1.006e+03, threshold=6.485e+02, percent-clipped=2.0 2023-05-16 05:59:08,829 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=171552.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 05:59:08,949 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.3299, 2.5874, 3.6182, 4.3751, 3.8853, 4.2796, 3.8316, 2.8998], device='cuda:0'), covar=tensor([0.0036, 0.0346, 0.0126, 0.0027, 0.0106, 0.0068, 0.0090, 0.0367], device='cuda:0'), in_proj_covar=tensor([0.0083, 0.0120, 0.0102, 0.0075, 0.0099, 0.0112, 0.0092, 0.0137], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 05:59:13,943 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=171559.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 05:59:20,901 INFO [finetune.py:992] (0/2) Epoch 6, batch 10400, loss[loss=0.1834, simple_loss=0.2769, pruned_loss=0.04495, over 12357.00 frames. ], tot_loss[loss=0.1725, simple_loss=0.2605, pruned_loss=0.04219, over 2377866.11 frames. ], batch size: 38, lr: 4.51e-03, grad_scale: 16.0 2023-05-16 05:59:30,255 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.5200, 3.6182, 3.3407, 3.2238, 2.9257, 2.7053, 3.6504, 2.3091], device='cuda:0'), covar=tensor([0.0354, 0.0120, 0.0142, 0.0154, 0.0356, 0.0329, 0.0100, 0.0428], device='cuda:0'), in_proj_covar=tensor([0.0188, 0.0155, 0.0151, 0.0180, 0.0196, 0.0190, 0.0159, 0.0196], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 05:59:31,542 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=171584.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 05:59:38,705 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.54 vs. limit=2.0 2023-05-16 05:59:56,464 INFO [finetune.py:992] (0/2) Epoch 6, batch 10450, loss[loss=0.2328, simple_loss=0.3109, pruned_loss=0.07737, over 7916.00 frames. ], tot_loss[loss=0.1733, simple_loss=0.2615, pruned_loss=0.04254, over 2374072.44 frames. ], batch size: 97, lr: 4.51e-03, grad_scale: 16.0 2023-05-16 06:00:03,615 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.4701, 2.3647, 3.1318, 4.3894, 2.2506, 4.3866, 4.3876, 4.5845], device='cuda:0'), covar=tensor([0.0108, 0.1249, 0.0510, 0.0130, 0.1239, 0.0187, 0.0144, 0.0093], device='cuda:0'), in_proj_covar=tensor([0.0121, 0.0203, 0.0185, 0.0116, 0.0187, 0.0177, 0.0172, 0.0122], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 06:00:07,046 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=171632.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 06:00:14,130 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.168e+02 2.841e+02 3.503e+02 4.038e+02 1.380e+03, threshold=7.007e+02, percent-clipped=4.0 2023-05-16 06:00:14,539 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 2023-05-16 06:00:16,728 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2023-05-16 06:00:33,296 INFO [finetune.py:992] (0/2) Epoch 6, batch 10500, loss[loss=0.1542, simple_loss=0.2372, pruned_loss=0.0356, over 12296.00 frames. ], tot_loss[loss=0.1734, simple_loss=0.2615, pruned_loss=0.0427, over 2370898.03 frames. ], batch size: 28, lr: 4.50e-03, grad_scale: 16.0 2023-05-16 06:00:56,180 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=171701.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 06:01:09,087 INFO [finetune.py:992] (0/2) Epoch 6, batch 10550, loss[loss=0.1497, simple_loss=0.249, pruned_loss=0.02522, over 12119.00 frames. ], tot_loss[loss=0.1728, simple_loss=0.2612, pruned_loss=0.04218, over 2374218.01 frames. ], batch size: 33, lr: 4.50e-03, grad_scale: 16.0 2023-05-16 06:01:18,789 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.2339, 3.5781, 3.5655, 4.0021, 2.8478, 3.5622, 2.5514, 3.4682], device='cuda:0'), covar=tensor([0.1522, 0.0752, 0.0895, 0.0663, 0.1105, 0.0675, 0.1651, 0.1048], device='cuda:0'), in_proj_covar=tensor([0.0226, 0.0263, 0.0299, 0.0353, 0.0238, 0.0240, 0.0255, 0.0380], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 06:01:25,600 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.991e+02 2.657e+02 3.113e+02 3.809e+02 6.399e+02, threshold=6.226e+02, percent-clipped=0.0 2023-05-16 06:01:29,948 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=171748.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 06:01:44,986 INFO [finetune.py:992] (0/2) Epoch 6, batch 10600, loss[loss=0.1371, simple_loss=0.213, pruned_loss=0.03055, over 11893.00 frames. ], tot_loss[loss=0.1725, simple_loss=0.2608, pruned_loss=0.04211, over 2373427.79 frames. ], batch size: 26, lr: 4.50e-03, grad_scale: 16.0 2023-05-16 06:02:05,089 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=171796.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 06:02:10,406 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.3375, 4.1162, 4.0569, 4.5372, 3.2048, 3.9445, 2.7271, 3.9968], device='cuda:0'), covar=tensor([0.1615, 0.0727, 0.1089, 0.0626, 0.1139, 0.0640, 0.1744, 0.1741], device='cuda:0'), in_proj_covar=tensor([0.0224, 0.0261, 0.0297, 0.0350, 0.0236, 0.0238, 0.0253, 0.0377], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 06:02:21,636 INFO [finetune.py:992] (0/2) Epoch 6, batch 10650, loss[loss=0.1788, simple_loss=0.2694, pruned_loss=0.04404, over 12115.00 frames. ], tot_loss[loss=0.1729, simple_loss=0.2611, pruned_loss=0.04235, over 2369806.33 frames. ], batch size: 33, lr: 4.50e-03, grad_scale: 16.0 2023-05-16 06:02:34,120 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.5721, 4.1833, 3.9442, 4.5608, 3.6618, 4.2294, 2.6331, 4.3367], device='cuda:0'), covar=tensor([0.1239, 0.0635, 0.1155, 0.0660, 0.0729, 0.0476, 0.1510, 0.1352], device='cuda:0'), in_proj_covar=tensor([0.0224, 0.0261, 0.0296, 0.0349, 0.0235, 0.0238, 0.0253, 0.0377], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 06:02:38,041 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.944e+02 2.731e+02 3.265e+02 4.072e+02 6.938e+02, threshold=6.531e+02, percent-clipped=2.0 2023-05-16 06:02:45,330 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=171852.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 06:02:46,596 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=171854.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 06:02:57,137 INFO [finetune.py:992] (0/2) Epoch 6, batch 10700, loss[loss=0.1743, simple_loss=0.2644, pruned_loss=0.04214, over 11669.00 frames. ], tot_loss[loss=0.1741, simple_loss=0.2622, pruned_loss=0.04302, over 2373905.62 frames. ], batch size: 48, lr: 4.50e-03, grad_scale: 16.0 2023-05-16 06:03:18,363 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.1095, 3.7459, 5.3490, 2.9175, 3.1428, 3.9796, 3.5643, 4.1188], device='cuda:0'), covar=tensor([0.0400, 0.0906, 0.0340, 0.1086, 0.1651, 0.1504, 0.1130, 0.0924], device='cuda:0'), in_proj_covar=tensor([0.0232, 0.0228, 0.0235, 0.0180, 0.0235, 0.0281, 0.0222, 0.0252], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:0') 2023-05-16 06:03:18,893 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=171900.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 06:03:33,719 INFO [finetune.py:992] (0/2) Epoch 6, batch 10750, loss[loss=0.171, simple_loss=0.2558, pruned_loss=0.04308, over 12032.00 frames. ], tot_loss[loss=0.1745, simple_loss=0.2623, pruned_loss=0.04335, over 2363752.10 frames. ], batch size: 31, lr: 4.50e-03, grad_scale: 16.0 2023-05-16 06:03:46,706 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=171937.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 06:03:50,062 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.851e+02 2.826e+02 3.370e+02 4.038e+02 8.088e+02, threshold=6.740e+02, percent-clipped=7.0 2023-05-16 06:04:09,487 INFO [finetune.py:992] (0/2) Epoch 6, batch 10800, loss[loss=0.1446, simple_loss=0.2347, pruned_loss=0.0272, over 12348.00 frames. ], tot_loss[loss=0.1743, simple_loss=0.2624, pruned_loss=0.04308, over 2365211.02 frames. ], batch size: 31, lr: 4.50e-03, grad_scale: 16.0 2023-05-16 06:04:14,561 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.8901, 4.8059, 4.7185, 4.8527, 3.7767, 4.9462, 5.0146, 5.0292], device='cuda:0'), covar=tensor([0.0259, 0.0186, 0.0198, 0.0351, 0.1274, 0.0346, 0.0183, 0.0210], device='cuda:0'), in_proj_covar=tensor([0.0183, 0.0188, 0.0183, 0.0235, 0.0233, 0.0207, 0.0168, 0.0216], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0005, 0.0004, 0.0005, 0.0004, 0.0005], device='cuda:0') 2023-05-16 06:04:30,105 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=171998.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 06:04:31,656 INFO [checkpoint.py:75] (0/2) Saving checkpoint to pruned_transducer_stateless7/exp_giga_finetune/checkpoint-72000.pt 2023-05-16 06:04:35,425 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=172001.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 06:04:48,139 INFO [finetune.py:992] (0/2) Epoch 6, batch 10850, loss[loss=0.1796, simple_loss=0.2745, pruned_loss=0.04235, over 11796.00 frames. ], tot_loss[loss=0.1741, simple_loss=0.2619, pruned_loss=0.04315, over 2370960.46 frames. ], batch size: 44, lr: 4.50e-03, grad_scale: 16.0 2023-05-16 06:05:05,512 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.678e+02 2.817e+02 3.464e+02 4.128e+02 6.888e+02, threshold=6.928e+02, percent-clipped=1.0 2023-05-16 06:05:10,649 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=172049.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 06:05:25,732 INFO [finetune.py:992] (0/2) Epoch 6, batch 10900, loss[loss=0.2944, simple_loss=0.343, pruned_loss=0.1229, over 7845.00 frames. ], tot_loss[loss=0.1755, simple_loss=0.263, pruned_loss=0.044, over 2362571.39 frames. ], batch size: 97, lr: 4.50e-03, grad_scale: 16.0 2023-05-16 06:06:00,423 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=172118.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 06:06:00,952 INFO [finetune.py:992] (0/2) Epoch 6, batch 10950, loss[loss=0.178, simple_loss=0.2756, pruned_loss=0.04021, over 12122.00 frames. ], tot_loss[loss=0.1764, simple_loss=0.2638, pruned_loss=0.04446, over 2364418.27 frames. ], batch size: 34, lr: 4.50e-03, grad_scale: 16.0 2023-05-16 06:06:07,702 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2023-05-16 06:06:17,104 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.080e+02 2.767e+02 3.256e+02 4.075e+02 8.444e+02, threshold=6.513e+02, percent-clipped=1.0 2023-05-16 06:06:25,646 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=172154.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 06:06:35,006 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.5544, 4.5384, 4.4449, 4.0930, 4.1322, 4.5447, 4.2274, 4.0438], device='cuda:0'), covar=tensor([0.0826, 0.0970, 0.0649, 0.1261, 0.2778, 0.0842, 0.1497, 0.1199], device='cuda:0'), in_proj_covar=tensor([0.0586, 0.0513, 0.0493, 0.0598, 0.0399, 0.0689, 0.0743, 0.0545], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0004, 0.0004, 0.0003], device='cuda:0') 2023-05-16 06:06:36,277 INFO [finetune.py:992] (0/2) Epoch 6, batch 11000, loss[loss=0.184, simple_loss=0.2767, pruned_loss=0.0457, over 11655.00 frames. ], tot_loss[loss=0.1789, simple_loss=0.2661, pruned_loss=0.04586, over 2346638.54 frames. ], batch size: 48, lr: 4.50e-03, grad_scale: 16.0 2023-05-16 06:06:43,979 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=172179.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 06:07:01,139 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=172202.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 06:07:12,988 INFO [finetune.py:992] (0/2) Epoch 6, batch 11050, loss[loss=0.2472, simple_loss=0.335, pruned_loss=0.07965, over 10174.00 frames. ], tot_loss[loss=0.1834, simple_loss=0.27, pruned_loss=0.04842, over 2308297.27 frames. ], batch size: 68, lr: 4.50e-03, grad_scale: 16.0 2023-05-16 06:07:24,863 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.3003, 3.2438, 3.2737, 3.6013, 2.6858, 3.2917, 2.6005, 3.0870], device='cuda:0'), covar=tensor([0.1402, 0.0776, 0.0857, 0.0618, 0.0920, 0.0630, 0.1446, 0.1009], device='cuda:0'), in_proj_covar=tensor([0.0222, 0.0259, 0.0294, 0.0348, 0.0234, 0.0235, 0.0252, 0.0373], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 06:07:28,818 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.158e+02 3.130e+02 3.953e+02 4.943e+02 1.066e+03, threshold=7.907e+02, percent-clipped=6.0 2023-05-16 06:07:47,799 INFO [finetune.py:992] (0/2) Epoch 6, batch 11100, loss[loss=0.1594, simple_loss=0.2495, pruned_loss=0.03465, over 12097.00 frames. ], tot_loss[loss=0.186, simple_loss=0.2726, pruned_loss=0.0497, over 2272568.00 frames. ], batch size: 33, lr: 4.50e-03, grad_scale: 16.0 2023-05-16 06:08:05,782 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=172293.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 06:08:24,082 INFO [finetune.py:992] (0/2) Epoch 6, batch 11150, loss[loss=0.2304, simple_loss=0.3162, pruned_loss=0.07233, over 10360.00 frames. ], tot_loss[loss=0.1933, simple_loss=0.2789, pruned_loss=0.0538, over 2218654.29 frames. ], batch size: 68, lr: 4.50e-03, grad_scale: 16.0 2023-05-16 06:08:40,182 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.396e+02 3.335e+02 4.017e+02 5.044e+02 1.045e+03, threshold=8.034e+02, percent-clipped=3.0 2023-05-16 06:08:55,084 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.9992, 3.0789, 4.4266, 2.4629, 2.7022, 3.5852, 3.0565, 3.7239], device='cuda:0'), covar=tensor([0.0496, 0.1119, 0.0203, 0.1230, 0.1790, 0.1097, 0.1268, 0.0720], device='cuda:0'), in_proj_covar=tensor([0.0229, 0.0226, 0.0232, 0.0177, 0.0233, 0.0276, 0.0219, 0.0248], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:0') 2023-05-16 06:08:56,516 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 2023-05-16 06:09:00,372 INFO [finetune.py:992] (0/2) Epoch 6, batch 11200, loss[loss=0.1689, simple_loss=0.2638, pruned_loss=0.037, over 12147.00 frames. ], tot_loss[loss=0.2, simple_loss=0.2844, pruned_loss=0.05782, over 2151241.73 frames. ], batch size: 36, lr: 4.50e-03, grad_scale: 16.0 2023-05-16 06:09:22,974 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.0102, 2.0039, 2.5244, 3.0125, 2.1870, 3.1280, 3.0186, 3.1482], device='cuda:0'), covar=tensor([0.0145, 0.1100, 0.0450, 0.0139, 0.0957, 0.0223, 0.0294, 0.0113], device='cuda:0'), in_proj_covar=tensor([0.0118, 0.0198, 0.0182, 0.0112, 0.0184, 0.0172, 0.0167, 0.0118], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 06:09:35,113 INFO [finetune.py:992] (0/2) Epoch 6, batch 11250, loss[loss=0.3265, simple_loss=0.3692, pruned_loss=0.1419, over 6324.00 frames. ], tot_loss[loss=0.209, simple_loss=0.2921, pruned_loss=0.06299, over 2078466.49 frames. ], batch size: 99, lr: 4.50e-03, grad_scale: 16.0 2023-05-16 06:09:36,314 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.56 vs. limit=2.0 2023-05-16 06:09:42,174 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([1.9120, 2.2119, 2.1893, 2.1753, 1.9938, 1.9902, 2.0453, 1.7481], device='cuda:0'), covar=tensor([0.0265, 0.0123, 0.0123, 0.0181, 0.0270, 0.0181, 0.0162, 0.0330], device='cuda:0'), in_proj_covar=tensor([0.0184, 0.0151, 0.0147, 0.0176, 0.0192, 0.0187, 0.0156, 0.0194], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], device='cuda:0') 2023-05-16 06:09:51,989 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.255e+02 3.510e+02 4.248e+02 5.135e+02 1.202e+03, threshold=8.497e+02, percent-clipped=3.0 2023-05-16 06:10:10,375 INFO [finetune.py:992] (0/2) Epoch 6, batch 11300, loss[loss=0.247, simple_loss=0.3295, pruned_loss=0.08228, over 10131.00 frames. ], tot_loss[loss=0.2166, simple_loss=0.2989, pruned_loss=0.06716, over 2024166.52 frames. ], batch size: 68, lr: 4.50e-03, grad_scale: 16.0 2023-05-16 06:10:13,967 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=172474.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 06:10:40,792 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.87 vs. limit=2.0 2023-05-16 06:10:45,716 INFO [finetune.py:992] (0/2) Epoch 6, batch 11350, loss[loss=0.27, simple_loss=0.3369, pruned_loss=0.1015, over 7012.00 frames. ], tot_loss[loss=0.2219, simple_loss=0.3034, pruned_loss=0.07018, over 1971648.26 frames. ], batch size: 99, lr: 4.50e-03, grad_scale: 16.0 2023-05-16 06:10:52,892 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2023-05-16 06:11:01,337 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.533e+02 3.462e+02 4.203e+02 5.363e+02 9.829e+02, threshold=8.405e+02, percent-clipped=2.0 2023-05-16 06:11:20,365 INFO [finetune.py:992] (0/2) Epoch 6, batch 11400, loss[loss=0.196, simple_loss=0.2832, pruned_loss=0.05436, over 12214.00 frames. ], tot_loss[loss=0.2267, simple_loss=0.3073, pruned_loss=0.07303, over 1922644.76 frames. ], batch size: 35, lr: 4.50e-03, grad_scale: 16.0 2023-05-16 06:11:29,269 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.8280, 4.3642, 3.9692, 4.0966, 4.4486, 3.8894, 4.1164, 3.9531], device='cuda:0'), covar=tensor([0.1436, 0.0981, 0.1399, 0.1525, 0.0979, 0.1845, 0.1429, 0.1214], device='cuda:0'), in_proj_covar=tensor([0.0335, 0.0456, 0.0361, 0.0410, 0.0439, 0.0416, 0.0373, 0.0357], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 06:11:37,256 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=172593.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 06:11:40,833 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=172598.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 06:11:55,219 INFO [finetune.py:992] (0/2) Epoch 6, batch 11450, loss[loss=0.1816, simple_loss=0.2774, pruned_loss=0.0429, over 12250.00 frames. ], tot_loss[loss=0.2323, simple_loss=0.3115, pruned_loss=0.0765, over 1876490.79 frames. ], batch size: 32, lr: 4.49e-03, grad_scale: 16.0 2023-05-16 06:11:59,377 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=172625.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 06:12:11,420 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=172641.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 06:12:11,962 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.535e+02 3.484e+02 4.035e+02 4.703e+02 1.176e+03, threshold=8.069e+02, percent-clipped=1.0 2023-05-16 06:12:23,761 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=172659.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 06:12:30,344 INFO [finetune.py:992] (0/2) Epoch 6, batch 11500, loss[loss=0.2202, simple_loss=0.2952, pruned_loss=0.0726, over 11411.00 frames. ], tot_loss[loss=0.2367, simple_loss=0.3145, pruned_loss=0.0794, over 1842158.26 frames. ], batch size: 48, lr: 4.49e-03, grad_scale: 16.0 2023-05-16 06:12:31,156 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.6053, 4.3738, 4.6089, 4.1646, 4.3953, 4.2037, 4.5662, 4.1905], device='cuda:0'), covar=tensor([0.0295, 0.0322, 0.0285, 0.0260, 0.0354, 0.0307, 0.0273, 0.0580], device='cuda:0'), in_proj_covar=tensor([0.0229, 0.0231, 0.0254, 0.0232, 0.0226, 0.0227, 0.0209, 0.0192], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 06:12:41,974 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=172686.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 06:13:05,364 INFO [finetune.py:992] (0/2) Epoch 6, batch 11550, loss[loss=0.2049, simple_loss=0.2967, pruned_loss=0.05657, over 11147.00 frames. ], tot_loss[loss=0.239, simple_loss=0.3163, pruned_loss=0.08088, over 1828814.64 frames. ], batch size: 55, lr: 4.49e-03, grad_scale: 16.0 2023-05-16 06:13:12,157 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.5144, 4.5135, 4.3629, 4.0746, 4.1588, 4.4970, 4.2263, 4.0472], device='cuda:0'), covar=tensor([0.0712, 0.0809, 0.0722, 0.1258, 0.1722, 0.0745, 0.1447, 0.1104], device='cuda:0'), in_proj_covar=tensor([0.0555, 0.0487, 0.0469, 0.0568, 0.0380, 0.0648, 0.0696, 0.0517], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 06:13:15,636 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=172734.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 06:13:16,958 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.2365, 2.3622, 3.7524, 3.1115, 3.5613, 3.4001, 2.4241, 3.5986], device='cuda:0'), covar=tensor([0.0094, 0.0333, 0.0074, 0.0212, 0.0103, 0.0110, 0.0320, 0.0086], device='cuda:0'), in_proj_covar=tensor([0.0160, 0.0192, 0.0172, 0.0168, 0.0194, 0.0147, 0.0182, 0.0163], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 06:13:20,748 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.324e+02 3.468e+02 4.001e+02 4.655e+02 7.886e+02, threshold=8.001e+02, percent-clipped=0.0 2023-05-16 06:13:33,354 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.72 vs. limit=2.0 2023-05-16 06:13:40,052 INFO [finetune.py:992] (0/2) Epoch 6, batch 11600, loss[loss=0.259, simple_loss=0.337, pruned_loss=0.09056, over 6752.00 frames. ], tot_loss[loss=0.2411, simple_loss=0.3175, pruned_loss=0.08237, over 1799386.13 frames. ], batch size: 98, lr: 4.49e-03, grad_scale: 16.0 2023-05-16 06:13:43,503 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=172774.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 06:13:45,684 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2023-05-16 06:13:57,772 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=172795.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 06:14:16,291 INFO [finetune.py:992] (0/2) Epoch 6, batch 11650, loss[loss=0.2843, simple_loss=0.3429, pruned_loss=0.1128, over 7098.00 frames. ], tot_loss[loss=0.2403, simple_loss=0.3164, pruned_loss=0.08207, over 1792832.47 frames. ], batch size: 97, lr: 4.49e-03, grad_scale: 16.0 2023-05-16 06:14:18,720 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=172822.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 06:14:32,712 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.738e+02 3.666e+02 4.160e+02 4.980e+02 7.102e+02, threshold=8.321e+02, percent-clipped=0.0 2023-05-16 06:14:34,949 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=172845.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 06:14:49,240 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=4.50 vs. limit=5.0 2023-05-16 06:14:50,850 INFO [finetune.py:992] (0/2) Epoch 6, batch 11700, loss[loss=0.2257, simple_loss=0.3077, pruned_loss=0.07187, over 10394.00 frames. ], tot_loss[loss=0.2408, simple_loss=0.3165, pruned_loss=0.08253, over 1780441.68 frames. ], batch size: 68, lr: 4.49e-03, grad_scale: 16.0 2023-05-16 06:14:54,952 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.7185, 2.0366, 2.8128, 3.7188, 2.1799, 3.8356, 3.6658, 3.8482], device='cuda:0'), covar=tensor([0.0126, 0.1428, 0.0462, 0.0119, 0.1380, 0.0176, 0.0215, 0.0097], device='cuda:0'), in_proj_covar=tensor([0.0115, 0.0196, 0.0179, 0.0111, 0.0182, 0.0168, 0.0163, 0.0116], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 06:14:56,945 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.9087, 2.1997, 2.6946, 2.8334, 2.8221, 2.9270, 2.7956, 2.4110], device='cuda:0'), covar=tensor([0.0063, 0.0328, 0.0149, 0.0062, 0.0100, 0.0082, 0.0112, 0.0336], device='cuda:0'), in_proj_covar=tensor([0.0077, 0.0115, 0.0097, 0.0071, 0.0095, 0.0107, 0.0087, 0.0130], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 06:15:17,222 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=172906.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 06:15:25,691 INFO [finetune.py:992] (0/2) Epoch 6, batch 11750, loss[loss=0.2551, simple_loss=0.3197, pruned_loss=0.09521, over 6681.00 frames. ], tot_loss[loss=0.2417, simple_loss=0.3166, pruned_loss=0.0834, over 1742637.78 frames. ], batch size: 97, lr: 4.49e-03, grad_scale: 16.0 2023-05-16 06:15:42,078 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.367e+02 3.625e+02 4.400e+02 5.167e+02 1.076e+03, threshold=8.801e+02, percent-clipped=4.0 2023-05-16 06:15:50,872 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=172954.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 06:16:00,827 INFO [finetune.py:992] (0/2) Epoch 6, batch 11800, loss[loss=0.2778, simple_loss=0.3342, pruned_loss=0.1107, over 6773.00 frames. ], tot_loss[loss=0.245, simple_loss=0.3193, pruned_loss=0.08531, over 1721163.37 frames. ], batch size: 98, lr: 4.49e-03, grad_scale: 16.0 2023-05-16 06:16:03,217 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.5731, 2.6958, 4.2766, 4.4863, 3.0449, 2.5791, 2.7703, 1.8504], device='cuda:0'), covar=tensor([0.1651, 0.3090, 0.0501, 0.0388, 0.1180, 0.2488, 0.2780, 0.4889], device='cuda:0'), in_proj_covar=tensor([0.0271, 0.0361, 0.0256, 0.0281, 0.0249, 0.0280, 0.0350, 0.0351], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-16 06:16:09,220 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=172981.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 06:16:36,307 INFO [finetune.py:992] (0/2) Epoch 6, batch 11850, loss[loss=0.2839, simple_loss=0.3433, pruned_loss=0.1122, over 6873.00 frames. ], tot_loss[loss=0.2464, simple_loss=0.3211, pruned_loss=0.08583, over 1710385.35 frames. ], batch size: 98, lr: 4.49e-03, grad_scale: 16.0 2023-05-16 06:16:51,795 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.651e+02 3.587e+02 4.059e+02 5.000e+02 7.524e+02, threshold=8.117e+02, percent-clipped=0.0 2023-05-16 06:17:10,809 INFO [finetune.py:992] (0/2) Epoch 6, batch 11900, loss[loss=0.1959, simple_loss=0.2897, pruned_loss=0.05104, over 10268.00 frames. ], tot_loss[loss=0.246, simple_loss=0.3209, pruned_loss=0.08552, over 1692049.41 frames. ], batch size: 68, lr: 4.49e-03, grad_scale: 16.0 2023-05-16 06:17:12,376 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.7468, 3.5055, 3.5910, 3.6462, 3.6390, 3.7236, 3.7147, 2.7002], device='cuda:0'), covar=tensor([0.0084, 0.0095, 0.0130, 0.0083, 0.0071, 0.0114, 0.0090, 0.0653], device='cuda:0'), in_proj_covar=tensor([0.0061, 0.0069, 0.0072, 0.0065, 0.0054, 0.0082, 0.0070, 0.0086], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-16 06:17:21,147 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.9136, 3.7715, 3.9089, 3.6761, 3.7871, 3.6715, 3.8684, 3.5503], device='cuda:0'), covar=tensor([0.0331, 0.0329, 0.0334, 0.0244, 0.0330, 0.0279, 0.0299, 0.1152], device='cuda:0'), in_proj_covar=tensor([0.0222, 0.0224, 0.0245, 0.0224, 0.0219, 0.0220, 0.0200, 0.0185], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 06:17:25,878 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=173090.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 06:17:41,990 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.6063, 2.5185, 4.0039, 4.1113, 2.9634, 2.5982, 2.6125, 2.0066], device='cuda:0'), covar=tensor([0.1517, 0.2907, 0.0488, 0.0429, 0.1085, 0.2251, 0.2872, 0.4504], device='cuda:0'), in_proj_covar=tensor([0.0270, 0.0359, 0.0256, 0.0280, 0.0247, 0.0279, 0.0350, 0.0351], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-16 06:17:45,699 INFO [finetune.py:992] (0/2) Epoch 6, batch 11950, loss[loss=0.1916, simple_loss=0.2811, pruned_loss=0.05105, over 11531.00 frames. ], tot_loss[loss=0.2413, simple_loss=0.3173, pruned_loss=0.08266, over 1679575.42 frames. ], batch size: 48, lr: 4.49e-03, grad_scale: 16.0 2023-05-16 06:18:01,877 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.108e+02 3.189e+02 3.702e+02 4.307e+02 7.149e+02, threshold=7.404e+02, percent-clipped=0.0 2023-05-16 06:18:07,494 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=173150.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 06:18:19,099 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 2023-05-16 06:18:20,086 INFO [finetune.py:992] (0/2) Epoch 6, batch 12000, loss[loss=0.2195, simple_loss=0.2977, pruned_loss=0.0707, over 6934.00 frames. ], tot_loss[loss=0.2349, simple_loss=0.3124, pruned_loss=0.07866, over 1675317.15 frames. ], batch size: 99, lr: 4.49e-03, grad_scale: 16.0 2023-05-16 06:18:20,087 INFO [finetune.py:1017] (0/2) Computing validation loss 2023-05-16 06:18:38,370 INFO [finetune.py:1026] (0/2) Epoch 6, validation: loss=0.2919, simple_loss=0.3683, pruned_loss=0.1078, over 1020973.00 frames. 2023-05-16 06:18:38,371 INFO [finetune.py:1027] (0/2) Maximum memory allocated so far is 12745MB 2023-05-16 06:18:46,150 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.7348, 2.3867, 3.5254, 3.5826, 2.9066, 2.6494, 2.5935, 2.2576], device='cuda:0'), covar=tensor([0.1208, 0.2863, 0.0625, 0.0472, 0.0891, 0.2005, 0.2448, 0.3600], device='cuda:0'), in_proj_covar=tensor([0.0267, 0.0355, 0.0253, 0.0277, 0.0244, 0.0276, 0.0346, 0.0347], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-16 06:18:51,724 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=2.10 vs. limit=2.0 2023-05-16 06:19:00,960 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=173201.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 06:19:08,003 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=173211.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 06:19:13,097 INFO [finetune.py:992] (0/2) Epoch 6, batch 12050, loss[loss=0.2201, simple_loss=0.2994, pruned_loss=0.07039, over 7213.00 frames. ], tot_loss[loss=0.2283, simple_loss=0.3072, pruned_loss=0.07467, over 1682581.76 frames. ], batch size: 98, lr: 4.49e-03, grad_scale: 16.0 2023-05-16 06:19:29,348 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.293e+02 2.981e+02 3.503e+02 4.122e+02 6.105e+02, threshold=7.005e+02, percent-clipped=0.0 2023-05-16 06:19:36,842 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=173254.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 06:19:46,372 INFO [finetune.py:992] (0/2) Epoch 6, batch 12100, loss[loss=0.1877, simple_loss=0.2921, pruned_loss=0.04169, over 10229.00 frames. ], tot_loss[loss=0.2264, simple_loss=0.306, pruned_loss=0.07334, over 1689450.73 frames. ], batch size: 68, lr: 4.49e-03, grad_scale: 16.0 2023-05-16 06:19:52,277 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.5993, 5.1842, 4.9055, 4.8339, 5.2430, 4.7073, 4.8662, 4.7714], device='cuda:0'), covar=tensor([0.1225, 0.0896, 0.0909, 0.1699, 0.0804, 0.1545, 0.1523, 0.1012], device='cuda:0'), in_proj_covar=tensor([0.0318, 0.0428, 0.0341, 0.0386, 0.0411, 0.0390, 0.0348, 0.0332], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002], device='cuda:0') 2023-05-16 06:19:54,112 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=173281.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 06:20:07,475 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=173302.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 06:20:12,076 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=173309.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 06:20:18,250 INFO [finetune.py:992] (0/2) Epoch 6, batch 12150, loss[loss=0.2651, simple_loss=0.3293, pruned_loss=0.1004, over 6840.00 frames. ], tot_loss[loss=0.2265, simple_loss=0.3064, pruned_loss=0.07326, over 1705223.28 frames. ], batch size: 100, lr: 4.49e-03, grad_scale: 32.0 2023-05-16 06:20:24,584 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=173329.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 06:20:32,637 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.014e+02 3.251e+02 3.790e+02 4.347e+02 7.490e+02, threshold=7.581e+02, percent-clipped=1.0 2023-05-16 06:20:49,397 INFO [finetune.py:992] (0/2) Epoch 6, batch 12200, loss[loss=0.285, simple_loss=0.336, pruned_loss=0.117, over 7003.00 frames. ], tot_loss[loss=0.2279, simple_loss=0.3074, pruned_loss=0.07416, over 1696157.72 frames. ], batch size: 98, lr: 4.49e-03, grad_scale: 32.0 2023-05-16 06:20:50,194 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=173370.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 06:20:56,125 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.8981, 4.4984, 4.1479, 4.1493, 4.5400, 4.1591, 4.2276, 4.1014], device='cuda:0'), covar=tensor([0.1820, 0.1139, 0.1280, 0.2248, 0.1025, 0.1977, 0.1546, 0.1169], device='cuda:0'), in_proj_covar=tensor([0.0321, 0.0430, 0.0344, 0.0391, 0.0415, 0.0392, 0.0349, 0.0334], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002], device='cuda:0') 2023-05-16 06:21:02,920 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=173390.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 06:21:12,041 INFO [checkpoint.py:75] (0/2) Saving checkpoint to pruned_transducer_stateless7/exp_giga_finetune/epoch-6.pt 2023-05-16 06:21:37,542 INFO [finetune.py:992] (0/2) Epoch 7, batch 0, loss[loss=0.1561, simple_loss=0.2472, pruned_loss=0.03248, over 12354.00 frames. ], tot_loss[loss=0.1561, simple_loss=0.2472, pruned_loss=0.03248, over 12354.00 frames. ], batch size: 30, lr: 4.49e-03, grad_scale: 32.0 2023-05-16 06:21:37,542 INFO [finetune.py:1017] (0/2) Computing validation loss 2023-05-16 06:21:55,240 INFO [finetune.py:1026] (0/2) Epoch 7, validation: loss=0.2955, simple_loss=0.3679, pruned_loss=0.1116, over 1020973.00 frames. 2023-05-16 06:21:55,241 INFO [finetune.py:1027] (0/2) Maximum memory allocated so far is 12745MB 2023-05-16 06:22:20,786 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=173438.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 06:22:23,613 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.092e+02 3.197e+02 3.790e+02 4.651e+02 1.246e+03, threshold=7.581e+02, percent-clipped=2.0 2023-05-16 06:22:31,487 INFO [finetune.py:992] (0/2) Epoch 7, batch 50, loss[loss=0.1851, simple_loss=0.2785, pruned_loss=0.04588, over 12096.00 frames. ], tot_loss[loss=0.1832, simple_loss=0.271, pruned_loss=0.04769, over 537547.25 frames. ], batch size: 32, lr: 4.48e-03, grad_scale: 32.0 2023-05-16 06:22:35,152 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([6.2267, 6.1500, 5.9793, 5.4266, 5.0644, 6.0931, 5.7778, 5.5405], device='cuda:0'), covar=tensor([0.0606, 0.0919, 0.0656, 0.1433, 0.0736, 0.0680, 0.1433, 0.1108], device='cuda:0'), in_proj_covar=tensor([0.0539, 0.0479, 0.0456, 0.0551, 0.0366, 0.0624, 0.0673, 0.0503], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002], device='cuda:0') 2023-05-16 06:22:54,818 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=4.57 vs. limit=5.0 2023-05-16 06:22:59,439 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=173491.0, num_to_drop=1, layers_to_drop={1} 2023-05-16 06:23:06,386 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=173501.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 06:23:07,713 INFO [finetune.py:992] (0/2) Epoch 7, batch 100, loss[loss=0.2114, simple_loss=0.3057, pruned_loss=0.0586, over 12125.00 frames. ], tot_loss[loss=0.1819, simple_loss=0.2705, pruned_loss=0.04662, over 950031.89 frames. ], batch size: 38, lr: 4.48e-03, grad_scale: 16.0 2023-05-16 06:23:09,911 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=173506.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 06:23:29,068 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([6.1795, 6.0690, 5.9259, 5.3466, 5.1390, 6.0661, 5.7122, 5.4814], device='cuda:0'), covar=tensor([0.0733, 0.1167, 0.0731, 0.1671, 0.0617, 0.0752, 0.1553, 0.1086], device='cuda:0'), in_proj_covar=tensor([0.0546, 0.0485, 0.0460, 0.0560, 0.0370, 0.0632, 0.0683, 0.0509], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002], device='cuda:0') 2023-05-16 06:23:35,919 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.898e+02 2.918e+02 3.480e+02 4.159e+02 1.047e+03, threshold=6.960e+02, percent-clipped=3.0 2023-05-16 06:23:40,257 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=173549.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 06:23:42,531 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=173552.0, num_to_drop=1, layers_to_drop={2} 2023-05-16 06:23:42,993 INFO [finetune.py:992] (0/2) Epoch 7, batch 150, loss[loss=0.1776, simple_loss=0.2685, pruned_loss=0.04337, over 12083.00 frames. ], tot_loss[loss=0.181, simple_loss=0.2694, pruned_loss=0.04634, over 1268358.88 frames. ], batch size: 42, lr: 4.48e-03, grad_scale: 16.0 2023-05-16 06:24:19,661 INFO [finetune.py:992] (0/2) Epoch 7, batch 200, loss[loss=0.2037, simple_loss=0.2914, pruned_loss=0.05796, over 12118.00 frames. ], tot_loss[loss=0.1789, simple_loss=0.2673, pruned_loss=0.04525, over 1511669.24 frames. ], batch size: 39, lr: 4.48e-03, grad_scale: 16.0 2023-05-16 06:24:27,584 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=173614.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 06:24:45,352 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.2104, 4.5485, 3.8453, 4.8895, 4.2684, 2.7998, 4.2779, 2.8109], device='cuda:0'), covar=tensor([0.0785, 0.0806, 0.1482, 0.0346, 0.1206, 0.1701, 0.0914, 0.3871], device='cuda:0'), in_proj_covar=tensor([0.0290, 0.0357, 0.0335, 0.0256, 0.0346, 0.0257, 0.0325, 0.0349], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 06:24:48,600 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.259e+02 2.761e+02 3.100e+02 3.797e+02 5.593e+02, threshold=6.201e+02, percent-clipped=0.0 2023-05-16 06:24:49,574 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.3785, 2.4865, 3.3565, 4.2823, 3.7901, 4.1330, 3.7167, 2.7087], device='cuda:0'), covar=tensor([0.0029, 0.0381, 0.0161, 0.0039, 0.0124, 0.0082, 0.0110, 0.0421], device='cuda:0'), in_proj_covar=tensor([0.0078, 0.0117, 0.0098, 0.0072, 0.0098, 0.0110, 0.0089, 0.0133], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 06:24:55,734 INFO [finetune.py:992] (0/2) Epoch 7, batch 250, loss[loss=0.1591, simple_loss=0.2391, pruned_loss=0.03952, over 12262.00 frames. ], tot_loss[loss=0.178, simple_loss=0.2662, pruned_loss=0.0449, over 1705070.93 frames. ], batch size: 32, lr: 4.48e-03, grad_scale: 16.0 2023-05-16 06:25:04,503 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=173665.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 06:25:11,681 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=173675.0, num_to_drop=1, layers_to_drop={0} 2023-05-16 06:25:12,399 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.3058, 4.6646, 2.8458, 2.6505, 4.0641, 2.7100, 3.9308, 3.3048], device='cuda:0'), covar=tensor([0.0657, 0.0472, 0.1118, 0.1460, 0.0233, 0.1227, 0.0480, 0.0754], device='cuda:0'), in_proj_covar=tensor([0.0177, 0.0231, 0.0168, 0.0190, 0.0129, 0.0172, 0.0183, 0.0165], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 06:25:15,197 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.4795, 3.4821, 3.2856, 3.1958, 2.8360, 2.7646, 3.5702, 2.3730], device='cuda:0'), covar=tensor([0.0379, 0.0125, 0.0183, 0.0190, 0.0391, 0.0361, 0.0117, 0.0457], device='cuda:0'), in_proj_covar=tensor([0.0180, 0.0146, 0.0142, 0.0173, 0.0188, 0.0183, 0.0150, 0.0189], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], device='cuda:0') 2023-05-16 06:25:25,013 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.0212, 2.3422, 3.4761, 3.0464, 3.3331, 3.1990, 2.2988, 3.3839], device='cuda:0'), covar=tensor([0.0119, 0.0332, 0.0145, 0.0199, 0.0141, 0.0145, 0.0348, 0.0129], device='cuda:0'), in_proj_covar=tensor([0.0158, 0.0189, 0.0167, 0.0165, 0.0190, 0.0145, 0.0180, 0.0160], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 06:25:31,206 INFO [finetune.py:992] (0/2) Epoch 7, batch 300, loss[loss=0.1866, simple_loss=0.2724, pruned_loss=0.05041, over 11621.00 frames. ], tot_loss[loss=0.178, simple_loss=0.2664, pruned_loss=0.04486, over 1856965.13 frames. ], batch size: 48, lr: 4.48e-03, grad_scale: 16.0 2023-05-16 06:25:55,830 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.9033, 4.5946, 4.6403, 4.8817, 4.7013, 4.8071, 4.7423, 2.3954], device='cuda:0'), covar=tensor([0.0101, 0.0065, 0.0101, 0.0054, 0.0048, 0.0093, 0.0070, 0.0877], device='cuda:0'), in_proj_covar=tensor([0.0062, 0.0071, 0.0075, 0.0067, 0.0056, 0.0085, 0.0072, 0.0089], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-16 06:26:00,436 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.313e+02 2.879e+02 3.329e+02 4.149e+02 7.087e+02, threshold=6.658e+02, percent-clipped=4.0 2023-05-16 06:26:06,396 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.0835, 2.2232, 3.0550, 4.0115, 2.2999, 4.1089, 3.9817, 4.2390], device='cuda:0'), covar=tensor([0.0109, 0.1285, 0.0465, 0.0121, 0.1249, 0.0214, 0.0194, 0.0081], device='cuda:0'), in_proj_covar=tensor([0.0115, 0.0197, 0.0177, 0.0110, 0.0184, 0.0166, 0.0162, 0.0115], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 06:26:07,591 INFO [finetune.py:992] (0/2) Epoch 7, batch 350, loss[loss=0.1585, simple_loss=0.2412, pruned_loss=0.03793, over 12302.00 frames. ], tot_loss[loss=0.1764, simple_loss=0.2645, pruned_loss=0.04414, over 1976007.53 frames. ], batch size: 28, lr: 4.48e-03, grad_scale: 16.0 2023-05-16 06:26:11,893 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.5746, 2.4930, 3.3113, 4.5341, 2.2728, 4.5989, 4.5322, 4.7815], device='cuda:0'), covar=tensor([0.0125, 0.1252, 0.0410, 0.0116, 0.1326, 0.0152, 0.0129, 0.0073], device='cuda:0'), in_proj_covar=tensor([0.0115, 0.0197, 0.0177, 0.0109, 0.0183, 0.0166, 0.0162, 0.0115], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 06:26:28,353 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.43 vs. limit=2.0 2023-05-16 06:26:36,261 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.98 vs. limit=2.0 2023-05-16 06:26:43,880 INFO [finetune.py:992] (0/2) Epoch 7, batch 400, loss[loss=0.1767, simple_loss=0.2553, pruned_loss=0.04903, over 12193.00 frames. ], tot_loss[loss=0.1764, simple_loss=0.2644, pruned_loss=0.04425, over 2065972.05 frames. ], batch size: 29, lr: 4.48e-03, grad_scale: 16.0 2023-05-16 06:26:46,086 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=173806.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 06:27:12,084 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.093e+02 2.932e+02 3.449e+02 4.062e+02 1.100e+03, threshold=6.898e+02, percent-clipped=4.0 2023-05-16 06:27:15,104 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=173847.0, num_to_drop=1, layers_to_drop={1} 2023-05-16 06:27:19,203 INFO [finetune.py:992] (0/2) Epoch 7, batch 450, loss[loss=0.1919, simple_loss=0.2758, pruned_loss=0.05395, over 12114.00 frames. ], tot_loss[loss=0.1767, simple_loss=0.2646, pruned_loss=0.04443, over 2136101.64 frames. ], batch size: 33, lr: 4.48e-03, grad_scale: 16.0 2023-05-16 06:27:19,916 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=173854.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 06:27:50,159 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.9837, 3.8699, 3.9367, 3.9744, 3.7109, 3.8346, 3.6597, 3.9462], device='cuda:0'), covar=tensor([0.0836, 0.0875, 0.1138, 0.0836, 0.2191, 0.1387, 0.0717, 0.1045], device='cuda:0'), in_proj_covar=tensor([0.0481, 0.0635, 0.0545, 0.0571, 0.0758, 0.0683, 0.0508, 0.0458], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0005, 0.0005, 0.0003, 0.0003], device='cuda:0') 2023-05-16 06:27:54,960 INFO [finetune.py:992] (0/2) Epoch 7, batch 500, loss[loss=0.1857, simple_loss=0.2768, pruned_loss=0.04731, over 12363.00 frames. ], tot_loss[loss=0.1767, simple_loss=0.2642, pruned_loss=0.04456, over 2184357.05 frames. ], batch size: 35, lr: 4.48e-03, grad_scale: 16.0 2023-05-16 06:28:23,982 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.896e+02 2.971e+02 3.357e+02 3.784e+02 6.011e+02, threshold=6.715e+02, percent-clipped=0.0 2023-05-16 06:28:30,232 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.33 vs. limit=2.0 2023-05-16 06:28:31,215 INFO [finetune.py:992] (0/2) Epoch 7, batch 550, loss[loss=0.1806, simple_loss=0.2749, pruned_loss=0.0432, over 12152.00 frames. ], tot_loss[loss=0.1767, simple_loss=0.2649, pruned_loss=0.04425, over 2227799.86 frames. ], batch size: 36, lr: 4.48e-03, grad_scale: 16.0 2023-05-16 06:28:39,872 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=173965.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 06:28:43,416 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=173970.0, num_to_drop=1, layers_to_drop={3} 2023-05-16 06:29:04,839 INFO [checkpoint.py:75] (0/2) Saving checkpoint to pruned_transducer_stateless7/exp_giga_finetune/checkpoint-74000.pt 2023-05-16 06:29:08,007 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.2416, 5.1979, 5.0890, 5.0776, 4.8011, 5.1540, 5.2392, 5.4241], device='cuda:0'), covar=tensor([0.0247, 0.0150, 0.0152, 0.0318, 0.0698, 0.0320, 0.0136, 0.0152], device='cuda:0'), in_proj_covar=tensor([0.0174, 0.0177, 0.0174, 0.0222, 0.0221, 0.0197, 0.0160, 0.0205], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0005, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:0') 2023-05-16 06:29:09,944 INFO [finetune.py:992] (0/2) Epoch 7, batch 600, loss[loss=0.1866, simple_loss=0.2834, pruned_loss=0.04488, over 11168.00 frames. ], tot_loss[loss=0.1765, simple_loss=0.2647, pruned_loss=0.04416, over 2253588.72 frames. ], batch size: 55, lr: 4.48e-03, grad_scale: 16.0 2023-05-16 06:29:17,303 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=174013.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 06:29:18,850 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=174015.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 06:29:20,200 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=174017.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 06:29:39,004 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.027e+02 2.752e+02 3.364e+02 3.945e+02 6.201e+02, threshold=6.728e+02, percent-clipped=0.0 2023-05-16 06:29:46,141 INFO [finetune.py:992] (0/2) Epoch 7, batch 650, loss[loss=0.1699, simple_loss=0.26, pruned_loss=0.03991, over 12295.00 frames. ], tot_loss[loss=0.1754, simple_loss=0.2638, pruned_loss=0.04344, over 2289266.35 frames. ], batch size: 34, lr: 4.48e-03, grad_scale: 16.0 2023-05-16 06:30:03,409 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=174076.0, num_to_drop=1, layers_to_drop={0} 2023-05-16 06:30:04,737 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=174078.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 06:30:22,238 INFO [finetune.py:992] (0/2) Epoch 7, batch 700, loss[loss=0.1751, simple_loss=0.271, pruned_loss=0.03958, over 12119.00 frames. ], tot_loss[loss=0.1761, simple_loss=0.2644, pruned_loss=0.04384, over 2296003.78 frames. ], batch size: 33, lr: 4.48e-03, grad_scale: 16.0 2023-05-16 06:30:34,530 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.35 vs. limit=2.0 2023-05-16 06:30:47,760 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=174139.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 06:30:50,940 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.236e+02 2.951e+02 3.360e+02 4.029e+02 6.160e+02, threshold=6.721e+02, percent-clipped=0.0 2023-05-16 06:30:52,986 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=174147.0, num_to_drop=1, layers_to_drop={2} 2023-05-16 06:30:57,125 INFO [finetune.py:992] (0/2) Epoch 7, batch 750, loss[loss=0.1865, simple_loss=0.2748, pruned_loss=0.04912, over 12039.00 frames. ], tot_loss[loss=0.176, simple_loss=0.2643, pruned_loss=0.04382, over 2310183.16 frames. ], batch size: 40, lr: 4.48e-03, grad_scale: 8.0 2023-05-16 06:31:06,203 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.6015, 2.6555, 3.4073, 4.4738, 2.3398, 4.3999, 4.5709, 4.7542], device='cuda:0'), covar=tensor([0.0140, 0.1190, 0.0402, 0.0141, 0.1377, 0.0216, 0.0160, 0.0090], device='cuda:0'), in_proj_covar=tensor([0.0120, 0.0205, 0.0184, 0.0115, 0.0190, 0.0174, 0.0171, 0.0119], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 06:31:27,572 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=174195.0, num_to_drop=1, layers_to_drop={1} 2023-05-16 06:31:31,574 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=174200.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 06:31:33,590 INFO [finetune.py:992] (0/2) Epoch 7, batch 800, loss[loss=0.1551, simple_loss=0.2406, pruned_loss=0.03476, over 12025.00 frames. ], tot_loss[loss=0.1757, simple_loss=0.2636, pruned_loss=0.04384, over 2325187.66 frames. ], batch size: 28, lr: 4.48e-03, grad_scale: 8.0 2023-05-16 06:31:38,006 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.7460, 5.7149, 5.4740, 4.9534, 4.9555, 5.6381, 5.2976, 5.0139], device='cuda:0'), covar=tensor([0.0737, 0.0901, 0.0694, 0.1543, 0.0832, 0.0784, 0.1596, 0.1111], device='cuda:0'), in_proj_covar=tensor([0.0572, 0.0508, 0.0484, 0.0590, 0.0389, 0.0671, 0.0724, 0.0536], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0004, 0.0004, 0.0003], device='cuda:0') 2023-05-16 06:32:03,472 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.995e+02 2.898e+02 3.453e+02 4.336e+02 9.975e+02, threshold=6.906e+02, percent-clipped=2.0 2023-05-16 06:32:09,685 INFO [finetune.py:992] (0/2) Epoch 7, batch 850, loss[loss=0.1809, simple_loss=0.2792, pruned_loss=0.04135, over 12192.00 frames. ], tot_loss[loss=0.1768, simple_loss=0.2647, pruned_loss=0.04443, over 2326749.61 frames. ], batch size: 35, lr: 4.48e-03, grad_scale: 8.0 2023-05-16 06:32:21,760 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=174270.0, num_to_drop=1, layers_to_drop={0} 2023-05-16 06:32:28,412 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-05-16 06:32:31,212 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=174283.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 06:32:36,659 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=174291.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 06:32:45,171 INFO [finetune.py:992] (0/2) Epoch 7, batch 900, loss[loss=0.1775, simple_loss=0.2621, pruned_loss=0.04639, over 12185.00 frames. ], tot_loss[loss=0.1759, simple_loss=0.2636, pruned_loss=0.04415, over 2328086.64 frames. ], batch size: 31, lr: 4.48e-03, grad_scale: 8.0 2023-05-16 06:32:51,387 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=2.02 vs. limit=2.0 2023-05-16 06:32:56,723 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=174318.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 06:33:08,919 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.96 vs. limit=2.0 2023-05-16 06:33:14,758 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.956e+02 2.707e+02 3.185e+02 3.751e+02 7.083e+02, threshold=6.370e+02, percent-clipped=1.0 2023-05-16 06:33:14,993 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=174344.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 06:33:20,680 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=174352.0, num_to_drop=1, layers_to_drop={2} 2023-05-16 06:33:21,150 INFO [finetune.py:992] (0/2) Epoch 7, batch 950, loss[loss=0.1754, simple_loss=0.2752, pruned_loss=0.03782, over 12260.00 frames. ], tot_loss[loss=0.1756, simple_loss=0.2632, pruned_loss=0.04397, over 2343831.20 frames. ], batch size: 37, lr: 4.48e-03, grad_scale: 8.0 2023-05-16 06:33:24,337 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.1797, 4.5476, 4.0038, 4.9363, 4.5183, 2.3936, 3.8455, 2.9272], device='cuda:0'), covar=tensor([0.0871, 0.0790, 0.1490, 0.0472, 0.1176, 0.2043, 0.1356, 0.3586], device='cuda:0'), in_proj_covar=tensor([0.0297, 0.0365, 0.0344, 0.0266, 0.0355, 0.0261, 0.0333, 0.0357], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 06:33:34,771 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=174371.0, num_to_drop=1, layers_to_drop={3} 2023-05-16 06:33:36,176 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=174373.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 06:33:47,012 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=174388.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 06:33:57,760 INFO [finetune.py:992] (0/2) Epoch 7, batch 1000, loss[loss=0.1803, simple_loss=0.2715, pruned_loss=0.04456, over 12083.00 frames. ], tot_loss[loss=0.1753, simple_loss=0.2633, pruned_loss=0.04368, over 2352632.94 frames. ], batch size: 32, lr: 4.47e-03, grad_scale: 8.0 2023-05-16 06:34:27,209 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.815e+02 2.792e+02 3.331e+02 4.189e+02 8.922e+02, threshold=6.662e+02, percent-clipped=2.0 2023-05-16 06:34:30,950 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=174449.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 06:34:33,389 INFO [finetune.py:992] (0/2) Epoch 7, batch 1050, loss[loss=0.197, simple_loss=0.2817, pruned_loss=0.05615, over 11732.00 frames. ], tot_loss[loss=0.1758, simple_loss=0.2639, pruned_loss=0.0439, over 2351099.93 frames. ], batch size: 44, lr: 4.47e-03, grad_scale: 8.0 2023-05-16 06:34:47,501 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.95 vs. limit=2.0 2023-05-16 06:35:03,854 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=174495.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 06:35:09,338 INFO [finetune.py:992] (0/2) Epoch 7, batch 1100, loss[loss=0.1654, simple_loss=0.2564, pruned_loss=0.03723, over 12124.00 frames. ], tot_loss[loss=0.176, simple_loss=0.2641, pruned_loss=0.04394, over 2349194.14 frames. ], batch size: 33, lr: 4.47e-03, grad_scale: 8.0 2023-05-16 06:35:29,730 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=4.72 vs. limit=5.0 2023-05-16 06:35:39,029 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.904e+02 2.923e+02 3.347e+02 4.160e+02 1.417e+03, threshold=6.693e+02, percent-clipped=6.0 2023-05-16 06:35:45,182 INFO [finetune.py:992] (0/2) Epoch 7, batch 1150, loss[loss=0.1893, simple_loss=0.2768, pruned_loss=0.05091, over 10404.00 frames. ], tot_loss[loss=0.1759, simple_loss=0.2638, pruned_loss=0.04401, over 2347812.07 frames. ], batch size: 68, lr: 4.47e-03, grad_scale: 8.0 2023-05-16 06:36:12,142 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.6482, 2.9613, 4.5812, 4.8004, 2.7579, 2.6625, 3.0436, 2.0577], device='cuda:0'), covar=tensor([0.1517, 0.2939, 0.0425, 0.0377, 0.1293, 0.2163, 0.2566, 0.4190], device='cuda:0'), in_proj_covar=tensor([0.0276, 0.0369, 0.0259, 0.0286, 0.0253, 0.0285, 0.0358, 0.0358], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-16 06:36:21,265 INFO [finetune.py:992] (0/2) Epoch 7, batch 1200, loss[loss=0.1555, simple_loss=0.2467, pruned_loss=0.0322, over 12365.00 frames. ], tot_loss[loss=0.1756, simple_loss=0.2638, pruned_loss=0.04368, over 2359916.40 frames. ], batch size: 31, lr: 4.47e-03, grad_scale: 8.0 2023-05-16 06:36:47,059 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=174639.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 06:36:50,614 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.028e+02 2.797e+02 3.322e+02 3.820e+02 7.213e+02, threshold=6.643e+02, percent-clipped=1.0 2023-05-16 06:36:52,723 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=174647.0, num_to_drop=1, layers_to_drop={1} 2023-05-16 06:36:56,932 INFO [finetune.py:992] (0/2) Epoch 7, batch 1250, loss[loss=0.2236, simple_loss=0.2886, pruned_loss=0.0793, over 8036.00 frames. ], tot_loss[loss=0.1745, simple_loss=0.2627, pruned_loss=0.04319, over 2362578.88 frames. ], batch size: 98, lr: 4.47e-03, grad_scale: 8.0 2023-05-16 06:37:10,694 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=174671.0, num_to_drop=1, layers_to_drop={2} 2023-05-16 06:37:12,025 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=174673.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 06:37:33,264 INFO [finetune.py:992] (0/2) Epoch 7, batch 1300, loss[loss=0.1518, simple_loss=0.2287, pruned_loss=0.03743, over 11775.00 frames. ], tot_loss[loss=0.1745, simple_loss=0.2627, pruned_loss=0.04317, over 2367695.66 frames. ], batch size: 26, lr: 4.47e-03, grad_scale: 8.0 2023-05-16 06:37:34,925 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=174705.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 06:37:45,131 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=174719.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 06:37:46,519 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=174721.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 06:37:59,152 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.35 vs. limit=2.0 2023-05-16 06:38:02,386 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.3025, 4.7820, 5.2614, 4.5883, 4.8546, 4.6930, 5.2804, 4.9518], device='cuda:0'), covar=tensor([0.0241, 0.0366, 0.0249, 0.0271, 0.0382, 0.0290, 0.0178, 0.0286], device='cuda:0'), in_proj_covar=tensor([0.0241, 0.0245, 0.0269, 0.0243, 0.0240, 0.0238, 0.0219, 0.0203], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 06:38:02,926 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.091e+02 2.824e+02 3.276e+02 4.021e+02 7.300e+02, threshold=6.552e+02, percent-clipped=2.0 2023-05-16 06:38:03,024 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=174744.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 06:38:10,034 INFO [finetune.py:992] (0/2) Epoch 7, batch 1350, loss[loss=0.1627, simple_loss=0.2446, pruned_loss=0.04036, over 12131.00 frames. ], tot_loss[loss=0.1742, simple_loss=0.2626, pruned_loss=0.04295, over 2370380.47 frames. ], batch size: 30, lr: 4.47e-03, grad_scale: 8.0 2023-05-16 06:38:16,849 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.1555, 4.6855, 4.8386, 4.9403, 4.6618, 4.9709, 4.8955, 2.7749], device='cuda:0'), covar=tensor([0.0110, 0.0071, 0.0084, 0.0053, 0.0054, 0.0085, 0.0084, 0.0655], device='cuda:0'), in_proj_covar=tensor([0.0064, 0.0073, 0.0077, 0.0068, 0.0058, 0.0087, 0.0075, 0.0092], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-16 06:38:19,604 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=174766.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 06:38:33,824 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.4983, 2.4864, 3.1794, 4.4705, 2.3458, 4.4016, 4.4188, 4.7020], device='cuda:0'), covar=tensor([0.0142, 0.1236, 0.0472, 0.0109, 0.1273, 0.0248, 0.0215, 0.0065], device='cuda:0'), in_proj_covar=tensor([0.0118, 0.0200, 0.0182, 0.0112, 0.0186, 0.0172, 0.0170, 0.0117], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 06:38:40,153 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=174795.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 06:38:46,020 INFO [finetune.py:992] (0/2) Epoch 7, batch 1400, loss[loss=0.16, simple_loss=0.2454, pruned_loss=0.03724, over 12121.00 frames. ], tot_loss[loss=0.1747, simple_loss=0.2631, pruned_loss=0.04311, over 2372573.03 frames. ], batch size: 30, lr: 4.47e-03, grad_scale: 8.0 2023-05-16 06:39:08,768 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=4.83 vs. limit=5.0 2023-05-16 06:39:15,449 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=174843.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 06:39:16,073 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.826e+02 2.886e+02 3.412e+02 3.861e+02 6.454e+02, threshold=6.823e+02, percent-clipped=0.0 2023-05-16 06:39:22,180 INFO [finetune.py:992] (0/2) Epoch 7, batch 1450, loss[loss=0.1571, simple_loss=0.242, pruned_loss=0.03609, over 12182.00 frames. ], tot_loss[loss=0.1749, simple_loss=0.2634, pruned_loss=0.04318, over 2371677.31 frames. ], batch size: 31, lr: 4.47e-03, grad_scale: 8.0 2023-05-16 06:39:40,235 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.36 vs. limit=2.0 2023-05-16 06:39:57,715 INFO [finetune.py:992] (0/2) Epoch 7, batch 1500, loss[loss=0.1843, simple_loss=0.2762, pruned_loss=0.04617, over 10549.00 frames. ], tot_loss[loss=0.1746, simple_loss=0.2629, pruned_loss=0.04309, over 2380516.63 frames. ], batch size: 68, lr: 4.47e-03, grad_scale: 8.0 2023-05-16 06:40:09,975 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.7234, 2.5848, 4.7630, 5.1122, 3.2457, 2.5832, 2.8852, 1.9828], device='cuda:0'), covar=tensor([0.1569, 0.3489, 0.0409, 0.0319, 0.1057, 0.2427, 0.2985, 0.5114], device='cuda:0'), in_proj_covar=tensor([0.0278, 0.0371, 0.0260, 0.0287, 0.0255, 0.0287, 0.0360, 0.0360], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-16 06:40:23,163 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=174939.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 06:40:26,545 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.978e+02 2.872e+02 3.588e+02 4.496e+02 1.019e+03, threshold=7.175e+02, percent-clipped=4.0 2023-05-16 06:40:28,823 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=174947.0, num_to_drop=1, layers_to_drop={2} 2023-05-16 06:40:30,805 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([6.1656, 6.1095, 5.8893, 5.4279, 5.1094, 6.0394, 5.6691, 5.4731], device='cuda:0'), covar=tensor([0.0632, 0.0955, 0.0610, 0.1566, 0.0683, 0.0721, 0.1587, 0.1062], device='cuda:0'), in_proj_covar=tensor([0.0573, 0.0512, 0.0486, 0.0588, 0.0390, 0.0674, 0.0728, 0.0537], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0004, 0.0004, 0.0003], device='cuda:0') 2023-05-16 06:40:32,862 INFO [finetune.py:992] (0/2) Epoch 7, batch 1550, loss[loss=0.1788, simple_loss=0.2746, pruned_loss=0.04146, over 12361.00 frames. ], tot_loss[loss=0.1742, simple_loss=0.2627, pruned_loss=0.04287, over 2375341.05 frames. ], batch size: 35, lr: 4.47e-03, grad_scale: 8.0 2023-05-16 06:40:57,956 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=174987.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 06:41:03,695 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=174995.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 06:41:09,683 INFO [finetune.py:992] (0/2) Epoch 7, batch 1600, loss[loss=0.1582, simple_loss=0.2444, pruned_loss=0.03602, over 12015.00 frames. ], tot_loss[loss=0.1749, simple_loss=0.2634, pruned_loss=0.04318, over 2367319.43 frames. ], batch size: 28, lr: 4.47e-03, grad_scale: 8.0 2023-05-16 06:41:39,461 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.171e+02 2.821e+02 3.320e+02 3.769e+02 7.996e+02, threshold=6.639e+02, percent-clipped=1.0 2023-05-16 06:41:39,614 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=175044.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 06:41:45,898 INFO [finetune.py:992] (0/2) Epoch 7, batch 1650, loss[loss=0.1931, simple_loss=0.2856, pruned_loss=0.05024, over 12046.00 frames. ], tot_loss[loss=0.1748, simple_loss=0.2632, pruned_loss=0.04324, over 2376465.79 frames. ], batch size: 37, lr: 4.47e-03, grad_scale: 8.0 2023-05-16 06:41:51,857 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=175061.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 06:42:13,610 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=175092.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 06:42:22,027 INFO [finetune.py:992] (0/2) Epoch 7, batch 1700, loss[loss=0.1542, simple_loss=0.2431, pruned_loss=0.03262, over 12290.00 frames. ], tot_loss[loss=0.1747, simple_loss=0.2631, pruned_loss=0.04315, over 2383549.37 frames. ], batch size: 33, lr: 4.47e-03, grad_scale: 8.0 2023-05-16 06:42:33,802 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.0911, 2.4678, 3.7792, 3.1034, 3.6097, 3.3854, 2.4961, 3.6539], device='cuda:0'), covar=tensor([0.0117, 0.0313, 0.0114, 0.0215, 0.0111, 0.0132, 0.0317, 0.0092], device='cuda:0'), in_proj_covar=tensor([0.0166, 0.0196, 0.0176, 0.0171, 0.0200, 0.0151, 0.0185, 0.0170], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 06:42:42,177 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.0146, 4.8160, 4.9519, 5.0084, 4.5974, 4.6579, 4.4514, 4.9545], device='cuda:0'), covar=tensor([0.0697, 0.0591, 0.0812, 0.0579, 0.1804, 0.1283, 0.0571, 0.0944], device='cuda:0'), in_proj_covar=tensor([0.0496, 0.0642, 0.0553, 0.0584, 0.0783, 0.0694, 0.0520, 0.0467], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0005, 0.0005, 0.0004, 0.0003], device='cuda:0') 2023-05-16 06:42:47,921 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=175139.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 06:42:51,241 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.869e+02 2.856e+02 3.271e+02 4.086e+02 1.198e+03, threshold=6.542e+02, percent-clipped=2.0 2023-05-16 06:42:52,246 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.5552, 2.7259, 3.7090, 4.4456, 3.8707, 4.4380, 3.6880, 3.2070], device='cuda:0'), covar=tensor([0.0030, 0.0334, 0.0127, 0.0048, 0.0107, 0.0069, 0.0146, 0.0322], device='cuda:0'), in_proj_covar=tensor([0.0081, 0.0121, 0.0102, 0.0075, 0.0102, 0.0113, 0.0094, 0.0138], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 06:42:57,748 INFO [finetune.py:992] (0/2) Epoch 7, batch 1750, loss[loss=0.1551, simple_loss=0.2339, pruned_loss=0.03811, over 12138.00 frames. ], tot_loss[loss=0.1736, simple_loss=0.262, pruned_loss=0.04262, over 2385152.29 frames. ], batch size: 30, lr: 4.47e-03, grad_scale: 8.0 2023-05-16 06:43:24,898 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.9834, 2.2405, 3.5754, 3.0200, 3.4176, 3.2173, 2.3829, 3.4875], device='cuda:0'), covar=tensor([0.0127, 0.0367, 0.0144, 0.0206, 0.0129, 0.0148, 0.0361, 0.0112], device='cuda:0'), in_proj_covar=tensor([0.0165, 0.0196, 0.0176, 0.0171, 0.0198, 0.0150, 0.0185, 0.0169], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 06:43:32,383 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=175200.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 06:43:34,293 INFO [finetune.py:992] (0/2) Epoch 7, batch 1800, loss[loss=0.1779, simple_loss=0.2666, pruned_loss=0.0446, over 11631.00 frames. ], tot_loss[loss=0.1727, simple_loss=0.261, pruned_loss=0.04224, over 2384574.51 frames. ], batch size: 48, lr: 4.47e-03, grad_scale: 8.0 2023-05-16 06:43:37,346 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=175207.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 06:43:48,101 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.7874, 2.8758, 4.5733, 4.8246, 3.0226, 2.6555, 2.9247, 2.0705], device='cuda:0'), covar=tensor([0.1402, 0.3043, 0.0457, 0.0342, 0.1149, 0.2134, 0.2672, 0.4020], device='cuda:0'), in_proj_covar=tensor([0.0277, 0.0371, 0.0259, 0.0287, 0.0253, 0.0287, 0.0359, 0.0359], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-16 06:44:03,187 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.932e+02 2.769e+02 3.289e+02 3.861e+02 5.923e+02, threshold=6.578e+02, percent-clipped=0.0 2023-05-16 06:44:09,692 INFO [finetune.py:992] (0/2) Epoch 7, batch 1850, loss[loss=0.1576, simple_loss=0.2432, pruned_loss=0.03601, over 12351.00 frames. ], tot_loss[loss=0.1726, simple_loss=0.2611, pruned_loss=0.04204, over 2387554.61 frames. ], batch size: 30, lr: 4.47e-03, grad_scale: 8.0 2023-05-16 06:44:17,235 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=4.68 vs. limit=5.0 2023-05-16 06:44:21,034 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=175268.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 06:44:45,931 INFO [finetune.py:992] (0/2) Epoch 7, batch 1900, loss[loss=0.2032, simple_loss=0.2926, pruned_loss=0.05691, over 11824.00 frames. ], tot_loss[loss=0.1727, simple_loss=0.261, pruned_loss=0.04217, over 2379332.32 frames. ], batch size: 44, lr: 4.46e-03, grad_scale: 8.0 2023-05-16 06:45:15,870 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.077e+02 2.899e+02 3.482e+02 3.970e+02 5.767e+02, threshold=6.964e+02, percent-clipped=0.0 2023-05-16 06:45:22,011 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.14 vs. limit=2.0 2023-05-16 06:45:22,362 INFO [finetune.py:992] (0/2) Epoch 7, batch 1950, loss[loss=0.19, simple_loss=0.283, pruned_loss=0.04846, over 12160.00 frames. ], tot_loss[loss=0.173, simple_loss=0.2616, pruned_loss=0.04222, over 2381710.47 frames. ], batch size: 34, lr: 4.46e-03, grad_scale: 8.0 2023-05-16 06:45:26,939 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-05-16 06:45:28,183 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=175361.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 06:45:34,608 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=175370.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 06:45:51,423 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=175394.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 06:45:58,657 INFO [finetune.py:992] (0/2) Epoch 7, batch 2000, loss[loss=0.1749, simple_loss=0.2695, pruned_loss=0.04017, over 12192.00 frames. ], tot_loss[loss=0.1731, simple_loss=0.2615, pruned_loss=0.04235, over 2374472.48 frames. ], batch size: 35, lr: 4.46e-03, grad_scale: 8.0 2023-05-16 06:46:02,907 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=175409.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 06:46:05,531 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=5.02 vs. limit=5.0 2023-05-16 06:46:10,105 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.7769, 4.0678, 3.6402, 4.3462, 3.9968, 2.5964, 3.6551, 2.9241], device='cuda:0'), covar=tensor([0.0808, 0.0810, 0.1268, 0.0505, 0.1071, 0.1549, 0.1018, 0.2804], device='cuda:0'), in_proj_covar=tensor([0.0296, 0.0368, 0.0346, 0.0267, 0.0356, 0.0262, 0.0332, 0.0358], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 06:46:15,248 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=2.45 vs. limit=5.0 2023-05-16 06:46:18,669 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=175431.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 06:46:27,809 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.855e+02 2.722e+02 3.209e+02 4.046e+02 9.248e+02, threshold=6.418e+02, percent-clipped=0.0 2023-05-16 06:46:34,316 INFO [finetune.py:992] (0/2) Epoch 7, batch 2050, loss[loss=0.1722, simple_loss=0.261, pruned_loss=0.04174, over 12062.00 frames. ], tot_loss[loss=0.1729, simple_loss=0.2614, pruned_loss=0.04225, over 2376456.29 frames. ], batch size: 40, lr: 4.46e-03, grad_scale: 8.0 2023-05-16 06:46:35,957 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=175455.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 06:46:55,095 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=175481.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 06:47:04,918 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=175495.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 06:47:10,606 INFO [finetune.py:992] (0/2) Epoch 7, batch 2100, loss[loss=0.2103, simple_loss=0.2946, pruned_loss=0.06302, over 8583.00 frames. ], tot_loss[loss=0.1729, simple_loss=0.2614, pruned_loss=0.04223, over 2370104.95 frames. ], batch size: 98, lr: 4.46e-03, grad_scale: 8.0 2023-05-16 06:47:33,481 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 2023-05-16 06:47:38,351 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=175542.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 06:47:39,579 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.017e+02 2.816e+02 3.398e+02 4.015e+02 1.115e+03, threshold=6.795e+02, percent-clipped=4.0 2023-05-16 06:47:46,722 INFO [finetune.py:992] (0/2) Epoch 7, batch 2150, loss[loss=0.181, simple_loss=0.2764, pruned_loss=0.04279, over 11758.00 frames. ], tot_loss[loss=0.1727, simple_loss=0.2612, pruned_loss=0.04216, over 2373048.73 frames. ], batch size: 44, lr: 4.46e-03, grad_scale: 8.0 2023-05-16 06:47:53,912 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=175563.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 06:48:19,812 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.45 vs. limit=2.0 2023-05-16 06:48:22,512 INFO [finetune.py:992] (0/2) Epoch 7, batch 2200, loss[loss=0.1742, simple_loss=0.2586, pruned_loss=0.04488, over 12203.00 frames. ], tot_loss[loss=0.1726, simple_loss=0.2606, pruned_loss=0.04232, over 2380043.52 frames. ], batch size: 35, lr: 4.46e-03, grad_scale: 8.0 2023-05-16 06:48:52,283 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.946e+02 2.766e+02 3.230e+02 3.814e+02 1.083e+03, threshold=6.460e+02, percent-clipped=3.0 2023-05-16 06:48:57,539 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.0736, 5.0480, 4.8844, 4.9578, 4.5539, 5.0721, 5.1195, 5.2977], device='cuda:0'), covar=tensor([0.0194, 0.0143, 0.0190, 0.0272, 0.0739, 0.0279, 0.0131, 0.0151], device='cuda:0'), in_proj_covar=tensor([0.0184, 0.0187, 0.0184, 0.0234, 0.0235, 0.0208, 0.0169, 0.0214], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0005, 0.0004, 0.0005, 0.0004, 0.0005], device='cuda:0') 2023-05-16 06:48:58,832 INFO [finetune.py:992] (0/2) Epoch 7, batch 2250, loss[loss=0.1775, simple_loss=0.2694, pruned_loss=0.04281, over 10671.00 frames. ], tot_loss[loss=0.1729, simple_loss=0.2611, pruned_loss=0.04234, over 2376153.43 frames. ], batch size: 68, lr: 4.46e-03, grad_scale: 8.0 2023-05-16 06:49:35,039 INFO [finetune.py:992] (0/2) Epoch 7, batch 2300, loss[loss=0.1618, simple_loss=0.2493, pruned_loss=0.03715, over 12262.00 frames. ], tot_loss[loss=0.1727, simple_loss=0.2609, pruned_loss=0.04231, over 2379524.16 frames. ], batch size: 32, lr: 4.46e-03, grad_scale: 8.0 2023-05-16 06:49:40,120 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=175710.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 06:49:45,422 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.86 vs. limit=2.0 2023-05-16 06:49:51,214 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=175726.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 06:50:04,150 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.887e+02 2.742e+02 3.217e+02 3.891e+02 8.665e+02, threshold=6.435e+02, percent-clipped=4.0 2023-05-16 06:50:09,141 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=175750.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 06:50:11,133 INFO [finetune.py:992] (0/2) Epoch 7, batch 2350, loss[loss=0.1805, simple_loss=0.2741, pruned_loss=0.0435, over 12188.00 frames. ], tot_loss[loss=0.1733, simple_loss=0.2615, pruned_loss=0.04258, over 2378705.63 frames. ], batch size: 35, lr: 4.46e-03, grad_scale: 8.0 2023-05-16 06:50:14,821 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.0017, 4.5666, 4.8080, 4.7578, 4.6231, 4.8217, 4.7180, 2.6192], device='cuda:0'), covar=tensor([0.0103, 0.0068, 0.0081, 0.0062, 0.0050, 0.0090, 0.0080, 0.0748], device='cuda:0'), in_proj_covar=tensor([0.0064, 0.0074, 0.0077, 0.0069, 0.0057, 0.0087, 0.0075, 0.0092], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-16 06:50:20,717 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-05-16 06:50:23,993 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=175771.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 06:50:41,181 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=175795.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 06:50:42,252 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.40 vs. limit=2.0 2023-05-16 06:50:46,827 INFO [finetune.py:992] (0/2) Epoch 7, batch 2400, loss[loss=0.1886, simple_loss=0.2748, pruned_loss=0.05125, over 12120.00 frames. ], tot_loss[loss=0.1728, simple_loss=0.2614, pruned_loss=0.04216, over 2389312.76 frames. ], batch size: 39, lr: 4.46e-03, grad_scale: 8.0 2023-05-16 06:50:56,143 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.9210, 3.7369, 5.3161, 2.4889, 2.8167, 3.9170, 3.4428, 4.0107], device='cuda:0'), covar=tensor([0.0324, 0.0886, 0.0214, 0.1212, 0.1860, 0.1292, 0.1055, 0.0878], device='cuda:0'), in_proj_covar=tensor([0.0233, 0.0231, 0.0236, 0.0181, 0.0236, 0.0282, 0.0224, 0.0252], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:0') 2023-05-16 06:50:56,773 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.5616, 5.4176, 5.4979, 5.5158, 5.1426, 5.2303, 4.9921, 5.4763], device='cuda:0'), covar=tensor([0.0627, 0.0555, 0.0760, 0.0627, 0.1975, 0.1137, 0.0520, 0.1042], device='cuda:0'), in_proj_covar=tensor([0.0503, 0.0652, 0.0558, 0.0594, 0.0792, 0.0698, 0.0524, 0.0470], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0005, 0.0004, 0.0004, 0.0005, 0.0005, 0.0004, 0.0003], device='cuda:0') 2023-05-16 06:51:10,877 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=175837.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 06:51:15,146 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=175843.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 06:51:15,828 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.517e+02 2.658e+02 2.949e+02 3.599e+02 5.995e+02, threshold=5.899e+02, percent-clipped=0.0 2023-05-16 06:51:22,880 INFO [finetune.py:992] (0/2) Epoch 7, batch 2450, loss[loss=0.2018, simple_loss=0.2839, pruned_loss=0.05988, over 8072.00 frames. ], tot_loss[loss=0.1731, simple_loss=0.2615, pruned_loss=0.04234, over 2377686.28 frames. ], batch size: 98, lr: 4.46e-03, grad_scale: 8.0 2023-05-16 06:51:30,115 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=175863.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 06:51:35,968 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.6195, 3.4411, 3.1373, 3.0769, 2.8323, 2.6020, 3.4753, 2.1763], device='cuda:0'), covar=tensor([0.0305, 0.0125, 0.0170, 0.0194, 0.0319, 0.0347, 0.0107, 0.0429], device='cuda:0'), in_proj_covar=tensor([0.0186, 0.0154, 0.0151, 0.0182, 0.0196, 0.0192, 0.0157, 0.0198], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 06:51:43,125 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=175881.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 06:51:44,692 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.15 vs. limit=2.0 2023-05-16 06:51:59,035 INFO [finetune.py:992] (0/2) Epoch 7, batch 2500, loss[loss=0.1717, simple_loss=0.2593, pruned_loss=0.04202, over 12286.00 frames. ], tot_loss[loss=0.1726, simple_loss=0.261, pruned_loss=0.04212, over 2382164.57 frames. ], batch size: 33, lr: 4.46e-03, grad_scale: 8.0 2023-05-16 06:52:04,623 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=175911.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 06:52:16,222 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.4786, 2.4319, 3.1548, 4.4469, 2.2080, 4.4401, 4.4369, 4.6784], device='cuda:0'), covar=tensor([0.0139, 0.1341, 0.0490, 0.0127, 0.1380, 0.0228, 0.0139, 0.0065], device='cuda:0'), in_proj_covar=tensor([0.0120, 0.0204, 0.0183, 0.0114, 0.0188, 0.0176, 0.0172, 0.0118], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 06:52:22,729 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.3576, 3.1610, 3.0124, 2.8900, 2.6531, 2.5788, 3.1733, 1.9916], device='cuda:0'), covar=tensor([0.0337, 0.0121, 0.0165, 0.0184, 0.0366, 0.0296, 0.0112, 0.0452], device='cuda:0'), in_proj_covar=tensor([0.0187, 0.0155, 0.0151, 0.0182, 0.0197, 0.0193, 0.0158, 0.0199], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 06:52:26,822 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=175942.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 06:52:28,042 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.964e+02 2.858e+02 3.344e+02 3.806e+02 7.648e+02, threshold=6.688e+02, percent-clipped=5.0 2023-05-16 06:52:28,926 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=175945.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 06:52:34,498 INFO [finetune.py:992] (0/2) Epoch 7, batch 2550, loss[loss=0.1806, simple_loss=0.2686, pruned_loss=0.04632, over 12304.00 frames. ], tot_loss[loss=0.1724, simple_loss=0.2606, pruned_loss=0.04206, over 2375991.62 frames. ], batch size: 34, lr: 4.46e-03, grad_scale: 8.0 2023-05-16 06:52:36,336 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.94 vs. limit=2.0 2023-05-16 06:53:03,267 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.0804, 2.2343, 3.7584, 3.0916, 3.5686, 3.2360, 2.3773, 3.6633], device='cuda:0'), covar=tensor([0.0173, 0.0431, 0.0128, 0.0265, 0.0159, 0.0163, 0.0442, 0.0159], device='cuda:0'), in_proj_covar=tensor([0.0165, 0.0197, 0.0177, 0.0172, 0.0200, 0.0152, 0.0186, 0.0170], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 06:53:09,169 INFO [checkpoint.py:75] (0/2) Saving checkpoint to pruned_transducer_stateless7/exp_giga_finetune/checkpoint-76000.pt 2023-05-16 06:53:14,409 INFO [finetune.py:992] (0/2) Epoch 7, batch 2600, loss[loss=0.1661, simple_loss=0.2575, pruned_loss=0.03739, over 12359.00 frames. ], tot_loss[loss=0.1717, simple_loss=0.26, pruned_loss=0.0417, over 2368475.83 frames. ], batch size: 36, lr: 4.46e-03, grad_scale: 8.0 2023-05-16 06:53:16,717 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=176006.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 06:53:30,748 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=176026.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 06:53:42,387 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=2.95 vs. limit=5.0 2023-05-16 06:53:44,044 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.056e+02 2.756e+02 3.186e+02 3.876e+02 8.146e+02, threshold=6.371e+02, percent-clipped=1.0 2023-05-16 06:53:48,547 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=176050.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 06:53:50,360 INFO [finetune.py:992] (0/2) Epoch 7, batch 2650, loss[loss=0.1789, simple_loss=0.2674, pruned_loss=0.04524, over 12125.00 frames. ], tot_loss[loss=0.1726, simple_loss=0.2606, pruned_loss=0.04226, over 2366286.66 frames. ], batch size: 38, lr: 4.46e-03, grad_scale: 8.0 2023-05-16 06:53:59,656 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=176066.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 06:54:05,307 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=176074.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 06:54:13,995 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.5487, 4.4293, 4.5401, 4.5496, 4.2286, 4.3211, 4.1088, 4.5197], device='cuda:0'), covar=tensor([0.0784, 0.0634, 0.0752, 0.0664, 0.1841, 0.1193, 0.0610, 0.0972], device='cuda:0'), in_proj_covar=tensor([0.0499, 0.0648, 0.0553, 0.0587, 0.0783, 0.0696, 0.0518, 0.0465], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0005, 0.0004, 0.0004, 0.0005, 0.0005, 0.0004, 0.0003], device='cuda:0') 2023-05-16 06:54:22,508 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=176098.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 06:54:25,946 INFO [finetune.py:992] (0/2) Epoch 7, batch 2700, loss[loss=0.2021, simple_loss=0.2877, pruned_loss=0.05821, over 12340.00 frames. ], tot_loss[loss=0.1719, simple_loss=0.2603, pruned_loss=0.04179, over 2362522.51 frames. ], batch size: 36, lr: 4.46e-03, grad_scale: 8.0 2023-05-16 06:54:50,719 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=176137.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 06:54:55,648 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.944e+02 2.627e+02 3.045e+02 3.681e+02 6.868e+02, threshold=6.091e+02, percent-clipped=1.0 2023-05-16 06:54:55,874 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.1132, 2.0713, 2.8406, 3.0139, 2.9037, 3.0211, 2.8248, 2.4031], device='cuda:0'), covar=tensor([0.0070, 0.0374, 0.0145, 0.0068, 0.0118, 0.0097, 0.0113, 0.0327], device='cuda:0'), in_proj_covar=tensor([0.0082, 0.0121, 0.0102, 0.0075, 0.0103, 0.0114, 0.0095, 0.0139], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 06:55:00,297 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.80 vs. limit=2.0 2023-05-16 06:55:01,878 INFO [finetune.py:992] (0/2) Epoch 7, batch 2750, loss[loss=0.156, simple_loss=0.238, pruned_loss=0.037, over 12351.00 frames. ], tot_loss[loss=0.1724, simple_loss=0.2602, pruned_loss=0.04225, over 2363759.36 frames. ], batch size: 30, lr: 4.46e-03, grad_scale: 16.0 2023-05-16 06:55:25,336 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=176185.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 06:55:27,593 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.1896, 4.6713, 2.7904, 2.5037, 3.9728, 2.5816, 3.9550, 3.0333], device='cuda:0'), covar=tensor([0.0739, 0.0447, 0.1195, 0.1662, 0.0244, 0.1331, 0.0478, 0.0915], device='cuda:0'), in_proj_covar=tensor([0.0184, 0.0244, 0.0173, 0.0195, 0.0134, 0.0179, 0.0192, 0.0171], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 06:55:36,642 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.0688, 2.4394, 3.5875, 3.0024, 3.4587, 3.1400, 2.3445, 3.5334], device='cuda:0'), covar=tensor([0.0127, 0.0334, 0.0120, 0.0196, 0.0115, 0.0173, 0.0347, 0.0089], device='cuda:0'), in_proj_covar=tensor([0.0165, 0.0196, 0.0177, 0.0172, 0.0199, 0.0151, 0.0187, 0.0168], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 06:55:38,583 INFO [finetune.py:992] (0/2) Epoch 7, batch 2800, loss[loss=0.1901, simple_loss=0.2787, pruned_loss=0.05073, over 12110.00 frames. ], tot_loss[loss=0.1724, simple_loss=0.2601, pruned_loss=0.04234, over 2367186.26 frames. ], batch size: 38, lr: 4.46e-03, grad_scale: 8.0 2023-05-16 06:55:40,158 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.2208, 4.4722, 3.9334, 4.8401, 4.3582, 2.8788, 4.2273, 3.0932], device='cuda:0'), covar=tensor([0.0799, 0.0803, 0.1506, 0.0410, 0.1205, 0.1516, 0.0933, 0.3184], device='cuda:0'), in_proj_covar=tensor([0.0297, 0.0371, 0.0348, 0.0270, 0.0356, 0.0263, 0.0334, 0.0357], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 06:55:42,398 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.67 vs. limit=2.0 2023-05-16 06:55:51,365 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.2706, 4.8226, 5.2194, 4.6075, 4.8364, 4.6073, 5.2708, 4.9747], device='cuda:0'), covar=tensor([0.0238, 0.0362, 0.0289, 0.0254, 0.0366, 0.0366, 0.0219, 0.0267], device='cuda:0'), in_proj_covar=tensor([0.0244, 0.0249, 0.0273, 0.0245, 0.0242, 0.0244, 0.0222, 0.0207], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 06:56:00,142 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.0245, 4.7063, 4.8672, 4.8188, 4.6430, 4.8452, 4.8194, 2.6572], device='cuda:0'), covar=tensor([0.0101, 0.0060, 0.0082, 0.0062, 0.0064, 0.0107, 0.0077, 0.0699], device='cuda:0'), in_proj_covar=tensor([0.0063, 0.0073, 0.0077, 0.0068, 0.0057, 0.0087, 0.0074, 0.0091], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-16 06:56:02,877 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=176237.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 06:56:06,351 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=176242.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 06:56:08,231 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.888e+02 2.792e+02 3.403e+02 4.133e+02 6.843e+02, threshold=6.805e+02, percent-clipped=5.0 2023-05-16 06:56:14,076 INFO [finetune.py:992] (0/2) Epoch 7, batch 2850, loss[loss=0.1577, simple_loss=0.245, pruned_loss=0.03521, over 12191.00 frames. ], tot_loss[loss=0.173, simple_loss=0.2606, pruned_loss=0.04268, over 2368226.71 frames. ], batch size: 31, lr: 4.45e-03, grad_scale: 8.0 2023-05-16 06:56:14,240 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.9059, 4.5793, 4.8974, 4.4037, 4.5710, 4.4247, 4.9040, 4.5913], device='cuda:0'), covar=tensor([0.0255, 0.0315, 0.0255, 0.0235, 0.0363, 0.0290, 0.0228, 0.0390], device='cuda:0'), in_proj_covar=tensor([0.0242, 0.0247, 0.0271, 0.0243, 0.0240, 0.0242, 0.0221, 0.0205], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 06:56:15,282 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=4.82 vs. limit=5.0 2023-05-16 06:56:17,186 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=176257.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 06:56:48,889 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=176301.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 06:56:50,161 INFO [finetune.py:992] (0/2) Epoch 7, batch 2900, loss[loss=0.1976, simple_loss=0.2822, pruned_loss=0.05651, over 12119.00 frames. ], tot_loss[loss=0.1732, simple_loss=0.2611, pruned_loss=0.04268, over 2377173.35 frames. ], batch size: 39, lr: 4.45e-03, grad_scale: 8.0 2023-05-16 06:56:50,364 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=176303.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 06:57:01,106 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=176318.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 06:57:03,043 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.9943, 4.8228, 4.9443, 4.9822, 4.5530, 4.6123, 4.4488, 4.8982], device='cuda:0'), covar=tensor([0.0681, 0.0649, 0.0807, 0.0602, 0.2011, 0.1366, 0.0575, 0.1042], device='cuda:0'), in_proj_covar=tensor([0.0503, 0.0654, 0.0558, 0.0593, 0.0796, 0.0702, 0.0524, 0.0468], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0005, 0.0004, 0.0004, 0.0005, 0.0005, 0.0004, 0.0003], device='cuda:0') 2023-05-16 06:57:20,291 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.997e+02 2.685e+02 3.197e+02 3.783e+02 7.095e+02, threshold=6.394e+02, percent-clipped=1.0 2023-05-16 06:57:25,836 INFO [finetune.py:992] (0/2) Epoch 7, batch 2950, loss[loss=0.172, simple_loss=0.2639, pruned_loss=0.04009, over 12112.00 frames. ], tot_loss[loss=0.1736, simple_loss=0.2615, pruned_loss=0.04288, over 2373068.72 frames. ], batch size: 39, lr: 4.45e-03, grad_scale: 8.0 2023-05-16 06:57:35,264 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=176366.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 06:57:54,287 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.1858, 4.8709, 5.0664, 4.9809, 4.8367, 5.0642, 5.0008, 2.6587], device='cuda:0'), covar=tensor([0.0064, 0.0051, 0.0060, 0.0053, 0.0043, 0.0068, 0.0059, 0.0647], device='cuda:0'), in_proj_covar=tensor([0.0063, 0.0073, 0.0076, 0.0068, 0.0057, 0.0086, 0.0074, 0.0090], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-16 06:58:01,730 INFO [finetune.py:992] (0/2) Epoch 7, batch 3000, loss[loss=0.1911, simple_loss=0.2818, pruned_loss=0.05017, over 11867.00 frames. ], tot_loss[loss=0.1731, simple_loss=0.261, pruned_loss=0.04259, over 2370513.28 frames. ], batch size: 44, lr: 4.45e-03, grad_scale: 8.0 2023-05-16 06:58:01,731 INFO [finetune.py:1017] (0/2) Computing validation loss 2023-05-16 06:58:10,637 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.9276, 3.9972, 3.7024, 4.4933, 4.2456, 2.5714, 3.7752, 2.7129], device='cuda:0'), covar=tensor([0.0828, 0.1025, 0.1430, 0.0566, 0.0989, 0.1844, 0.1194, 0.3937], device='cuda:0'), in_proj_covar=tensor([0.0299, 0.0372, 0.0352, 0.0272, 0.0360, 0.0265, 0.0337, 0.0360], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 06:58:21,072 INFO [finetune.py:1026] (0/2) Epoch 7, validation: loss=0.325, simple_loss=0.3994, pruned_loss=0.1253, over 1020973.00 frames. 2023-05-16 06:58:21,073 INFO [finetune.py:1027] (0/2) Maximum memory allocated so far is 12745MB 2023-05-16 06:58:25,442 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([6.2255, 6.2399, 6.0388, 5.4737, 5.2292, 6.0845, 5.7041, 5.5582], device='cuda:0'), covar=tensor([0.0585, 0.0733, 0.0532, 0.1487, 0.0528, 0.0716, 0.1450, 0.1009], device='cuda:0'), in_proj_covar=tensor([0.0586, 0.0518, 0.0497, 0.0596, 0.0395, 0.0683, 0.0739, 0.0545], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0004, 0.0004, 0.0003], device='cuda:0') 2023-05-16 06:58:29,052 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=176414.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 06:58:51,833 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.120e+02 2.842e+02 3.241e+02 3.824e+02 6.982e+02, threshold=6.482e+02, percent-clipped=1.0 2023-05-16 06:58:57,479 INFO [finetune.py:992] (0/2) Epoch 7, batch 3050, loss[loss=0.1652, simple_loss=0.2601, pruned_loss=0.03511, over 11241.00 frames. ], tot_loss[loss=0.1725, simple_loss=0.2606, pruned_loss=0.04218, over 2379905.97 frames. ], batch size: 55, lr: 4.45e-03, grad_scale: 8.0 2023-05-16 06:59:06,804 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=176466.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 06:59:33,007 INFO [finetune.py:992] (0/2) Epoch 7, batch 3100, loss[loss=0.1818, simple_loss=0.2579, pruned_loss=0.05281, over 11992.00 frames. ], tot_loss[loss=0.1725, simple_loss=0.2604, pruned_loss=0.04229, over 2378458.23 frames. ], batch size: 28, lr: 4.45e-03, grad_scale: 8.0 2023-05-16 06:59:50,500 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=176527.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 06:59:57,544 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=176537.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 07:00:03,724 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.503e+02 2.708e+02 3.147e+02 3.794e+02 6.652e+02, threshold=6.294e+02, percent-clipped=2.0 2023-05-16 07:00:09,310 INFO [finetune.py:992] (0/2) Epoch 7, batch 3150, loss[loss=0.1809, simple_loss=0.2673, pruned_loss=0.04725, over 10481.00 frames. ], tot_loss[loss=0.1726, simple_loss=0.2607, pruned_loss=0.04227, over 2376402.55 frames. ], batch size: 68, lr: 4.45e-03, grad_scale: 8.0 2023-05-16 07:00:09,557 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.5019, 4.3277, 4.3945, 4.7628, 3.5702, 4.1163, 2.9328, 4.2469], device='cuda:0'), covar=tensor([0.1578, 0.0607, 0.0782, 0.0521, 0.0808, 0.0632, 0.1534, 0.1692], device='cuda:0'), in_proj_covar=tensor([0.0223, 0.0258, 0.0291, 0.0348, 0.0232, 0.0236, 0.0253, 0.0374], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 07:00:26,725 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.82 vs. limit=2.0 2023-05-16 07:00:32,573 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=176585.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 07:00:41,645 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=176598.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 07:00:43,930 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=176601.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 07:00:45,248 INFO [finetune.py:992] (0/2) Epoch 7, batch 3200, loss[loss=0.1755, simple_loss=0.2544, pruned_loss=0.0483, over 12077.00 frames. ], tot_loss[loss=0.1731, simple_loss=0.2611, pruned_loss=0.04254, over 2376064.76 frames. ], batch size: 32, lr: 4.45e-03, grad_scale: 8.0 2023-05-16 07:00:48,428 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.6349, 2.6637, 4.3520, 4.6168, 2.8100, 2.6378, 2.8739, 2.1469], device='cuda:0'), covar=tensor([0.1531, 0.2875, 0.0510, 0.0392, 0.1279, 0.2199, 0.2693, 0.3971], device='cuda:0'), in_proj_covar=tensor([0.0279, 0.0372, 0.0261, 0.0292, 0.0257, 0.0287, 0.0359, 0.0361], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-16 07:00:52,651 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=176613.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 07:01:01,519 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=176625.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 07:01:15,393 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.065e+02 2.748e+02 3.297e+02 3.969e+02 7.576e+02, threshold=6.593e+02, percent-clipped=1.0 2023-05-16 07:01:18,173 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=176649.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 07:01:19,000 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.2081, 4.5809, 2.7531, 2.5885, 3.8846, 2.4743, 3.8514, 3.0465], device='cuda:0'), covar=tensor([0.0700, 0.0332, 0.1263, 0.1450, 0.0239, 0.1334, 0.0517, 0.0885], device='cuda:0'), in_proj_covar=tensor([0.0186, 0.0246, 0.0175, 0.0196, 0.0135, 0.0180, 0.0195, 0.0173], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 07:01:20,945 INFO [finetune.py:992] (0/2) Epoch 7, batch 3250, loss[loss=0.1964, simple_loss=0.2838, pruned_loss=0.05449, over 12279.00 frames. ], tot_loss[loss=0.1729, simple_loss=0.2609, pruned_loss=0.04244, over 2371211.42 frames. ], batch size: 37, lr: 4.45e-03, grad_scale: 8.0 2023-05-16 07:01:29,735 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=176665.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 07:01:45,012 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=176686.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 07:01:57,486 INFO [finetune.py:992] (0/2) Epoch 7, batch 3300, loss[loss=0.1633, simple_loss=0.2447, pruned_loss=0.04094, over 12030.00 frames. ], tot_loss[loss=0.1727, simple_loss=0.261, pruned_loss=0.04217, over 2375637.09 frames. ], batch size: 28, lr: 4.45e-03, grad_scale: 8.0 2023-05-16 07:01:57,718 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.7624, 2.7176, 3.3312, 4.7198, 2.6404, 4.6335, 4.5308, 4.9480], device='cuda:0'), covar=tensor([0.0120, 0.1229, 0.0451, 0.0092, 0.1206, 0.0203, 0.0192, 0.0055], device='cuda:0'), in_proj_covar=tensor([0.0119, 0.0201, 0.0182, 0.0113, 0.0187, 0.0175, 0.0172, 0.0117], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 07:02:03,346 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.1948, 4.5661, 2.8279, 2.5239, 3.8154, 2.4508, 3.8645, 3.0152], device='cuda:0'), covar=tensor([0.0752, 0.0401, 0.1202, 0.1504, 0.0269, 0.1434, 0.0492, 0.0926], device='cuda:0'), in_proj_covar=tensor([0.0185, 0.0245, 0.0175, 0.0196, 0.0135, 0.0180, 0.0194, 0.0173], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 07:02:13,258 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.4379, 2.3864, 3.1541, 4.3728, 2.2287, 4.2961, 4.2398, 4.6017], device='cuda:0'), covar=tensor([0.0148, 0.1259, 0.0484, 0.0119, 0.1349, 0.0243, 0.0233, 0.0079], device='cuda:0'), in_proj_covar=tensor([0.0120, 0.0201, 0.0183, 0.0113, 0.0188, 0.0175, 0.0172, 0.0117], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 07:02:13,948 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=176726.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 07:02:27,480 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.958e+02 2.802e+02 3.191e+02 3.872e+02 5.594e+02, threshold=6.381e+02, percent-clipped=0.0 2023-05-16 07:02:33,113 INFO [finetune.py:992] (0/2) Epoch 7, batch 3350, loss[loss=0.1777, simple_loss=0.2624, pruned_loss=0.04656, over 10684.00 frames. ], tot_loss[loss=0.173, simple_loss=0.2613, pruned_loss=0.04238, over 2374880.90 frames. ], batch size: 69, lr: 4.45e-03, grad_scale: 8.0 2023-05-16 07:03:08,914 INFO [finetune.py:992] (0/2) Epoch 7, batch 3400, loss[loss=0.1532, simple_loss=0.2403, pruned_loss=0.03304, over 12322.00 frames. ], tot_loss[loss=0.1723, simple_loss=0.2605, pruned_loss=0.0421, over 2379349.97 frames. ], batch size: 31, lr: 4.45e-03, grad_scale: 8.0 2023-05-16 07:03:22,447 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=176822.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 07:03:26,063 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.0708, 4.0236, 4.0268, 4.4534, 2.8144, 3.8949, 2.5998, 4.0178], device='cuda:0'), covar=tensor([0.1692, 0.0641, 0.0908, 0.0507, 0.1137, 0.0636, 0.1740, 0.1099], device='cuda:0'), in_proj_covar=tensor([0.0224, 0.0260, 0.0292, 0.0350, 0.0234, 0.0239, 0.0256, 0.0376], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 07:03:39,292 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.123e+02 2.870e+02 3.341e+02 4.078e+02 8.005e+02, threshold=6.681e+02, percent-clipped=1.0 2023-05-16 07:03:45,536 INFO [finetune.py:992] (0/2) Epoch 7, batch 3450, loss[loss=0.1948, simple_loss=0.2814, pruned_loss=0.05412, over 12104.00 frames. ], tot_loss[loss=0.1716, simple_loss=0.2594, pruned_loss=0.04187, over 2380160.28 frames. ], batch size: 39, lr: 4.45e-03, grad_scale: 8.0 2023-05-16 07:03:45,710 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=176853.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 07:04:17,258 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=176898.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 07:04:20,629 INFO [finetune.py:992] (0/2) Epoch 7, batch 3500, loss[loss=0.1521, simple_loss=0.248, pruned_loss=0.02804, over 10444.00 frames. ], tot_loss[loss=0.1718, simple_loss=0.2601, pruned_loss=0.04173, over 2378885.26 frames. ], batch size: 68, lr: 4.45e-03, grad_scale: 8.0 2023-05-16 07:04:27,818 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=176913.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 07:04:28,600 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=176914.0, num_to_drop=1, layers_to_drop={1} 2023-05-16 07:04:38,271 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.91 vs. limit=2.0 2023-05-16 07:04:39,243 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([6.2095, 6.0979, 5.9238, 5.3788, 5.2117, 6.0868, 5.7872, 5.4303], device='cuda:0'), covar=tensor([0.0521, 0.0850, 0.0660, 0.1653, 0.0662, 0.0698, 0.1292, 0.1057], device='cuda:0'), in_proj_covar=tensor([0.0581, 0.0517, 0.0496, 0.0600, 0.0394, 0.0682, 0.0745, 0.0545], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0004, 0.0004, 0.0003], device='cuda:0') 2023-05-16 07:04:50,074 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.761e+02 2.880e+02 3.425e+02 4.316e+02 1.226e+03, threshold=6.849e+02, percent-clipped=3.0 2023-05-16 07:04:50,920 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=176946.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 07:04:55,836 INFO [finetune.py:992] (0/2) Epoch 7, batch 3550, loss[loss=0.1476, simple_loss=0.2263, pruned_loss=0.03444, over 12262.00 frames. ], tot_loss[loss=0.1725, simple_loss=0.2604, pruned_loss=0.04227, over 2366417.59 frames. ], batch size: 28, lr: 4.45e-03, grad_scale: 8.0 2023-05-16 07:05:01,443 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=176961.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 07:05:16,451 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=176981.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 07:05:33,146 INFO [finetune.py:992] (0/2) Epoch 7, batch 3600, loss[loss=0.1725, simple_loss=0.2626, pruned_loss=0.04125, over 12118.00 frames. ], tot_loss[loss=0.1731, simple_loss=0.2612, pruned_loss=0.04253, over 2353272.41 frames. ], batch size: 39, lr: 4.45e-03, grad_scale: 8.0 2023-05-16 07:05:46,123 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=177021.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 07:06:03,233 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.139e+02 2.813e+02 3.293e+02 3.837e+02 6.918e+02, threshold=6.586e+02, percent-clipped=1.0 2023-05-16 07:06:08,814 INFO [finetune.py:992] (0/2) Epoch 7, batch 3650, loss[loss=0.1899, simple_loss=0.2791, pruned_loss=0.05037, over 12062.00 frames. ], tot_loss[loss=0.1725, simple_loss=0.2608, pruned_loss=0.04208, over 2361516.89 frames. ], batch size: 42, lr: 4.45e-03, grad_scale: 8.0 2023-05-16 07:06:16,357 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=3.99 vs. limit=5.0 2023-05-16 07:06:44,367 INFO [finetune.py:992] (0/2) Epoch 7, batch 3700, loss[loss=0.1709, simple_loss=0.2504, pruned_loss=0.04575, over 12329.00 frames. ], tot_loss[loss=0.1724, simple_loss=0.2606, pruned_loss=0.0421, over 2362211.75 frames. ], batch size: 30, lr: 4.45e-03, grad_scale: 8.0 2023-05-16 07:06:48,766 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=177109.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 07:06:50,219 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=177111.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 07:06:58,698 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=177122.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 07:07:00,947 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.8987, 3.9514, 3.9028, 4.4089, 3.6539, 4.1266, 2.7760, 4.4345], device='cuda:0'), covar=tensor([0.1011, 0.0673, 0.1330, 0.0875, 0.0760, 0.0481, 0.1484, 0.1254], device='cuda:0'), in_proj_covar=tensor([0.0222, 0.0258, 0.0292, 0.0348, 0.0234, 0.0237, 0.0254, 0.0375], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 07:07:15,341 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.903e+02 2.904e+02 3.352e+02 3.873e+02 6.220e+02, threshold=6.704e+02, percent-clipped=0.0 2023-05-16 07:07:21,109 INFO [finetune.py:992] (0/2) Epoch 7, batch 3750, loss[loss=0.1656, simple_loss=0.2466, pruned_loss=0.04235, over 12140.00 frames. ], tot_loss[loss=0.1728, simple_loss=0.2608, pruned_loss=0.04237, over 2362363.73 frames. ], batch size: 30, lr: 4.45e-03, grad_scale: 8.0 2023-05-16 07:07:33,210 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=177170.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 07:07:33,318 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=177170.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 07:07:34,744 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=177172.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 07:07:36,946 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.5055, 4.8418, 3.0261, 2.8018, 4.1610, 2.7465, 4.1272, 3.4101], device='cuda:0'), covar=tensor([0.0646, 0.0567, 0.1150, 0.1462, 0.0266, 0.1227, 0.0469, 0.0764], device='cuda:0'), in_proj_covar=tensor([0.0188, 0.0249, 0.0177, 0.0198, 0.0136, 0.0181, 0.0196, 0.0174], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 07:07:56,594 INFO [finetune.py:992] (0/2) Epoch 7, batch 3800, loss[loss=0.1902, simple_loss=0.2771, pruned_loss=0.05164, over 12122.00 frames. ], tot_loss[loss=0.1732, simple_loss=0.2611, pruned_loss=0.0426, over 2367588.06 frames. ], batch size: 39, lr: 4.44e-03, grad_scale: 8.0 2023-05-16 07:08:00,822 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=177209.0, num_to_drop=1, layers_to_drop={3} 2023-05-16 07:08:26,089 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.914e+02 2.889e+02 3.301e+02 4.159e+02 8.427e+02, threshold=6.603e+02, percent-clipped=1.0 2023-05-16 07:08:31,517 INFO [finetune.py:992] (0/2) Epoch 7, batch 3850, loss[loss=0.1687, simple_loss=0.2629, pruned_loss=0.03724, over 10565.00 frames. ], tot_loss[loss=0.1742, simple_loss=0.2621, pruned_loss=0.04312, over 2362787.65 frames. ], batch size: 68, lr: 4.44e-03, grad_scale: 8.0 2023-05-16 07:08:53,071 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=177281.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 07:09:04,575 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.3568, 4.7346, 2.9692, 2.5936, 4.0979, 2.6765, 3.9880, 3.3148], device='cuda:0'), covar=tensor([0.0718, 0.0462, 0.1116, 0.1599, 0.0243, 0.1242, 0.0440, 0.0743], device='cuda:0'), in_proj_covar=tensor([0.0188, 0.0249, 0.0176, 0.0197, 0.0136, 0.0181, 0.0195, 0.0174], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 07:09:08,639 INFO [finetune.py:992] (0/2) Epoch 7, batch 3900, loss[loss=0.1778, simple_loss=0.2709, pruned_loss=0.04232, over 11982.00 frames. ], tot_loss[loss=0.1732, simple_loss=0.2611, pruned_loss=0.04268, over 2375609.64 frames. ], batch size: 42, lr: 4.44e-03, grad_scale: 8.0 2023-05-16 07:09:21,545 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=177321.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 07:09:27,115 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=177329.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 07:09:28,061 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.7321, 2.8312, 4.6356, 4.8843, 2.9221, 2.7069, 2.8821, 2.1764], device='cuda:0'), covar=tensor([0.1362, 0.2883, 0.0404, 0.0307, 0.1141, 0.1965, 0.2687, 0.3687], device='cuda:0'), in_proj_covar=tensor([0.0278, 0.0372, 0.0260, 0.0290, 0.0256, 0.0286, 0.0359, 0.0360], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-16 07:09:38,529 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.974e+02 2.746e+02 3.274e+02 3.920e+02 7.586e+02, threshold=6.548e+02, percent-clipped=2.0 2023-05-16 07:09:44,179 INFO [finetune.py:992] (0/2) Epoch 7, batch 3950, loss[loss=0.1721, simple_loss=0.2657, pruned_loss=0.03927, over 11831.00 frames. ], tot_loss[loss=0.1734, simple_loss=0.2614, pruned_loss=0.04274, over 2384015.32 frames. ], batch size: 44, lr: 4.44e-03, grad_scale: 8.0 2023-05-16 07:09:55,557 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=177369.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 07:10:20,360 INFO [finetune.py:992] (0/2) Epoch 7, batch 4000, loss[loss=0.1941, simple_loss=0.283, pruned_loss=0.05266, over 12361.00 frames. ], tot_loss[loss=0.1722, simple_loss=0.2601, pruned_loss=0.0422, over 2382286.54 frames. ], batch size: 35, lr: 4.44e-03, grad_scale: 8.0 2023-05-16 07:10:51,244 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.034e+02 2.807e+02 3.267e+02 3.740e+02 6.650e+02, threshold=6.534e+02, percent-clipped=2.0 2023-05-16 07:10:56,917 INFO [finetune.py:992] (0/2) Epoch 7, batch 4050, loss[loss=0.1749, simple_loss=0.2621, pruned_loss=0.04383, over 12023.00 frames. ], tot_loss[loss=0.1725, simple_loss=0.2601, pruned_loss=0.04241, over 2377252.51 frames. ], batch size: 42, lr: 4.44e-03, grad_scale: 8.0 2023-05-16 07:11:05,462 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=177465.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 07:11:06,778 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=177467.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 07:11:32,144 INFO [finetune.py:992] (0/2) Epoch 7, batch 4100, loss[loss=0.1716, simple_loss=0.2721, pruned_loss=0.03552, over 12204.00 frames. ], tot_loss[loss=0.1725, simple_loss=0.26, pruned_loss=0.04249, over 2373789.43 frames. ], batch size: 35, lr: 4.44e-03, grad_scale: 8.0 2023-05-16 07:11:36,280 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=177509.0, num_to_drop=1, layers_to_drop={1} 2023-05-16 07:11:41,930 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.7381, 2.6552, 3.4792, 4.6200, 2.4532, 4.6550, 4.6592, 4.9175], device='cuda:0'), covar=tensor([0.0117, 0.1186, 0.0397, 0.0157, 0.1296, 0.0220, 0.0154, 0.0063], device='cuda:0'), in_proj_covar=tensor([0.0120, 0.0201, 0.0183, 0.0113, 0.0188, 0.0175, 0.0171, 0.0117], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 07:11:42,654 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.3317, 3.4819, 3.2687, 3.1345, 2.8452, 2.6350, 3.6019, 2.2389], device='cuda:0'), covar=tensor([0.0395, 0.0122, 0.0162, 0.0184, 0.0401, 0.0367, 0.0121, 0.0435], device='cuda:0'), in_proj_covar=tensor([0.0190, 0.0158, 0.0153, 0.0184, 0.0202, 0.0195, 0.0161, 0.0201], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 07:12:01,221 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.947e+02 2.896e+02 3.356e+02 4.010e+02 8.247e+02, threshold=6.713e+02, percent-clipped=2.0 2023-05-16 07:12:06,836 INFO [finetune.py:992] (0/2) Epoch 7, batch 4150, loss[loss=0.1524, simple_loss=0.2421, pruned_loss=0.03135, over 12187.00 frames. ], tot_loss[loss=0.1723, simple_loss=0.2597, pruned_loss=0.04238, over 2376183.35 frames. ], batch size: 31, lr: 4.44e-03, grad_scale: 8.0 2023-05-16 07:12:06,969 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=177553.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 07:12:09,571 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=177557.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 07:12:30,429 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.7605, 2.8256, 4.7898, 4.9748, 2.9935, 2.6371, 3.0511, 2.2516], device='cuda:0'), covar=tensor([0.1473, 0.2870, 0.0392, 0.0364, 0.1148, 0.2148, 0.2502, 0.3764], device='cuda:0'), in_proj_covar=tensor([0.0280, 0.0373, 0.0263, 0.0293, 0.0258, 0.0288, 0.0362, 0.0361], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-16 07:12:43,644 INFO [finetune.py:992] (0/2) Epoch 7, batch 4200, loss[loss=0.1829, simple_loss=0.2735, pruned_loss=0.04614, over 12357.00 frames. ], tot_loss[loss=0.172, simple_loss=0.2597, pruned_loss=0.04218, over 2380412.23 frames. ], batch size: 38, lr: 4.44e-03, grad_scale: 8.0 2023-05-16 07:12:51,838 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=177614.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 07:13:13,544 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.083e+02 2.718e+02 3.241e+02 3.751e+02 5.947e+02, threshold=6.481e+02, percent-clipped=0.0 2023-05-16 07:13:19,391 INFO [finetune.py:992] (0/2) Epoch 7, batch 4250, loss[loss=0.1678, simple_loss=0.2624, pruned_loss=0.03657, over 12153.00 frames. ], tot_loss[loss=0.1721, simple_loss=0.2597, pruned_loss=0.04229, over 2379033.54 frames. ], batch size: 34, lr: 4.44e-03, grad_scale: 8.0 2023-05-16 07:13:27,259 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=177664.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 07:13:51,418 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([6.1529, 6.1511, 5.9516, 5.5461, 5.2665, 6.1038, 5.6290, 5.5495], device='cuda:0'), covar=tensor([0.0714, 0.0725, 0.0580, 0.1389, 0.0681, 0.0692, 0.1548, 0.1000], device='cuda:0'), in_proj_covar=tensor([0.0586, 0.0513, 0.0496, 0.0593, 0.0396, 0.0684, 0.0747, 0.0542], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0004, 0.0004, 0.0003], device='cuda:0') 2023-05-16 07:13:55,449 INFO [finetune.py:992] (0/2) Epoch 7, batch 4300, loss[loss=0.1383, simple_loss=0.2246, pruned_loss=0.02595, over 12259.00 frames. ], tot_loss[loss=0.1722, simple_loss=0.2598, pruned_loss=0.04227, over 2370317.32 frames. ], batch size: 28, lr: 4.44e-03, grad_scale: 8.0 2023-05-16 07:14:05,166 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 2023-05-16 07:14:12,023 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=177725.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 07:14:25,975 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.876e+02 2.775e+02 3.290e+02 3.803e+02 7.959e+02, threshold=6.580e+02, percent-clipped=1.0 2023-05-16 07:14:29,782 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.6909, 2.7746, 4.4582, 4.7000, 2.8649, 2.6391, 2.9560, 2.0639], device='cuda:0'), covar=tensor([0.1434, 0.2935, 0.0449, 0.0355, 0.1172, 0.2025, 0.2563, 0.3918], device='cuda:0'), in_proj_covar=tensor([0.0280, 0.0372, 0.0263, 0.0293, 0.0257, 0.0287, 0.0362, 0.0361], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-16 07:14:31,609 INFO [finetune.py:992] (0/2) Epoch 7, batch 4350, loss[loss=0.1433, simple_loss=0.2274, pruned_loss=0.02961, over 12372.00 frames. ], tot_loss[loss=0.173, simple_loss=0.2608, pruned_loss=0.04262, over 2369766.61 frames. ], batch size: 30, lr: 4.44e-03, grad_scale: 8.0 2023-05-16 07:14:39,976 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=177765.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 07:14:41,428 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=177767.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 07:14:50,776 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=177780.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 07:15:02,933 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=177797.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 07:15:07,196 INFO [finetune.py:992] (0/2) Epoch 7, batch 4400, loss[loss=0.132, simple_loss=0.2161, pruned_loss=0.02394, over 12124.00 frames. ], tot_loss[loss=0.1726, simple_loss=0.2604, pruned_loss=0.0424, over 2377526.13 frames. ], batch size: 30, lr: 4.44e-03, grad_scale: 8.0 2023-05-16 07:15:14,373 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=177813.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 07:15:15,721 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=177815.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 07:15:28,785 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.2991, 4.4870, 4.0837, 4.9538, 4.6417, 2.8801, 4.2032, 3.1020], device='cuda:0'), covar=tensor([0.0758, 0.0898, 0.1374, 0.0511, 0.0883, 0.1581, 0.0994, 0.3073], device='cuda:0'), in_proj_covar=tensor([0.0297, 0.0370, 0.0349, 0.0272, 0.0358, 0.0265, 0.0336, 0.0359], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 07:15:34,389 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=177841.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 07:15:37,038 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.989e+02 2.909e+02 3.339e+02 4.089e+02 6.461e+02, threshold=6.677e+02, percent-clipped=0.0 2023-05-16 07:15:43,238 INFO [finetune.py:992] (0/2) Epoch 7, batch 4450, loss[loss=0.1746, simple_loss=0.2722, pruned_loss=0.03846, over 12190.00 frames. ], tot_loss[loss=0.1726, simple_loss=0.2604, pruned_loss=0.04246, over 2380172.58 frames. ], batch size: 35, lr: 4.44e-03, grad_scale: 8.0 2023-05-16 07:15:46,944 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=177858.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 07:16:02,106 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.3450, 4.9108, 5.3235, 4.6354, 4.9145, 4.7361, 5.3695, 4.9868], device='cuda:0'), covar=tensor([0.0249, 0.0337, 0.0258, 0.0267, 0.0356, 0.0342, 0.0215, 0.0293], device='cuda:0'), in_proj_covar=tensor([0.0244, 0.0252, 0.0273, 0.0246, 0.0243, 0.0245, 0.0223, 0.0207], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 07:16:15,462 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.7874, 3.6030, 3.6434, 3.8106, 3.4988, 3.8945, 3.8260, 3.9797], device='cuda:0'), covar=tensor([0.0253, 0.0205, 0.0212, 0.0338, 0.0652, 0.0349, 0.0203, 0.0222], device='cuda:0'), in_proj_covar=tensor([0.0185, 0.0189, 0.0189, 0.0237, 0.0239, 0.0211, 0.0172, 0.0219], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0005, 0.0004, 0.0005, 0.0004, 0.0005], device='cuda:0') 2023-05-16 07:16:19,608 INFO [finetune.py:992] (0/2) Epoch 7, batch 4500, loss[loss=0.175, simple_loss=0.2718, pruned_loss=0.03906, over 12153.00 frames. ], tot_loss[loss=0.1732, simple_loss=0.261, pruned_loss=0.04271, over 2374190.14 frames. ], batch size: 36, lr: 4.44e-03, grad_scale: 8.0 2023-05-16 07:16:24,084 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=177909.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 07:16:49,415 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.110e+02 2.795e+02 3.211e+02 3.845e+02 9.886e+02, threshold=6.421e+02, percent-clipped=1.0 2023-05-16 07:16:55,112 INFO [finetune.py:992] (0/2) Epoch 7, batch 4550, loss[loss=0.1885, simple_loss=0.2779, pruned_loss=0.04954, over 11813.00 frames. ], tot_loss[loss=0.174, simple_loss=0.2621, pruned_loss=0.04297, over 2365240.13 frames. ], batch size: 44, lr: 4.44e-03, grad_scale: 8.0 2023-05-16 07:16:56,001 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.6645, 2.5613, 3.8018, 4.5620, 4.0926, 4.5198, 3.8919, 3.3081], device='cuda:0'), covar=tensor([0.0025, 0.0386, 0.0108, 0.0037, 0.0082, 0.0071, 0.0107, 0.0335], device='cuda:0'), in_proj_covar=tensor([0.0082, 0.0120, 0.0101, 0.0075, 0.0102, 0.0113, 0.0095, 0.0137], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 07:17:29,279 INFO [checkpoint.py:75] (0/2) Saving checkpoint to pruned_transducer_stateless7/exp_giga_finetune/checkpoint-78000.pt 2023-05-16 07:17:34,551 INFO [finetune.py:992] (0/2) Epoch 7, batch 4600, loss[loss=0.1546, simple_loss=0.2347, pruned_loss=0.03722, over 12173.00 frames. ], tot_loss[loss=0.1733, simple_loss=0.261, pruned_loss=0.04277, over 2365707.99 frames. ], batch size: 29, lr: 4.44e-03, grad_scale: 8.0 2023-05-16 07:17:37,207 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.43 vs. limit=2.0 2023-05-16 07:17:47,481 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=178020.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 07:18:04,915 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.013e+02 2.769e+02 3.252e+02 3.820e+02 6.652e+02, threshold=6.503e+02, percent-clipped=1.0 2023-05-16 07:18:10,555 INFO [finetune.py:992] (0/2) Epoch 7, batch 4650, loss[loss=0.1741, simple_loss=0.2619, pruned_loss=0.04321, over 12273.00 frames. ], tot_loss[loss=0.1742, simple_loss=0.2622, pruned_loss=0.04314, over 2362980.37 frames. ], batch size: 33, lr: 4.44e-03, grad_scale: 8.0 2023-05-16 07:18:14,343 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.8176, 2.1630, 3.3268, 3.6808, 3.4781, 3.6875, 3.2866, 2.7198], device='cuda:0'), covar=tensor([0.0042, 0.0414, 0.0136, 0.0059, 0.0120, 0.0087, 0.0134, 0.0362], device='cuda:0'), in_proj_covar=tensor([0.0083, 0.0122, 0.0103, 0.0076, 0.0103, 0.0114, 0.0096, 0.0139], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 07:18:46,180 INFO [finetune.py:992] (0/2) Epoch 7, batch 4700, loss[loss=0.1986, simple_loss=0.2786, pruned_loss=0.05935, over 12097.00 frames. ], tot_loss[loss=0.174, simple_loss=0.262, pruned_loss=0.04295, over 2362006.48 frames. ], batch size: 38, lr: 4.43e-03, grad_scale: 8.0 2023-05-16 07:18:57,208 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.15 vs. limit=2.0 2023-05-16 07:19:09,573 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=178136.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 07:19:16,483 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.916e+02 2.991e+02 3.455e+02 4.205e+02 8.934e+02, threshold=6.910e+02, percent-clipped=1.0 2023-05-16 07:19:22,873 INFO [finetune.py:992] (0/2) Epoch 7, batch 4750, loss[loss=0.1635, simple_loss=0.2408, pruned_loss=0.04309, over 12245.00 frames. ], tot_loss[loss=0.1736, simple_loss=0.2614, pruned_loss=0.04291, over 2369920.65 frames. ], batch size: 32, lr: 4.43e-03, grad_scale: 8.0 2023-05-16 07:19:22,962 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=178153.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 07:19:58,753 INFO [finetune.py:992] (0/2) Epoch 7, batch 4800, loss[loss=0.1747, simple_loss=0.2657, pruned_loss=0.0418, over 12351.00 frames. ], tot_loss[loss=0.1732, simple_loss=0.2612, pruned_loss=0.04261, over 2377138.01 frames. ], batch size: 36, lr: 4.43e-03, grad_scale: 16.0 2023-05-16 07:20:02,971 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=178209.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 07:20:19,472 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.4898, 5.3515, 5.5442, 5.5555, 5.1531, 5.1726, 4.9707, 5.4555], device='cuda:0'), covar=tensor([0.0767, 0.0536, 0.0604, 0.0505, 0.1726, 0.1241, 0.0462, 0.1058], device='cuda:0'), in_proj_covar=tensor([0.0507, 0.0661, 0.0561, 0.0607, 0.0809, 0.0707, 0.0528, 0.0477], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0005, 0.0004, 0.0004, 0.0005, 0.0005, 0.0004, 0.0004], device='cuda:0') 2023-05-16 07:20:28,571 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.990e+02 2.827e+02 3.203e+02 3.957e+02 7.089e+02, threshold=6.407e+02, percent-clipped=1.0 2023-05-16 07:20:34,378 INFO [finetune.py:992] (0/2) Epoch 7, batch 4850, loss[loss=0.1934, simple_loss=0.2759, pruned_loss=0.05541, over 11803.00 frames. ], tot_loss[loss=0.1741, simple_loss=0.2619, pruned_loss=0.04318, over 2368158.91 frames. ], batch size: 44, lr: 4.43e-03, grad_scale: 16.0 2023-05-16 07:20:37,163 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=178257.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 07:20:45,885 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.2411, 5.0492, 5.2056, 5.2280, 4.8263, 4.8536, 4.6505, 5.1744], device='cuda:0'), covar=tensor([0.0730, 0.0676, 0.0803, 0.0628, 0.2167, 0.1498, 0.0578, 0.1065], device='cuda:0'), in_proj_covar=tensor([0.0507, 0.0660, 0.0562, 0.0606, 0.0809, 0.0707, 0.0529, 0.0477], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0005, 0.0004, 0.0004, 0.0005, 0.0005, 0.0004, 0.0004], device='cuda:0') 2023-05-16 07:20:53,054 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.59 vs. limit=2.0 2023-05-16 07:20:59,356 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.3714, 3.7340, 3.4586, 3.3431, 3.2212, 3.0147, 3.8067, 2.1209], device='cuda:0'), covar=tensor([0.0455, 0.0109, 0.0151, 0.0166, 0.0285, 0.0277, 0.0095, 0.0493], device='cuda:0'), in_proj_covar=tensor([0.0188, 0.0158, 0.0155, 0.0181, 0.0201, 0.0194, 0.0162, 0.0200], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 07:21:11,119 INFO [finetune.py:992] (0/2) Epoch 7, batch 4900, loss[loss=0.1714, simple_loss=0.2575, pruned_loss=0.04264, over 12025.00 frames. ], tot_loss[loss=0.1736, simple_loss=0.2614, pruned_loss=0.0429, over 2374701.01 frames. ], batch size: 31, lr: 4.43e-03, grad_scale: 16.0 2023-05-16 07:21:23,119 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=178320.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 07:21:40,632 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.836e+02 2.599e+02 3.080e+02 3.833e+02 7.580e+02, threshold=6.161e+02, percent-clipped=3.0 2023-05-16 07:21:46,464 INFO [finetune.py:992] (0/2) Epoch 7, batch 4950, loss[loss=0.1621, simple_loss=0.2588, pruned_loss=0.03271, over 11811.00 frames. ], tot_loss[loss=0.1732, simple_loss=0.2609, pruned_loss=0.04278, over 2375123.40 frames. ], batch size: 44, lr: 4.43e-03, grad_scale: 16.0 2023-05-16 07:21:57,229 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=178368.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 07:22:16,977 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.4628, 5.2719, 5.4129, 5.4492, 5.0610, 5.0814, 4.8532, 5.3891], device='cuda:0'), covar=tensor([0.0572, 0.0585, 0.0613, 0.0507, 0.1607, 0.1262, 0.0525, 0.0853], device='cuda:0'), in_proj_covar=tensor([0.0501, 0.0656, 0.0559, 0.0603, 0.0798, 0.0703, 0.0525, 0.0472], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0005, 0.0004, 0.0004, 0.0005, 0.0005, 0.0004, 0.0003], device='cuda:0') 2023-05-16 07:22:22,252 INFO [finetune.py:992] (0/2) Epoch 7, batch 5000, loss[loss=0.1953, simple_loss=0.2776, pruned_loss=0.05646, over 11245.00 frames. ], tot_loss[loss=0.1737, simple_loss=0.2614, pruned_loss=0.04304, over 2368257.64 frames. ], batch size: 55, lr: 4.43e-03, grad_scale: 16.0 2023-05-16 07:22:24,234 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.85 vs. limit=2.0 2023-05-16 07:22:38,596 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.1560, 4.7687, 5.1803, 4.4747, 4.7799, 4.5921, 5.2282, 4.8203], device='cuda:0'), covar=tensor([0.0250, 0.0365, 0.0258, 0.0291, 0.0356, 0.0336, 0.0190, 0.0337], device='cuda:0'), in_proj_covar=tensor([0.0244, 0.0251, 0.0272, 0.0247, 0.0244, 0.0245, 0.0221, 0.0206], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 07:22:46,511 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=178436.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 07:22:53,245 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.010e+02 2.791e+02 3.211e+02 3.917e+02 1.429e+03, threshold=6.422e+02, percent-clipped=4.0 2023-05-16 07:22:58,501 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=178452.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 07:22:59,019 INFO [finetune.py:992] (0/2) Epoch 7, batch 5050, loss[loss=0.1577, simple_loss=0.2376, pruned_loss=0.03893, over 12172.00 frames. ], tot_loss[loss=0.173, simple_loss=0.2605, pruned_loss=0.04273, over 2375090.57 frames. ], batch size: 29, lr: 4.43e-03, grad_scale: 16.0 2023-05-16 07:22:59,132 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=178453.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 07:23:21,085 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=178484.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 07:23:24,837 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.7844, 2.8993, 4.7872, 5.0210, 2.8429, 2.7336, 3.1179, 2.3210], device='cuda:0'), covar=tensor([0.1356, 0.2996, 0.0377, 0.0338, 0.1181, 0.2077, 0.2456, 0.3557], device='cuda:0'), in_proj_covar=tensor([0.0277, 0.0369, 0.0261, 0.0290, 0.0255, 0.0285, 0.0356, 0.0355], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-16 07:23:33,122 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=178501.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 07:23:34,473 INFO [finetune.py:992] (0/2) Epoch 7, batch 5100, loss[loss=0.1653, simple_loss=0.2573, pruned_loss=0.03662, over 12346.00 frames. ], tot_loss[loss=0.1727, simple_loss=0.2603, pruned_loss=0.04254, over 2376779.70 frames. ], batch size: 31, lr: 4.43e-03, grad_scale: 16.0 2023-05-16 07:23:41,731 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=178513.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 07:23:52,820 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.35 vs. limit=2.0 2023-05-16 07:23:53,215 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.3246, 4.7107, 2.9688, 2.7926, 4.0443, 2.5644, 3.9676, 3.4215], device='cuda:0'), covar=tensor([0.0698, 0.0473, 0.1032, 0.1385, 0.0281, 0.1319, 0.0498, 0.0727], device='cuda:0'), in_proj_covar=tensor([0.0190, 0.0253, 0.0176, 0.0198, 0.0139, 0.0182, 0.0197, 0.0177], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 07:24:03,023 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.4715, 4.8361, 3.0958, 2.8260, 4.1803, 2.5931, 4.0898, 3.4968], device='cuda:0'), covar=tensor([0.0688, 0.0497, 0.1004, 0.1493, 0.0251, 0.1358, 0.0472, 0.0733], device='cuda:0'), in_proj_covar=tensor([0.0190, 0.0253, 0.0176, 0.0198, 0.0139, 0.0182, 0.0197, 0.0177], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 07:24:04,296 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.912e+02 2.754e+02 3.289e+02 3.828e+02 9.604e+02, threshold=6.577e+02, percent-clipped=1.0 2023-05-16 07:24:09,915 INFO [finetune.py:992] (0/2) Epoch 7, batch 5150, loss[loss=0.1741, simple_loss=0.2763, pruned_loss=0.03596, over 11244.00 frames. ], tot_loss[loss=0.1718, simple_loss=0.2592, pruned_loss=0.04215, over 2374107.02 frames. ], batch size: 55, lr: 4.43e-03, grad_scale: 16.0 2023-05-16 07:24:11,424 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([6.0525, 6.0508, 5.8334, 5.2481, 5.1399, 5.9521, 5.5230, 5.3589], device='cuda:0'), covar=tensor([0.0788, 0.0843, 0.0652, 0.1423, 0.0672, 0.0654, 0.1523, 0.1001], device='cuda:0'), in_proj_covar=tensor([0.0586, 0.0519, 0.0496, 0.0602, 0.0396, 0.0691, 0.0752, 0.0550], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0004, 0.0004, 0.0003], device='cuda:0') 2023-05-16 07:24:33,325 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.2293, 4.8061, 5.2161, 4.5445, 4.8244, 4.6248, 5.2755, 4.9765], device='cuda:0'), covar=tensor([0.0262, 0.0364, 0.0265, 0.0254, 0.0353, 0.0337, 0.0184, 0.0289], device='cuda:0'), in_proj_covar=tensor([0.0244, 0.0251, 0.0272, 0.0246, 0.0244, 0.0244, 0.0222, 0.0205], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 07:24:46,929 INFO [finetune.py:992] (0/2) Epoch 7, batch 5200, loss[loss=0.1786, simple_loss=0.2693, pruned_loss=0.04392, over 12298.00 frames. ], tot_loss[loss=0.1731, simple_loss=0.2603, pruned_loss=0.04295, over 2367859.63 frames. ], batch size: 34, lr: 4.43e-03, grad_scale: 16.0 2023-05-16 07:24:58,502 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=178619.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 07:25:16,557 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.134e+02 2.748e+02 3.403e+02 4.136e+02 9.752e+02, threshold=6.807e+02, percent-clipped=3.0 2023-05-16 07:25:22,343 INFO [finetune.py:992] (0/2) Epoch 7, batch 5250, loss[loss=0.1673, simple_loss=0.2616, pruned_loss=0.03647, over 12268.00 frames. ], tot_loss[loss=0.173, simple_loss=0.2604, pruned_loss=0.04285, over 2377591.54 frames. ], batch size: 37, lr: 4.43e-03, grad_scale: 16.0 2023-05-16 07:25:41,482 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=178680.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 07:25:51,439 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=178694.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 07:25:53,537 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=178697.0, num_to_drop=1, layers_to_drop={0} 2023-05-16 07:25:57,699 INFO [finetune.py:992] (0/2) Epoch 7, batch 5300, loss[loss=0.1479, simple_loss=0.2274, pruned_loss=0.03422, over 12181.00 frames. ], tot_loss[loss=0.1727, simple_loss=0.2602, pruned_loss=0.04263, over 2375201.78 frames. ], batch size: 29, lr: 4.43e-03, grad_scale: 16.0 2023-05-16 07:26:29,082 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.949e+02 2.900e+02 3.366e+02 3.935e+02 8.707e+02, threshold=6.733e+02, percent-clipped=1.0 2023-05-16 07:26:31,512 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.9165, 4.8735, 4.7975, 4.8380, 4.4054, 4.9942, 4.8988, 5.1595], device='cuda:0'), covar=tensor([0.0237, 0.0155, 0.0206, 0.0295, 0.0867, 0.0306, 0.0164, 0.0167], device='cuda:0'), in_proj_covar=tensor([0.0187, 0.0190, 0.0189, 0.0241, 0.0240, 0.0211, 0.0172, 0.0219], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0005, 0.0005, 0.0005, 0.0004, 0.0005], device='cuda:0') 2023-05-16 07:26:34,930 INFO [finetune.py:992] (0/2) Epoch 7, batch 5350, loss[loss=0.1696, simple_loss=0.2681, pruned_loss=0.03555, over 12291.00 frames. ], tot_loss[loss=0.1726, simple_loss=0.2605, pruned_loss=0.04237, over 2367864.55 frames. ], batch size: 37, lr: 4.43e-03, grad_scale: 16.0 2023-05-16 07:26:36,578 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=178755.0, num_to_drop=1, layers_to_drop={2} 2023-05-16 07:26:38,628 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=178758.0, num_to_drop=1, layers_to_drop={1} 2023-05-16 07:26:47,130 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=178770.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 07:26:52,854 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([6.0166, 5.9813, 5.7694, 5.3227, 5.2177, 5.9183, 5.4830, 5.3250], device='cuda:0'), covar=tensor([0.0785, 0.0921, 0.0657, 0.1592, 0.0654, 0.0710, 0.1580, 0.1125], device='cuda:0'), in_proj_covar=tensor([0.0582, 0.0519, 0.0494, 0.0598, 0.0394, 0.0686, 0.0742, 0.0547], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0004, 0.0004, 0.0003], device='cuda:0') 2023-05-16 07:27:10,865 INFO [finetune.py:992] (0/2) Epoch 7, batch 5400, loss[loss=0.1972, simple_loss=0.2944, pruned_loss=0.05, over 10469.00 frames. ], tot_loss[loss=0.1735, simple_loss=0.2614, pruned_loss=0.04277, over 2368903.04 frames. ], batch size: 68, lr: 4.43e-03, grad_scale: 16.0 2023-05-16 07:27:14,520 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=178808.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 07:27:30,721 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=178831.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 07:27:40,558 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.833e+02 2.862e+02 3.225e+02 3.753e+02 6.699e+02, threshold=6.449e+02, percent-clipped=0.0 2023-05-16 07:27:46,223 INFO [finetune.py:992] (0/2) Epoch 7, batch 5450, loss[loss=0.1705, simple_loss=0.2616, pruned_loss=0.03976, over 12034.00 frames. ], tot_loss[loss=0.1727, simple_loss=0.2607, pruned_loss=0.04233, over 2375635.36 frames. ], batch size: 40, lr: 4.43e-03, grad_scale: 16.0 2023-05-16 07:28:15,090 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=4.83 vs. limit=5.0 2023-05-16 07:28:16,953 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=178894.0, num_to_drop=1, layers_to_drop={1} 2023-05-16 07:28:23,143 INFO [finetune.py:992] (0/2) Epoch 7, batch 5500, loss[loss=0.2133, simple_loss=0.2885, pruned_loss=0.06901, over 8290.00 frames. ], tot_loss[loss=0.1728, simple_loss=0.2609, pruned_loss=0.0424, over 2369773.19 frames. ], batch size: 98, lr: 4.43e-03, grad_scale: 16.0 2023-05-16 07:28:25,608 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.1824, 2.6129, 3.7714, 3.1321, 3.6415, 3.2170, 2.6871, 3.6964], device='cuda:0'), covar=tensor([0.0135, 0.0318, 0.0142, 0.0215, 0.0152, 0.0178, 0.0322, 0.0122], device='cuda:0'), in_proj_covar=tensor([0.0165, 0.0198, 0.0179, 0.0173, 0.0202, 0.0154, 0.0189, 0.0170], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 07:28:30,902 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.56 vs. limit=2.0 2023-05-16 07:28:43,453 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.8216, 4.7731, 4.6978, 4.6906, 4.3124, 4.8465, 4.7368, 5.0422], device='cuda:0'), covar=tensor([0.0249, 0.0146, 0.0194, 0.0299, 0.0834, 0.0322, 0.0165, 0.0162], device='cuda:0'), in_proj_covar=tensor([0.0188, 0.0191, 0.0190, 0.0244, 0.0243, 0.0213, 0.0175, 0.0222], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0005, 0.0005, 0.0005, 0.0004, 0.0005], device='cuda:0') 2023-05-16 07:28:49,711 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.3225, 4.8552, 5.3192, 4.6937, 4.9189, 4.7675, 5.3676, 4.9585], device='cuda:0'), covar=tensor([0.0263, 0.0344, 0.0254, 0.0251, 0.0396, 0.0338, 0.0209, 0.0314], device='cuda:0'), in_proj_covar=tensor([0.0245, 0.0253, 0.0273, 0.0247, 0.0246, 0.0246, 0.0222, 0.0206], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 07:28:53,206 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.724e+02 2.973e+02 3.388e+02 4.216e+02 1.873e+03, threshold=6.776e+02, percent-clipped=4.0 2023-05-16 07:28:58,954 INFO [finetune.py:992] (0/2) Epoch 7, batch 5550, loss[loss=0.1453, simple_loss=0.2327, pruned_loss=0.02892, over 12132.00 frames. ], tot_loss[loss=0.172, simple_loss=0.2603, pruned_loss=0.04182, over 2371739.70 frames. ], batch size: 30, lr: 4.43e-03, grad_scale: 16.0 2023-05-16 07:29:00,583 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=178955.0, num_to_drop=1, layers_to_drop={0} 2023-05-16 07:29:14,608 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=178975.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 07:29:34,598 INFO [finetune.py:992] (0/2) Epoch 7, batch 5600, loss[loss=0.1752, simple_loss=0.2609, pruned_loss=0.04476, over 12250.00 frames. ], tot_loss[loss=0.1721, simple_loss=0.2605, pruned_loss=0.04191, over 2373991.88 frames. ], batch size: 32, lr: 4.43e-03, grad_scale: 16.0 2023-05-16 07:29:51,782 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 2023-05-16 07:30:05,389 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.940e+02 2.735e+02 3.260e+02 3.853e+02 6.561e+02, threshold=6.520e+02, percent-clipped=0.0 2023-05-16 07:30:09,025 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=179050.0, num_to_drop=1, layers_to_drop={3} 2023-05-16 07:30:11,001 INFO [finetune.py:992] (0/2) Epoch 7, batch 5650, loss[loss=0.2123, simple_loss=0.3001, pruned_loss=0.06223, over 11123.00 frames. ], tot_loss[loss=0.1722, simple_loss=0.2606, pruned_loss=0.04188, over 2377925.49 frames. ], batch size: 55, lr: 4.42e-03, grad_scale: 16.0 2023-05-16 07:30:11,079 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=179053.0, num_to_drop=1, layers_to_drop={2} 2023-05-16 07:30:41,007 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.5957, 2.5206, 4.5838, 4.8742, 3.2403, 2.5247, 2.8140, 1.9104], device='cuda:0'), covar=tensor([0.1465, 0.3403, 0.0380, 0.0285, 0.0947, 0.2332, 0.2880, 0.4943], device='cuda:0'), in_proj_covar=tensor([0.0277, 0.0372, 0.0263, 0.0292, 0.0255, 0.0287, 0.0358, 0.0357], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-16 07:30:46,557 INFO [finetune.py:992] (0/2) Epoch 7, batch 5700, loss[loss=0.1727, simple_loss=0.2686, pruned_loss=0.03842, over 12341.00 frames. ], tot_loss[loss=0.1724, simple_loss=0.2606, pruned_loss=0.04209, over 2374247.11 frames. ], batch size: 36, lr: 4.42e-03, grad_scale: 16.0 2023-05-16 07:30:50,369 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=179108.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 07:30:52,649 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.6255, 2.8042, 4.3784, 4.5913, 2.9029, 2.5697, 2.8015, 2.1027], device='cuda:0'), covar=tensor([0.1455, 0.2870, 0.0486, 0.0377, 0.1165, 0.2107, 0.2504, 0.3836], device='cuda:0'), in_proj_covar=tensor([0.0277, 0.0374, 0.0263, 0.0292, 0.0256, 0.0287, 0.0359, 0.0358], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-16 07:31:02,879 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=179126.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 07:31:16,222 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.741e+02 2.766e+02 3.420e+02 4.407e+02 1.310e+03, threshold=6.840e+02, percent-clipped=1.0 2023-05-16 07:31:21,951 INFO [finetune.py:992] (0/2) Epoch 7, batch 5750, loss[loss=0.1661, simple_loss=0.2604, pruned_loss=0.03587, over 12028.00 frames. ], tot_loss[loss=0.1728, simple_loss=0.2615, pruned_loss=0.04205, over 2375406.02 frames. ], batch size: 40, lr: 4.42e-03, grad_scale: 16.0 2023-05-16 07:31:24,725 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=179156.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 07:31:59,185 INFO [finetune.py:992] (0/2) Epoch 7, batch 5800, loss[loss=0.209, simple_loss=0.2914, pruned_loss=0.06324, over 11681.00 frames. ], tot_loss[loss=0.1727, simple_loss=0.2611, pruned_loss=0.04208, over 2373836.14 frames. ], batch size: 48, lr: 4.42e-03, grad_scale: 16.0 2023-05-16 07:32:18,404 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.9713, 4.9049, 4.8378, 4.9284, 3.9026, 5.1036, 4.9992, 5.1452], device='cuda:0'), covar=tensor([0.0266, 0.0206, 0.0220, 0.0305, 0.1311, 0.0332, 0.0178, 0.0217], device='cuda:0'), in_proj_covar=tensor([0.0190, 0.0193, 0.0190, 0.0245, 0.0243, 0.0213, 0.0175, 0.0222], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0005, 0.0005, 0.0005, 0.0004, 0.0005], device='cuda:0') 2023-05-16 07:32:28,713 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.897e+02 2.755e+02 3.271e+02 4.027e+02 9.280e+02, threshold=6.542e+02, percent-clipped=3.0 2023-05-16 07:32:32,194 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=179250.0, num_to_drop=1, layers_to_drop={3} 2023-05-16 07:32:34,233 INFO [finetune.py:992] (0/2) Epoch 7, batch 5850, loss[loss=0.164, simple_loss=0.2637, pruned_loss=0.03213, over 11193.00 frames. ], tot_loss[loss=0.1742, simple_loss=0.2624, pruned_loss=0.04302, over 2373832.85 frames. ], batch size: 55, lr: 4.42e-03, grad_scale: 16.0 2023-05-16 07:32:39,042 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.85 vs. limit=2.0 2023-05-16 07:32:43,719 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([1.9708, 2.4108, 2.4054, 2.3328, 2.1719, 1.9887, 2.3654, 1.8349], device='cuda:0'), covar=tensor([0.0286, 0.0146, 0.0170, 0.0183, 0.0308, 0.0236, 0.0164, 0.0331], device='cuda:0'), in_proj_covar=tensor([0.0187, 0.0156, 0.0153, 0.0179, 0.0199, 0.0192, 0.0160, 0.0197], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 07:32:46,045 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.42 vs. limit=2.0 2023-05-16 07:32:49,990 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=179275.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 07:32:54,371 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.2869, 2.5584, 3.7650, 3.1982, 3.6955, 3.2667, 2.6456, 3.7655], device='cuda:0'), covar=tensor([0.0145, 0.0343, 0.0144, 0.0233, 0.0121, 0.0184, 0.0352, 0.0105], device='cuda:0'), in_proj_covar=tensor([0.0165, 0.0198, 0.0181, 0.0175, 0.0203, 0.0154, 0.0189, 0.0171], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 07:33:10,428 INFO [finetune.py:992] (0/2) Epoch 7, batch 5900, loss[loss=0.189, simple_loss=0.2682, pruned_loss=0.05491, over 11762.00 frames. ], tot_loss[loss=0.174, simple_loss=0.2621, pruned_loss=0.043, over 2372087.60 frames. ], batch size: 26, lr: 4.42e-03, grad_scale: 16.0 2023-05-16 07:33:24,031 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.0242, 4.7038, 4.8888, 4.7999, 4.6718, 4.9246, 4.7490, 2.6763], device='cuda:0'), covar=tensor([0.0115, 0.0073, 0.0101, 0.0085, 0.0057, 0.0095, 0.0089, 0.0822], device='cuda:0'), in_proj_covar=tensor([0.0065, 0.0075, 0.0078, 0.0071, 0.0058, 0.0088, 0.0076, 0.0092], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 07:33:24,586 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=179323.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 07:33:34,328 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.0988, 4.9780, 4.9590, 5.0162, 4.5917, 5.1073, 5.0849, 5.3666], device='cuda:0'), covar=tensor([0.0193, 0.0154, 0.0193, 0.0307, 0.0768, 0.0250, 0.0152, 0.0138], device='cuda:0'), in_proj_covar=tensor([0.0189, 0.0193, 0.0191, 0.0245, 0.0243, 0.0214, 0.0175, 0.0222], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0005, 0.0005, 0.0005, 0.0004, 0.0005], device='cuda:0') 2023-05-16 07:33:41,070 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.924e+02 2.863e+02 3.283e+02 4.029e+02 9.331e+02, threshold=6.565e+02, percent-clipped=3.0 2023-05-16 07:33:44,207 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=179350.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 07:33:46,126 INFO [finetune.py:992] (0/2) Epoch 7, batch 5950, loss[loss=0.1933, simple_loss=0.285, pruned_loss=0.05079, over 12269.00 frames. ], tot_loss[loss=0.1758, simple_loss=0.2642, pruned_loss=0.04368, over 2372071.91 frames. ], batch size: 37, lr: 4.42e-03, grad_scale: 8.0 2023-05-16 07:33:46,303 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=179353.0, num_to_drop=1, layers_to_drop={0} 2023-05-16 07:33:51,964 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=179361.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 07:34:07,858 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=2.55 vs. limit=5.0 2023-05-16 07:34:17,949 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=179398.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 07:34:20,259 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=179401.0, num_to_drop=1, layers_to_drop={1} 2023-05-16 07:34:21,575 INFO [finetune.py:992] (0/2) Epoch 7, batch 6000, loss[loss=0.1872, simple_loss=0.2785, pruned_loss=0.04792, over 11684.00 frames. ], tot_loss[loss=0.1764, simple_loss=0.2649, pruned_loss=0.04399, over 2374838.22 frames. ], batch size: 48, lr: 4.42e-03, grad_scale: 8.0 2023-05-16 07:34:21,575 INFO [finetune.py:1017] (0/2) Computing validation loss 2023-05-16 07:34:40,546 INFO [finetune.py:1026] (0/2) Epoch 7, validation: loss=0.3232, simple_loss=0.3987, pruned_loss=0.1238, over 1020973.00 frames. 2023-05-16 07:34:40,547 INFO [finetune.py:1027] (0/2) Maximum memory allocated so far is 12745MB 2023-05-16 07:34:41,043 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=4.58 vs. limit=5.0 2023-05-16 07:34:50,691 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.4890, 5.3037, 5.4480, 5.4557, 5.0739, 5.1029, 4.9245, 5.4441], device='cuda:0'), covar=tensor([0.0598, 0.0569, 0.0602, 0.0538, 0.1755, 0.1306, 0.0508, 0.0853], device='cuda:0'), in_proj_covar=tensor([0.0501, 0.0653, 0.0563, 0.0602, 0.0804, 0.0706, 0.0524, 0.0473], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0005, 0.0004, 0.0004, 0.0005, 0.0005, 0.0004, 0.0003], device='cuda:0') 2023-05-16 07:34:54,306 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=179422.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 07:34:57,029 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=179426.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 07:35:10,511 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=179444.0, num_to_drop=1, layers_to_drop={0} 2023-05-16 07:35:11,690 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.756e+02 2.747e+02 3.245e+02 4.203e+02 6.986e+02, threshold=6.491e+02, percent-clipped=1.0 2023-05-16 07:35:16,759 INFO [finetune.py:992] (0/2) Epoch 7, batch 6050, loss[loss=0.1644, simple_loss=0.2618, pruned_loss=0.03344, over 12306.00 frames. ], tot_loss[loss=0.175, simple_loss=0.2632, pruned_loss=0.04337, over 2382019.95 frames. ], batch size: 34, lr: 4.42e-03, grad_scale: 8.0 2023-05-16 07:35:19,942 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.32 vs. limit=2.0 2023-05-16 07:35:31,929 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=179474.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 07:35:52,394 INFO [finetune.py:992] (0/2) Epoch 7, batch 6100, loss[loss=0.1794, simple_loss=0.268, pruned_loss=0.04543, over 12192.00 frames. ], tot_loss[loss=0.175, simple_loss=0.2631, pruned_loss=0.04346, over 2374514.75 frames. ], batch size: 35, lr: 4.42e-03, grad_scale: 8.0 2023-05-16 07:35:52,515 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.3516, 6.1408, 5.7405, 5.7398, 6.2345, 5.5675, 5.7177, 5.7539], device='cuda:0'), covar=tensor([0.1466, 0.0884, 0.0879, 0.1797, 0.0912, 0.1989, 0.1631, 0.1013], device='cuda:0'), in_proj_covar=tensor([0.0356, 0.0486, 0.0387, 0.0432, 0.0455, 0.0439, 0.0390, 0.0371], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 07:35:52,698 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.0279, 3.6976, 5.3118, 2.8440, 3.0475, 3.8698, 3.6707, 3.9740], device='cuda:0'), covar=tensor([0.0315, 0.0927, 0.0239, 0.1043, 0.1697, 0.1360, 0.1062, 0.1052], device='cuda:0'), in_proj_covar=tensor([0.0229, 0.0230, 0.0234, 0.0177, 0.0231, 0.0279, 0.0220, 0.0251], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:0') 2023-05-16 07:35:54,068 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=179505.0, num_to_drop=1, layers_to_drop={3} 2023-05-16 07:36:23,254 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.099e+02 2.694e+02 3.293e+02 4.008e+02 9.136e+02, threshold=6.586e+02, percent-clipped=3.0 2023-05-16 07:36:26,200 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=179550.0, num_to_drop=1, layers_to_drop={0} 2023-05-16 07:36:28,145 INFO [finetune.py:992] (0/2) Epoch 7, batch 6150, loss[loss=0.1616, simple_loss=0.2568, pruned_loss=0.03314, over 12339.00 frames. ], tot_loss[loss=0.1749, simple_loss=0.2631, pruned_loss=0.04339, over 2379858.79 frames. ], batch size: 36, lr: 4.42e-03, grad_scale: 8.0 2023-05-16 07:36:34,922 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.38 vs. limit=2.0 2023-05-16 07:36:54,261 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.8208, 2.9871, 4.5116, 4.7028, 2.9859, 2.6823, 2.9642, 2.1138], device='cuda:0'), covar=tensor([0.1340, 0.2457, 0.0440, 0.0376, 0.1152, 0.2115, 0.2436, 0.3833], device='cuda:0'), in_proj_covar=tensor([0.0278, 0.0374, 0.0265, 0.0293, 0.0256, 0.0289, 0.0360, 0.0359], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-16 07:37:00,482 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=179598.0, num_to_drop=1, layers_to_drop={1} 2023-05-16 07:37:04,104 INFO [finetune.py:992] (0/2) Epoch 7, batch 6200, loss[loss=0.1667, simple_loss=0.2543, pruned_loss=0.03955, over 12253.00 frames. ], tot_loss[loss=0.1763, simple_loss=0.2642, pruned_loss=0.04423, over 2363548.42 frames. ], batch size: 32, lr: 4.42e-03, grad_scale: 8.0 2023-05-16 07:37:06,434 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([6.2474, 6.2106, 5.9872, 5.4508, 5.2702, 6.0989, 5.8029, 5.5146], device='cuda:0'), covar=tensor([0.0637, 0.0787, 0.0699, 0.1492, 0.0631, 0.0741, 0.1435, 0.1065], device='cuda:0'), in_proj_covar=tensor([0.0593, 0.0527, 0.0502, 0.0609, 0.0402, 0.0692, 0.0753, 0.0554], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0004, 0.0004, 0.0003], device='cuda:0') 2023-05-16 07:37:27,821 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=179636.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 07:37:34,801 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.914e+02 2.901e+02 3.290e+02 4.276e+02 6.618e+02, threshold=6.581e+02, percent-clipped=1.0 2023-05-16 07:37:39,654 INFO [finetune.py:992] (0/2) Epoch 7, batch 6250, loss[loss=0.1535, simple_loss=0.2464, pruned_loss=0.03032, over 12291.00 frames. ], tot_loss[loss=0.1757, simple_loss=0.2639, pruned_loss=0.04373, over 2368173.87 frames. ], batch size: 33, lr: 4.42e-03, grad_scale: 8.0 2023-05-16 07:37:51,220 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.5416, 4.3894, 4.3963, 4.5385, 3.5111, 4.6512, 4.4884, 4.7507], device='cuda:0'), covar=tensor([0.0291, 0.0228, 0.0230, 0.0298, 0.1284, 0.0288, 0.0206, 0.0223], device='cuda:0'), in_proj_covar=tensor([0.0187, 0.0190, 0.0188, 0.0242, 0.0240, 0.0211, 0.0173, 0.0220], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0005, 0.0004, 0.0005, 0.0004, 0.0005], device='cuda:0') 2023-05-16 07:38:00,287 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.3072, 5.0913, 5.2194, 5.2510, 4.8346, 4.8806, 4.6957, 5.1475], device='cuda:0'), covar=tensor([0.0540, 0.0579, 0.0669, 0.0579, 0.2129, 0.1344, 0.0611, 0.1088], device='cuda:0'), in_proj_covar=tensor([0.0502, 0.0654, 0.0557, 0.0600, 0.0805, 0.0706, 0.0522, 0.0472], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0005, 0.0004, 0.0004, 0.0005, 0.0005, 0.0004, 0.0003], device='cuda:0') 2023-05-16 07:38:07,308 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.2804, 5.0936, 5.2456, 5.2562, 4.8467, 4.8802, 4.7017, 5.2096], device='cuda:0'), covar=tensor([0.0611, 0.0601, 0.0631, 0.0626, 0.2217, 0.1406, 0.0623, 0.1007], device='cuda:0'), in_proj_covar=tensor([0.0503, 0.0655, 0.0559, 0.0601, 0.0807, 0.0707, 0.0523, 0.0473], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0005, 0.0004, 0.0004, 0.0005, 0.0005, 0.0004, 0.0003], device='cuda:0') 2023-05-16 07:38:11,783 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=179697.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 07:38:14,693 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.3464, 5.0641, 5.3904, 5.3192, 4.4771, 4.5725, 4.7582, 5.0636], device='cuda:0'), covar=tensor([0.0796, 0.1124, 0.0687, 0.0905, 0.3765, 0.2294, 0.0697, 0.1692], device='cuda:0'), in_proj_covar=tensor([0.0502, 0.0655, 0.0559, 0.0601, 0.0807, 0.0707, 0.0523, 0.0473], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0005, 0.0004, 0.0004, 0.0005, 0.0005, 0.0004, 0.0003], device='cuda:0') 2023-05-16 07:38:15,943 INFO [finetune.py:992] (0/2) Epoch 7, batch 6300, loss[loss=0.151, simple_loss=0.233, pruned_loss=0.03445, over 12206.00 frames. ], tot_loss[loss=0.1753, simple_loss=0.2636, pruned_loss=0.04347, over 2369615.50 frames. ], batch size: 29, lr: 4.42e-03, grad_scale: 8.0 2023-05-16 07:38:25,831 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=179717.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 07:38:47,012 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.024e+02 2.650e+02 3.175e+02 4.099e+02 8.156e+02, threshold=6.350e+02, percent-clipped=2.0 2023-05-16 07:38:52,168 INFO [finetune.py:992] (0/2) Epoch 7, batch 6350, loss[loss=0.1812, simple_loss=0.2688, pruned_loss=0.04678, over 11820.00 frames. ], tot_loss[loss=0.1749, simple_loss=0.2631, pruned_loss=0.04329, over 2370797.65 frames. ], batch size: 44, lr: 4.42e-03, grad_scale: 8.0 2023-05-16 07:39:13,090 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.4483, 3.4620, 3.1313, 3.1863, 2.8262, 2.6449, 3.5268, 2.2545], device='cuda:0'), covar=tensor([0.0326, 0.0122, 0.0147, 0.0168, 0.0309, 0.0309, 0.0111, 0.0414], device='cuda:0'), in_proj_covar=tensor([0.0185, 0.0155, 0.0152, 0.0178, 0.0196, 0.0191, 0.0160, 0.0196], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 07:39:23,196 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=179796.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 07:39:26,190 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=179800.0, num_to_drop=1, layers_to_drop={3} 2023-05-16 07:39:28,301 INFO [finetune.py:992] (0/2) Epoch 7, batch 6400, loss[loss=0.16, simple_loss=0.2527, pruned_loss=0.03361, over 12354.00 frames. ], tot_loss[loss=0.1747, simple_loss=0.2632, pruned_loss=0.04307, over 2372904.01 frames. ], batch size: 35, lr: 4.42e-03, grad_scale: 8.0 2023-05-16 07:39:48,458 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.9072, 4.5447, 4.2527, 4.1130, 4.6619, 4.0982, 4.2734, 4.0500], device='cuda:0'), covar=tensor([0.1591, 0.1133, 0.1313, 0.1969, 0.1069, 0.2023, 0.1660, 0.1473], device='cuda:0'), in_proj_covar=tensor([0.0353, 0.0483, 0.0385, 0.0428, 0.0451, 0.0434, 0.0387, 0.0369], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 07:39:55,545 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.4176, 5.2855, 5.3722, 5.4268, 4.9853, 5.0288, 4.8629, 5.3896], device='cuda:0'), covar=tensor([0.0701, 0.0564, 0.0743, 0.0707, 0.2047, 0.1445, 0.0597, 0.0880], device='cuda:0'), in_proj_covar=tensor([0.0507, 0.0659, 0.0564, 0.0604, 0.0813, 0.0714, 0.0526, 0.0475], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0005, 0.0004, 0.0004, 0.0005, 0.0005, 0.0004, 0.0004], device='cuda:0') 2023-05-16 07:39:59,594 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.822e+02 2.778e+02 3.173e+02 3.946e+02 6.917e+02, threshold=6.346e+02, percent-clipped=1.0 2023-05-16 07:40:03,369 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.9036, 4.8382, 4.7712, 4.8351, 4.4090, 4.8876, 4.8157, 5.1390], device='cuda:0'), covar=tensor([0.0286, 0.0161, 0.0192, 0.0265, 0.0787, 0.0279, 0.0177, 0.0164], device='cuda:0'), in_proj_covar=tensor([0.0190, 0.0192, 0.0190, 0.0245, 0.0241, 0.0214, 0.0174, 0.0222], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0005, 0.0005, 0.0005, 0.0004, 0.0005], device='cuda:0') 2023-05-16 07:40:04,645 INFO [finetune.py:992] (0/2) Epoch 7, batch 6450, loss[loss=0.1728, simple_loss=0.2544, pruned_loss=0.04562, over 12083.00 frames. ], tot_loss[loss=0.1746, simple_loss=0.263, pruned_loss=0.04311, over 2375256.66 frames. ], batch size: 32, lr: 4.42e-03, grad_scale: 8.0 2023-05-16 07:40:07,583 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=179857.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 07:40:15,998 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.2190, 4.8837, 5.0821, 5.0721, 4.9801, 5.0464, 4.9398, 2.8916], device='cuda:0'), covar=tensor([0.0102, 0.0064, 0.0063, 0.0059, 0.0040, 0.0084, 0.0075, 0.0593], device='cuda:0'), in_proj_covar=tensor([0.0065, 0.0075, 0.0078, 0.0071, 0.0058, 0.0089, 0.0077, 0.0093], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 07:40:25,968 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.94 vs. limit=2.0 2023-05-16 07:40:40,225 INFO [finetune.py:992] (0/2) Epoch 7, batch 6500, loss[loss=0.2347, simple_loss=0.3266, pruned_loss=0.07147, over 10566.00 frames. ], tot_loss[loss=0.1748, simple_loss=0.2632, pruned_loss=0.04323, over 2376369.27 frames. ], batch size: 68, lr: 4.42e-03, grad_scale: 8.0 2023-05-16 07:40:41,205 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.6313, 2.6156, 3.8907, 4.1099, 2.8979, 2.5308, 2.6738, 2.0803], device='cuda:0'), covar=tensor([0.1343, 0.2659, 0.0567, 0.0445, 0.1121, 0.2193, 0.2350, 0.3722], device='cuda:0'), in_proj_covar=tensor([0.0278, 0.0374, 0.0264, 0.0292, 0.0257, 0.0289, 0.0359, 0.0360], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-16 07:40:42,638 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=179906.0, num_to_drop=1, layers_to_drop={1} 2023-05-16 07:40:55,167 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=179924.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 07:40:58,015 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.3203, 3.3866, 3.0939, 3.0384, 2.8160, 2.5302, 3.4050, 2.1825], device='cuda:0'), covar=tensor([0.0370, 0.0134, 0.0178, 0.0180, 0.0324, 0.0399, 0.0152, 0.0447], device='cuda:0'), in_proj_covar=tensor([0.0187, 0.0157, 0.0154, 0.0181, 0.0198, 0.0193, 0.0162, 0.0198], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 07:41:10,717 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.948e+02 2.731e+02 3.331e+02 4.021e+02 7.219e+02, threshold=6.663e+02, percent-clipped=2.0 2023-05-16 07:41:15,764 INFO [finetune.py:992] (0/2) Epoch 7, batch 6550, loss[loss=0.1459, simple_loss=0.2327, pruned_loss=0.02952, over 12095.00 frames. ], tot_loss[loss=0.1747, simple_loss=0.2631, pruned_loss=0.0432, over 2373861.19 frames. ], batch size: 32, lr: 4.41e-03, grad_scale: 8.0 2023-05-16 07:41:25,915 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=179967.0, num_to_drop=1, layers_to_drop={2} 2023-05-16 07:41:38,880 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=179985.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 07:41:43,848 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=179992.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 07:41:49,787 INFO [checkpoint.py:75] (0/2) Saving checkpoint to pruned_transducer_stateless7/exp_giga_finetune/checkpoint-80000.pt 2023-05-16 07:41:54,840 INFO [finetune.py:992] (0/2) Epoch 7, batch 6600, loss[loss=0.18, simple_loss=0.2735, pruned_loss=0.04331, over 12306.00 frames. ], tot_loss[loss=0.1742, simple_loss=0.2626, pruned_loss=0.04286, over 2378756.21 frames. ], batch size: 34, lr: 4.41e-03, grad_scale: 8.0 2023-05-16 07:42:05,780 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=180017.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 07:42:12,176 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=180026.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 07:42:26,113 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.026e+02 2.656e+02 3.106e+02 3.718e+02 6.929e+02, threshold=6.212e+02, percent-clipped=1.0 2023-05-16 07:42:31,187 INFO [finetune.py:992] (0/2) Epoch 7, batch 6650, loss[loss=0.171, simple_loss=0.2646, pruned_loss=0.03871, over 12127.00 frames. ], tot_loss[loss=0.1737, simple_loss=0.2618, pruned_loss=0.04275, over 2381296.97 frames. ], batch size: 38, lr: 4.41e-03, grad_scale: 8.0 2023-05-16 07:42:39,684 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=180065.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 07:42:55,219 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=180087.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 07:43:04,152 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=180100.0, num_to_drop=1, layers_to_drop={1} 2023-05-16 07:43:06,143 INFO [finetune.py:992] (0/2) Epoch 7, batch 6700, loss[loss=0.1856, simple_loss=0.2732, pruned_loss=0.04902, over 12055.00 frames. ], tot_loss[loss=0.1739, simple_loss=0.2617, pruned_loss=0.04306, over 2372629.55 frames. ], batch size: 37, lr: 4.41e-03, grad_scale: 8.0 2023-05-16 07:43:37,582 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.027e+02 2.747e+02 3.239e+02 4.138e+02 6.736e+02, threshold=6.478e+02, percent-clipped=3.0 2023-05-16 07:43:39,127 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=180148.0, num_to_drop=1, layers_to_drop={0} 2023-05-16 07:43:41,667 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 2023-05-16 07:43:41,900 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=180152.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 07:43:42,557 INFO [finetune.py:992] (0/2) Epoch 7, batch 6750, loss[loss=0.1925, simple_loss=0.2824, pruned_loss=0.05133, over 11686.00 frames. ], tot_loss[loss=0.1747, simple_loss=0.2626, pruned_loss=0.04337, over 2365635.77 frames. ], batch size: 48, lr: 4.41e-03, grad_scale: 8.0 2023-05-16 07:44:10,590 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.4536, 2.5060, 3.6157, 4.3286, 3.8717, 4.3506, 3.7151, 3.0036], device='cuda:0'), covar=tensor([0.0034, 0.0356, 0.0147, 0.0040, 0.0130, 0.0078, 0.0136, 0.0349], device='cuda:0'), in_proj_covar=tensor([0.0082, 0.0120, 0.0102, 0.0074, 0.0100, 0.0113, 0.0093, 0.0135], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 07:44:10,630 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.0562, 2.4181, 3.5636, 3.0583, 3.4545, 3.1972, 2.5432, 3.5158], device='cuda:0'), covar=tensor([0.0136, 0.0325, 0.0140, 0.0211, 0.0137, 0.0144, 0.0309, 0.0104], device='cuda:0'), in_proj_covar=tensor([0.0167, 0.0199, 0.0183, 0.0176, 0.0205, 0.0156, 0.0191, 0.0173], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 07:44:18,441 INFO [finetune.py:992] (0/2) Epoch 7, batch 6800, loss[loss=0.1772, simple_loss=0.2733, pruned_loss=0.04056, over 11610.00 frames. ], tot_loss[loss=0.1753, simple_loss=0.2633, pruned_loss=0.04366, over 2365826.83 frames. ], batch size: 48, lr: 4.41e-03, grad_scale: 8.0 2023-05-16 07:44:20,042 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=180205.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 07:44:48,597 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.911e+02 2.845e+02 3.434e+02 4.080e+02 7.808e+02, threshold=6.868e+02, percent-clipped=1.0 2023-05-16 07:44:53,611 INFO [finetune.py:992] (0/2) Epoch 7, batch 6850, loss[loss=0.1952, simple_loss=0.2824, pruned_loss=0.05404, over 12369.00 frames. ], tot_loss[loss=0.175, simple_loss=0.2629, pruned_loss=0.04354, over 2369532.76 frames. ], batch size: 38, lr: 4.41e-03, grad_scale: 8.0 2023-05-16 07:45:00,562 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=180262.0, num_to_drop=1, layers_to_drop={3} 2023-05-16 07:45:03,551 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=180266.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 07:45:13,522 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=180280.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 07:45:14,361 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.4428, 4.7343, 2.8642, 2.7620, 4.0779, 2.6325, 3.9180, 3.2054], device='cuda:0'), covar=tensor([0.0584, 0.0455, 0.1122, 0.1426, 0.0246, 0.1267, 0.0481, 0.0769], device='cuda:0'), in_proj_covar=tensor([0.0186, 0.0248, 0.0173, 0.0195, 0.0136, 0.0177, 0.0193, 0.0171], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 07:45:22,011 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=180292.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 07:45:30,417 INFO [finetune.py:992] (0/2) Epoch 7, batch 6900, loss[loss=0.1715, simple_loss=0.2613, pruned_loss=0.04082, over 12350.00 frames. ], tot_loss[loss=0.1744, simple_loss=0.2622, pruned_loss=0.04333, over 2369234.91 frames. ], batch size: 36, lr: 4.41e-03, grad_scale: 8.0 2023-05-16 07:45:53,892 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=180336.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 07:45:56,663 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=180340.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 07:46:00,860 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.997e+02 2.614e+02 3.096e+02 3.766e+02 7.680e+02, threshold=6.192e+02, percent-clipped=1.0 2023-05-16 07:46:05,804 INFO [finetune.py:992] (0/2) Epoch 7, batch 6950, loss[loss=0.1843, simple_loss=0.2805, pruned_loss=0.04405, over 12064.00 frames. ], tot_loss[loss=0.174, simple_loss=0.2623, pruned_loss=0.04288, over 2371601.11 frames. ], batch size: 42, lr: 4.41e-03, grad_scale: 8.0 2023-05-16 07:46:14,235 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.5723, 2.8338, 4.4358, 4.7507, 2.9692, 2.5998, 2.7843, 2.0767], device='cuda:0'), covar=tensor([0.1418, 0.2821, 0.0458, 0.0335, 0.1093, 0.2033, 0.2773, 0.3842], device='cuda:0'), in_proj_covar=tensor([0.0280, 0.0374, 0.0265, 0.0293, 0.0256, 0.0291, 0.0361, 0.0361], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-16 07:46:19,310 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.4605, 4.8402, 4.3215, 5.0090, 4.7418, 2.9045, 4.3621, 3.1853], device='cuda:0'), covar=tensor([0.0725, 0.0654, 0.1365, 0.0464, 0.0893, 0.1518, 0.0953, 0.3120], device='cuda:0'), in_proj_covar=tensor([0.0297, 0.0372, 0.0349, 0.0276, 0.0359, 0.0264, 0.0337, 0.0360], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 07:46:26,753 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=180382.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 07:46:37,490 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=180397.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 07:46:41,839 INFO [finetune.py:992] (0/2) Epoch 7, batch 7000, loss[loss=0.1621, simple_loss=0.2439, pruned_loss=0.04015, over 11809.00 frames. ], tot_loss[loss=0.1728, simple_loss=0.261, pruned_loss=0.04234, over 2378467.07 frames. ], batch size: 26, lr: 4.41e-03, grad_scale: 8.0 2023-05-16 07:46:55,420 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.0123, 5.7877, 5.3756, 5.3181, 5.8877, 5.1918, 5.4394, 5.4195], device='cuda:0'), covar=tensor([0.1527, 0.0953, 0.1172, 0.2047, 0.0952, 0.2045, 0.1629, 0.1168], device='cuda:0'), in_proj_covar=tensor([0.0351, 0.0478, 0.0381, 0.0424, 0.0448, 0.0427, 0.0386, 0.0366], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 07:47:12,680 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.8756, 3.3190, 2.4069, 2.2368, 2.9537, 2.2200, 3.1740, 2.5652], device='cuda:0'), covar=tensor([0.0541, 0.0606, 0.0901, 0.1278, 0.0268, 0.1156, 0.0469, 0.0766], device='cuda:0'), in_proj_covar=tensor([0.0186, 0.0247, 0.0172, 0.0193, 0.0136, 0.0176, 0.0192, 0.0170], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 07:47:13,163 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.800e+02 2.795e+02 3.201e+02 3.809e+02 8.289e+02, threshold=6.402e+02, percent-clipped=2.0 2023-05-16 07:47:18,235 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=180452.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 07:47:18,865 INFO [finetune.py:992] (0/2) Epoch 7, batch 7050, loss[loss=0.1394, simple_loss=0.228, pruned_loss=0.02537, over 12404.00 frames. ], tot_loss[loss=0.1733, simple_loss=0.2613, pruned_loss=0.04266, over 2368670.76 frames. ], batch size: 32, lr: 4.41e-03, grad_scale: 8.0 2023-05-16 07:47:27,665 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.90 vs. limit=2.0 2023-05-16 07:47:51,880 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=180500.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 07:47:53,468 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.3147, 2.8835, 3.5939, 4.2273, 3.8558, 4.2446, 3.6957, 2.8685], device='cuda:0'), covar=tensor([0.0038, 0.0301, 0.0123, 0.0035, 0.0105, 0.0061, 0.0109, 0.0373], device='cuda:0'), in_proj_covar=tensor([0.0083, 0.0121, 0.0102, 0.0074, 0.0102, 0.0114, 0.0094, 0.0137], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 07:47:53,956 INFO [finetune.py:992] (0/2) Epoch 7, batch 7100, loss[loss=0.1975, simple_loss=0.2799, pruned_loss=0.0575, over 11860.00 frames. ], tot_loss[loss=0.1746, simple_loss=0.2624, pruned_loss=0.0434, over 2361467.80 frames. ], batch size: 44, lr: 4.41e-03, grad_scale: 8.0 2023-05-16 07:48:06,292 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.1649, 5.9520, 5.6291, 5.4298, 6.0462, 5.2971, 5.6300, 5.6214], device='cuda:0'), covar=tensor([0.1375, 0.0888, 0.0953, 0.1987, 0.0870, 0.2084, 0.1399, 0.1022], device='cuda:0'), in_proj_covar=tensor([0.0347, 0.0472, 0.0377, 0.0422, 0.0445, 0.0423, 0.0383, 0.0364], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 07:48:25,050 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.035e+02 2.914e+02 3.366e+02 4.178e+02 7.262e+02, threshold=6.731e+02, percent-clipped=1.0 2023-05-16 07:48:30,033 INFO [finetune.py:992] (0/2) Epoch 7, batch 7150, loss[loss=0.1957, simple_loss=0.2955, pruned_loss=0.04793, over 11618.00 frames. ], tot_loss[loss=0.1741, simple_loss=0.2624, pruned_loss=0.04295, over 2371689.77 frames. ], batch size: 48, lr: 4.41e-03, grad_scale: 8.0 2023-05-16 07:48:34,378 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.3693, 6.1442, 5.7802, 5.6385, 6.2095, 5.5910, 5.7966, 5.7413], device='cuda:0'), covar=tensor([0.1391, 0.0919, 0.0967, 0.1861, 0.0818, 0.1798, 0.1450, 0.0910], device='cuda:0'), in_proj_covar=tensor([0.0348, 0.0473, 0.0378, 0.0422, 0.0446, 0.0423, 0.0383, 0.0364], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 07:48:35,773 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=180561.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 07:48:36,556 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=180562.0, num_to_drop=1, layers_to_drop={0} 2023-05-16 07:48:42,339 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=180570.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 07:48:49,446 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=180580.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 07:48:51,795 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.6942, 2.8498, 4.4398, 4.5599, 3.0195, 2.7536, 2.8563, 2.0294], device='cuda:0'), covar=tensor([0.1335, 0.2606, 0.0449, 0.0385, 0.1105, 0.1905, 0.2555, 0.3783], device='cuda:0'), in_proj_covar=tensor([0.0279, 0.0374, 0.0264, 0.0293, 0.0256, 0.0289, 0.0359, 0.0359], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-16 07:48:56,635 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.3586, 4.9186, 5.3427, 4.6923, 4.9833, 4.7032, 5.3693, 5.0545], device='cuda:0'), covar=tensor([0.0215, 0.0322, 0.0263, 0.0242, 0.0315, 0.0326, 0.0216, 0.0263], device='cuda:0'), in_proj_covar=tensor([0.0243, 0.0251, 0.0273, 0.0247, 0.0243, 0.0246, 0.0222, 0.0207], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 07:49:06,841 INFO [finetune.py:992] (0/2) Epoch 7, batch 7200, loss[loss=0.1575, simple_loss=0.2435, pruned_loss=0.03576, over 12035.00 frames. ], tot_loss[loss=0.1734, simple_loss=0.2615, pruned_loss=0.04267, over 2377488.16 frames. ], batch size: 31, lr: 4.41e-03, grad_scale: 8.0 2023-05-16 07:49:11,722 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=180610.0, num_to_drop=1, layers_to_drop={0} 2023-05-16 07:49:24,557 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=180628.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 07:49:26,765 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=180631.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 07:49:37,258 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.070e+02 2.799e+02 3.250e+02 3.741e+02 6.052e+02, threshold=6.500e+02, percent-clipped=0.0 2023-05-16 07:49:42,173 INFO [finetune.py:992] (0/2) Epoch 7, batch 7250, loss[loss=0.1947, simple_loss=0.2762, pruned_loss=0.0566, over 11231.00 frames. ], tot_loss[loss=0.173, simple_loss=0.2609, pruned_loss=0.0426, over 2380801.34 frames. ], batch size: 55, lr: 4.41e-03, grad_scale: 8.0 2023-05-16 07:49:43,868 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.1585, 2.5214, 3.6710, 3.1831, 3.5864, 3.3378, 2.5609, 3.5557], device='cuda:0'), covar=tensor([0.0112, 0.0324, 0.0139, 0.0184, 0.0132, 0.0152, 0.0341, 0.0124], device='cuda:0'), in_proj_covar=tensor([0.0168, 0.0201, 0.0184, 0.0177, 0.0206, 0.0156, 0.0192, 0.0175], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 07:49:53,319 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 2023-05-16 07:50:03,002 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=180682.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 07:50:09,957 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=180692.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 07:50:18,532 INFO [finetune.py:992] (0/2) Epoch 7, batch 7300, loss[loss=0.1811, simple_loss=0.2805, pruned_loss=0.04086, over 12045.00 frames. ], tot_loss[loss=0.1736, simple_loss=0.2612, pruned_loss=0.04299, over 2382016.38 frames. ], batch size: 42, lr: 4.41e-03, grad_scale: 8.0 2023-05-16 07:50:37,617 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=180730.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 07:50:39,209 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=180732.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 07:50:48,108 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.33 vs. limit=2.0 2023-05-16 07:50:49,595 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.048e+02 2.837e+02 3.266e+02 3.945e+02 6.850e+02, threshold=6.533e+02, percent-clipped=1.0 2023-05-16 07:50:54,523 INFO [finetune.py:992] (0/2) Epoch 7, batch 7350, loss[loss=0.1524, simple_loss=0.2461, pruned_loss=0.02939, over 12112.00 frames. ], tot_loss[loss=0.1737, simple_loss=0.2614, pruned_loss=0.04307, over 2375742.74 frames. ], batch size: 33, lr: 4.41e-03, grad_scale: 8.0 2023-05-16 07:51:23,299 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=180793.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 07:51:30,454 INFO [finetune.py:992] (0/2) Epoch 7, batch 7400, loss[loss=0.1789, simple_loss=0.268, pruned_loss=0.04487, over 12349.00 frames. ], tot_loss[loss=0.1743, simple_loss=0.2617, pruned_loss=0.04339, over 2373613.76 frames. ], batch size: 35, lr: 4.41e-03, grad_scale: 8.0 2023-05-16 07:51:46,311 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=180825.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 07:51:51,248 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([1.9870, 4.1601, 4.1946, 4.4754, 3.1489, 4.0998, 2.5303, 4.2226], device='cuda:0'), covar=tensor([0.1694, 0.0652, 0.0881, 0.0586, 0.1084, 0.0587, 0.1988, 0.1275], device='cuda:0'), in_proj_covar=tensor([0.0225, 0.0265, 0.0297, 0.0354, 0.0239, 0.0240, 0.0260, 0.0381], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 07:52:01,406 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.980e+02 2.835e+02 3.344e+02 3.894e+02 5.964e+02, threshold=6.688e+02, percent-clipped=0.0 2023-05-16 07:52:06,460 INFO [finetune.py:992] (0/2) Epoch 7, batch 7450, loss[loss=0.1995, simple_loss=0.298, pruned_loss=0.05049, over 12193.00 frames. ], tot_loss[loss=0.1739, simple_loss=0.2617, pruned_loss=0.04304, over 2376692.79 frames. ], batch size: 35, lr: 4.41e-03, grad_scale: 8.0 2023-05-16 07:52:12,374 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=180861.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 07:52:20,940 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-05-16 07:52:29,806 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=180886.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 07:52:33,577 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-05-16 07:52:42,384 INFO [finetune.py:992] (0/2) Epoch 7, batch 7500, loss[loss=0.1792, simple_loss=0.2712, pruned_loss=0.04364, over 12003.00 frames. ], tot_loss[loss=0.1734, simple_loss=0.2611, pruned_loss=0.04283, over 2379206.34 frames. ], batch size: 40, lr: 4.40e-03, grad_scale: 8.0 2023-05-16 07:52:46,793 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=180909.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 07:52:54,232 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([1.9527, 3.9436, 4.1136, 4.4722, 2.7905, 3.9991, 2.4517, 3.9621], device='cuda:0'), covar=tensor([0.1819, 0.0758, 0.0866, 0.0560, 0.1344, 0.0609, 0.1983, 0.1206], device='cuda:0'), in_proj_covar=tensor([0.0225, 0.0265, 0.0296, 0.0353, 0.0238, 0.0240, 0.0259, 0.0380], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 07:52:57,424 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=5.02 vs. limit=5.0 2023-05-16 07:52:59,086 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=180926.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 07:53:13,091 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.952e+02 2.813e+02 3.239e+02 3.692e+02 7.434e+02, threshold=6.478e+02, percent-clipped=1.0 2023-05-16 07:53:18,160 INFO [finetune.py:992] (0/2) Epoch 7, batch 7550, loss[loss=0.1774, simple_loss=0.2741, pruned_loss=0.04033, over 11569.00 frames. ], tot_loss[loss=0.172, simple_loss=0.2599, pruned_loss=0.04207, over 2385015.93 frames. ], batch size: 48, lr: 4.40e-03, grad_scale: 8.0 2023-05-16 07:53:46,144 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=180992.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 07:53:54,162 INFO [finetune.py:992] (0/2) Epoch 7, batch 7600, loss[loss=0.165, simple_loss=0.2649, pruned_loss=0.03257, over 12326.00 frames. ], tot_loss[loss=0.1724, simple_loss=0.2601, pruned_loss=0.04239, over 2379469.00 frames. ], batch size: 34, lr: 4.40e-03, grad_scale: 8.0 2023-05-16 07:54:21,087 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=181040.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 07:54:25,181 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.639e+02 2.789e+02 3.341e+02 4.058e+02 7.799e+02, threshold=6.682e+02, percent-clipped=3.0 2023-05-16 07:54:26,959 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=2.06 vs. limit=2.0 2023-05-16 07:54:30,011 INFO [finetune.py:992] (0/2) Epoch 7, batch 7650, loss[loss=0.189, simple_loss=0.2867, pruned_loss=0.04567, over 12126.00 frames. ], tot_loss[loss=0.1721, simple_loss=0.26, pruned_loss=0.04216, over 2384116.24 frames. ], batch size: 38, lr: 4.40e-03, grad_scale: 8.0 2023-05-16 07:54:54,940 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=181088.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 07:55:05,451 INFO [finetune.py:992] (0/2) Epoch 7, batch 7700, loss[loss=0.1873, simple_loss=0.272, pruned_loss=0.05131, over 12291.00 frames. ], tot_loss[loss=0.172, simple_loss=0.2602, pruned_loss=0.04194, over 2383303.00 frames. ], batch size: 33, lr: 4.40e-03, grad_scale: 8.0 2023-05-16 07:55:36,614 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.931e+02 3.065e+02 3.525e+02 4.565e+02 2.580e+03, threshold=7.049e+02, percent-clipped=8.0 2023-05-16 07:55:41,682 INFO [finetune.py:992] (0/2) Epoch 7, batch 7750, loss[loss=0.1662, simple_loss=0.2583, pruned_loss=0.03702, over 12110.00 frames. ], tot_loss[loss=0.1734, simple_loss=0.2618, pruned_loss=0.04244, over 2385535.53 frames. ], batch size: 33, lr: 4.40e-03, grad_scale: 8.0 2023-05-16 07:56:02,071 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=181181.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 07:56:17,693 INFO [finetune.py:992] (0/2) Epoch 7, batch 7800, loss[loss=0.1832, simple_loss=0.2802, pruned_loss=0.04307, over 12052.00 frames. ], tot_loss[loss=0.1742, simple_loss=0.2623, pruned_loss=0.04305, over 2376526.60 frames. ], batch size: 40, lr: 4.40e-03, grad_scale: 8.0 2023-05-16 07:56:33,912 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=181226.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 07:56:47,949 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.083e+02 2.871e+02 3.508e+02 4.302e+02 1.050e+03, threshold=7.015e+02, percent-clipped=3.0 2023-05-16 07:56:52,969 INFO [finetune.py:992] (0/2) Epoch 7, batch 7850, loss[loss=0.1621, simple_loss=0.2559, pruned_loss=0.03418, over 12305.00 frames. ], tot_loss[loss=0.175, simple_loss=0.263, pruned_loss=0.0435, over 2369392.81 frames. ], batch size: 34, lr: 4.40e-03, grad_scale: 8.0 2023-05-16 07:57:08,351 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=181274.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 07:57:24,543 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=181297.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 07:57:26,962 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.38 vs. limit=2.0 2023-05-16 07:57:28,637 INFO [finetune.py:992] (0/2) Epoch 7, batch 7900, loss[loss=0.182, simple_loss=0.2728, pruned_loss=0.04555, over 12063.00 frames. ], tot_loss[loss=0.1755, simple_loss=0.2633, pruned_loss=0.04386, over 2363016.38 frames. ], batch size: 42, lr: 4.40e-03, grad_scale: 8.0 2023-05-16 07:57:33,070 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.6114, 2.5313, 3.2682, 4.5333, 2.5991, 4.5789, 4.6542, 4.7739], device='cuda:0'), covar=tensor([0.0131, 0.1215, 0.0420, 0.0140, 0.1216, 0.0187, 0.0117, 0.0101], device='cuda:0'), in_proj_covar=tensor([0.0117, 0.0198, 0.0180, 0.0113, 0.0185, 0.0175, 0.0170, 0.0118], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 07:57:58,489 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.5650, 2.3025, 3.3010, 4.5741, 2.2245, 4.6464, 4.7433, 4.8062], device='cuda:0'), covar=tensor([0.0162, 0.1286, 0.0401, 0.0153, 0.1347, 0.0163, 0.0105, 0.0111], device='cuda:0'), in_proj_covar=tensor([0.0117, 0.0198, 0.0179, 0.0113, 0.0184, 0.0174, 0.0170, 0.0118], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 07:57:59,794 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.935e+02 2.838e+02 3.458e+02 4.140e+02 1.351e+03, threshold=6.917e+02, percent-clipped=2.0 2023-05-16 07:58:04,674 INFO [finetune.py:992] (0/2) Epoch 7, batch 7950, loss[loss=0.2685, simple_loss=0.3289, pruned_loss=0.1041, over 8092.00 frames. ], tot_loss[loss=0.1763, simple_loss=0.2637, pruned_loss=0.04443, over 2355696.70 frames. ], batch size: 98, lr: 4.40e-03, grad_scale: 16.0 2023-05-16 07:58:08,392 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=181358.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 07:58:17,636 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.7208, 4.3782, 4.5778, 4.6191, 4.4316, 4.6323, 4.5599, 2.5288], device='cuda:0'), covar=tensor([0.0092, 0.0066, 0.0088, 0.0065, 0.0056, 0.0085, 0.0095, 0.0757], device='cuda:0'), in_proj_covar=tensor([0.0064, 0.0074, 0.0077, 0.0070, 0.0057, 0.0087, 0.0076, 0.0092], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 07:58:29,499 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=181388.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 07:58:40,218 INFO [finetune.py:992] (0/2) Epoch 7, batch 8000, loss[loss=0.1692, simple_loss=0.2543, pruned_loss=0.04205, over 12198.00 frames. ], tot_loss[loss=0.1772, simple_loss=0.2643, pruned_loss=0.045, over 2353871.37 frames. ], batch size: 35, lr: 4.40e-03, grad_scale: 16.0 2023-05-16 07:58:43,311 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.7943, 2.9455, 4.4855, 4.7839, 3.1595, 2.6870, 2.8805, 1.9915], device='cuda:0'), covar=tensor([0.1246, 0.2609, 0.0442, 0.0349, 0.1017, 0.2026, 0.2455, 0.3817], device='cuda:0'), in_proj_covar=tensor([0.0279, 0.0374, 0.0266, 0.0292, 0.0257, 0.0289, 0.0362, 0.0362], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-16 07:58:55,482 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.9639, 5.7721, 5.4133, 5.3490, 5.8672, 5.1888, 5.3672, 5.4593], device='cuda:0'), covar=tensor([0.1447, 0.0923, 0.1028, 0.1962, 0.0894, 0.2072, 0.1715, 0.1071], device='cuda:0'), in_proj_covar=tensor([0.0351, 0.0475, 0.0381, 0.0421, 0.0452, 0.0424, 0.0386, 0.0368], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 07:59:04,697 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=181436.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 07:59:11,694 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.051e+02 2.793e+02 3.335e+02 3.930e+02 8.631e+02, threshold=6.670e+02, percent-clipped=2.0 2023-05-16 07:59:16,534 INFO [finetune.py:992] (0/2) Epoch 7, batch 8050, loss[loss=0.166, simple_loss=0.2523, pruned_loss=0.03984, over 12291.00 frames. ], tot_loss[loss=0.1778, simple_loss=0.2647, pruned_loss=0.04544, over 2351769.67 frames. ], batch size: 33, lr: 4.40e-03, grad_scale: 16.0 2023-05-16 07:59:37,381 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=181481.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 07:59:52,913 INFO [finetune.py:992] (0/2) Epoch 7, batch 8100, loss[loss=0.1566, simple_loss=0.2492, pruned_loss=0.03198, over 12286.00 frames. ], tot_loss[loss=0.1775, simple_loss=0.2646, pruned_loss=0.04521, over 2354508.49 frames. ], batch size: 34, lr: 4.40e-03, grad_scale: 16.0 2023-05-16 08:00:11,415 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=181529.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 08:00:12,295 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=181530.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 08:00:15,815 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.3779, 3.4469, 3.1586, 3.1877, 2.7553, 2.5280, 3.5244, 2.2690], device='cuda:0'), covar=tensor([0.0368, 0.0137, 0.0189, 0.0166, 0.0418, 0.0441, 0.0134, 0.0443], device='cuda:0'), in_proj_covar=tensor([0.0188, 0.0158, 0.0153, 0.0180, 0.0200, 0.0193, 0.0162, 0.0199], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 08:00:23,305 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.160e+02 2.962e+02 3.368e+02 3.994e+02 1.143e+03, threshold=6.735e+02, percent-clipped=5.0 2023-05-16 08:00:28,349 INFO [finetune.py:992] (0/2) Epoch 7, batch 8150, loss[loss=0.253, simple_loss=0.31, pruned_loss=0.09804, over 8050.00 frames. ], tot_loss[loss=0.1772, simple_loss=0.2642, pruned_loss=0.04511, over 2351581.77 frames. ], batch size: 98, lr: 4.40e-03, grad_scale: 16.0 2023-05-16 08:00:55,859 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=181591.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 08:01:03,998 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.3129, 3.3962, 3.1681, 3.0996, 2.6470, 2.5705, 3.4703, 2.1430], device='cuda:0'), covar=tensor([0.0356, 0.0146, 0.0172, 0.0185, 0.0394, 0.0363, 0.0130, 0.0453], device='cuda:0'), in_proj_covar=tensor([0.0188, 0.0158, 0.0152, 0.0180, 0.0200, 0.0192, 0.0162, 0.0198], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 08:01:04,476 INFO [finetune.py:992] (0/2) Epoch 7, batch 8200, loss[loss=0.2081, simple_loss=0.2938, pruned_loss=0.06122, over 11998.00 frames. ], tot_loss[loss=0.177, simple_loss=0.2639, pruned_loss=0.04505, over 2355548.55 frames. ], batch size: 40, lr: 4.40e-03, grad_scale: 16.0 2023-05-16 08:01:12,871 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.0581, 4.7885, 4.9221, 4.9569, 4.8783, 4.9564, 4.7840, 2.8184], device='cuda:0'), covar=tensor([0.0100, 0.0063, 0.0082, 0.0052, 0.0035, 0.0089, 0.0091, 0.0643], device='cuda:0'), in_proj_covar=tensor([0.0064, 0.0075, 0.0078, 0.0070, 0.0058, 0.0088, 0.0077, 0.0093], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 08:01:15,922 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.73 vs. limit=2.0 2023-05-16 08:01:34,928 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.877e+02 3.045e+02 3.538e+02 4.343e+02 6.804e+02, threshold=7.077e+02, percent-clipped=1.0 2023-05-16 08:01:39,851 INFO [finetune.py:992] (0/2) Epoch 7, batch 8250, loss[loss=0.1538, simple_loss=0.2284, pruned_loss=0.03958, over 12300.00 frames. ], tot_loss[loss=0.177, simple_loss=0.2643, pruned_loss=0.04483, over 2356091.37 frames. ], batch size: 28, lr: 4.40e-03, grad_scale: 16.0 2023-05-16 08:01:39,928 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=181653.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 08:02:02,690 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.6088, 2.5878, 3.1820, 4.4759, 2.5110, 4.4691, 4.6332, 4.7524], device='cuda:0'), covar=tensor([0.0124, 0.1197, 0.0491, 0.0181, 0.1233, 0.0227, 0.0161, 0.0098], device='cuda:0'), in_proj_covar=tensor([0.0118, 0.0203, 0.0184, 0.0115, 0.0187, 0.0178, 0.0174, 0.0119], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 08:02:15,670 INFO [finetune.py:992] (0/2) Epoch 7, batch 8300, loss[loss=0.1989, simple_loss=0.2897, pruned_loss=0.05406, over 12279.00 frames. ], tot_loss[loss=0.1772, simple_loss=0.2648, pruned_loss=0.04481, over 2357401.08 frames. ], batch size: 33, lr: 4.40e-03, grad_scale: 16.0 2023-05-16 08:02:15,870 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.2114, 4.9017, 5.0975, 5.1132, 4.9050, 5.1120, 4.9466, 2.9380], device='cuda:0'), covar=tensor([0.0089, 0.0064, 0.0070, 0.0056, 0.0056, 0.0083, 0.0093, 0.0635], device='cuda:0'), in_proj_covar=tensor([0.0065, 0.0075, 0.0078, 0.0071, 0.0058, 0.0088, 0.0077, 0.0093], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 08:02:46,649 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.1113, 3.4339, 3.5889, 3.9665, 2.8110, 3.5674, 2.3247, 3.5601], device='cuda:0'), covar=tensor([0.1687, 0.0914, 0.0917, 0.0674, 0.1227, 0.0713, 0.2056, 0.1148], device='cuda:0'), in_proj_covar=tensor([0.0224, 0.0263, 0.0295, 0.0352, 0.0237, 0.0238, 0.0257, 0.0376], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 08:02:47,020 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.876e+02 2.829e+02 3.361e+02 4.143e+02 6.488e+02, threshold=6.723e+02, percent-clipped=0.0 2023-05-16 08:02:51,865 INFO [finetune.py:992] (0/2) Epoch 7, batch 8350, loss[loss=0.1506, simple_loss=0.2445, pruned_loss=0.02839, over 12353.00 frames. ], tot_loss[loss=0.1768, simple_loss=0.2645, pruned_loss=0.04449, over 2361020.29 frames. ], batch size: 31, lr: 4.40e-03, grad_scale: 16.0 2023-05-16 08:02:59,905 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.5701, 5.3841, 5.5392, 5.5775, 5.1712, 5.2299, 4.9999, 5.5142], device='cuda:0'), covar=tensor([0.0576, 0.0527, 0.0584, 0.0493, 0.1699, 0.1217, 0.0456, 0.0850], device='cuda:0'), in_proj_covar=tensor([0.0507, 0.0669, 0.0565, 0.0607, 0.0818, 0.0718, 0.0534, 0.0480], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0005, 0.0004, 0.0004, 0.0005, 0.0005, 0.0004, 0.0004], device='cuda:0') 2023-05-16 08:03:27,738 INFO [finetune.py:992] (0/2) Epoch 7, batch 8400, loss[loss=0.194, simple_loss=0.2839, pruned_loss=0.05199, over 12068.00 frames. ], tot_loss[loss=0.1751, simple_loss=0.2632, pruned_loss=0.04349, over 2368964.38 frames. ], batch size: 42, lr: 4.39e-03, grad_scale: 16.0 2023-05-16 08:03:42,222 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=3.99 vs. limit=5.0 2023-05-16 08:03:42,911 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 2023-05-16 08:03:59,180 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.056e+02 2.791e+02 3.242e+02 3.752e+02 6.527e+02, threshold=6.484e+02, percent-clipped=0.0 2023-05-16 08:04:04,011 INFO [finetune.py:992] (0/2) Epoch 7, batch 8450, loss[loss=0.235, simple_loss=0.3085, pruned_loss=0.0807, over 10474.00 frames. ], tot_loss[loss=0.1757, simple_loss=0.2635, pruned_loss=0.04395, over 2363136.54 frames. ], batch size: 68, lr: 4.39e-03, grad_scale: 16.0 2023-05-16 08:04:10,698 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.2482, 3.5635, 3.8463, 4.0982, 2.7386, 3.6074, 2.3161, 3.8139], device='cuda:0'), covar=tensor([0.1578, 0.1000, 0.1251, 0.0980, 0.1372, 0.0807, 0.2160, 0.1200], device='cuda:0'), in_proj_covar=tensor([0.0224, 0.0263, 0.0296, 0.0353, 0.0238, 0.0238, 0.0258, 0.0377], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 08:04:27,330 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=181886.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 08:04:34,523 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.5603, 2.8577, 4.4186, 4.6225, 2.8090, 2.5625, 2.9947, 2.0752], device='cuda:0'), covar=tensor([0.1530, 0.2783, 0.0464, 0.0391, 0.1254, 0.2223, 0.2527, 0.3919], device='cuda:0'), in_proj_covar=tensor([0.0278, 0.0373, 0.0265, 0.0290, 0.0256, 0.0287, 0.0361, 0.0360], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-16 08:04:40,043 INFO [finetune.py:992] (0/2) Epoch 7, batch 8500, loss[loss=0.152, simple_loss=0.2346, pruned_loss=0.03473, over 12017.00 frames. ], tot_loss[loss=0.1747, simple_loss=0.2624, pruned_loss=0.04349, over 2364981.67 frames. ], batch size: 31, lr: 4.39e-03, grad_scale: 16.0 2023-05-16 08:05:10,407 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.595e+02 2.826e+02 3.359e+02 3.857e+02 7.437e+02, threshold=6.718e+02, percent-clipped=3.0 2023-05-16 08:05:15,305 INFO [finetune.py:992] (0/2) Epoch 7, batch 8550, loss[loss=0.1682, simple_loss=0.2505, pruned_loss=0.04292, over 12091.00 frames. ], tot_loss[loss=0.1742, simple_loss=0.2618, pruned_loss=0.04329, over 2376411.74 frames. ], batch size: 32, lr: 4.39e-03, grad_scale: 16.0 2023-05-16 08:05:15,419 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=181953.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 08:05:33,164 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=181978.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 08:05:49,393 INFO [checkpoint.py:75] (0/2) Saving checkpoint to pruned_transducer_stateless7/exp_giga_finetune/checkpoint-82000.pt 2023-05-16 08:05:53,050 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=182001.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 08:05:54,371 INFO [finetune.py:992] (0/2) Epoch 7, batch 8600, loss[loss=0.1539, simple_loss=0.2322, pruned_loss=0.03778, over 11987.00 frames. ], tot_loss[loss=0.1744, simple_loss=0.2624, pruned_loss=0.04326, over 2375780.64 frames. ], batch size: 28, lr: 4.39e-03, grad_scale: 16.0 2023-05-16 08:05:59,708 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.3381, 4.5873, 4.1552, 5.0375, 4.6968, 3.0838, 4.2065, 3.1166], device='cuda:0'), covar=tensor([0.0815, 0.0879, 0.1282, 0.0394, 0.0994, 0.1488, 0.1015, 0.3408], device='cuda:0'), in_proj_covar=tensor([0.0301, 0.0373, 0.0354, 0.0279, 0.0361, 0.0265, 0.0341, 0.0363], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 08:06:21,080 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=182039.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 08:06:25,790 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.866e+02 2.615e+02 3.209e+02 4.105e+02 7.691e+02, threshold=6.418e+02, percent-clipped=1.0 2023-05-16 08:06:30,805 INFO [finetune.py:992] (0/2) Epoch 7, batch 8650, loss[loss=0.1471, simple_loss=0.2298, pruned_loss=0.03222, over 12287.00 frames. ], tot_loss[loss=0.1743, simple_loss=0.2621, pruned_loss=0.04325, over 2373399.59 frames. ], batch size: 33, lr: 4.39e-03, grad_scale: 16.0 2023-05-16 08:07:06,557 INFO [finetune.py:992] (0/2) Epoch 7, batch 8700, loss[loss=0.1755, simple_loss=0.2696, pruned_loss=0.04071, over 11790.00 frames. ], tot_loss[loss=0.1733, simple_loss=0.2611, pruned_loss=0.04271, over 2374824.32 frames. ], batch size: 44, lr: 4.39e-03, grad_scale: 16.0 2023-05-16 08:07:17,304 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=3.43 vs. limit=5.0 2023-05-16 08:07:31,907 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.4375, 4.7090, 2.8437, 2.8599, 4.0729, 2.4493, 3.9544, 3.2590], device='cuda:0'), covar=tensor([0.0643, 0.0428, 0.1053, 0.1256, 0.0242, 0.1390, 0.0423, 0.0720], device='cuda:0'), in_proj_covar=tensor([0.0187, 0.0251, 0.0175, 0.0197, 0.0139, 0.0180, 0.0195, 0.0172], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 08:07:37,401 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.020e+02 2.795e+02 3.270e+02 3.852e+02 5.826e+02, threshold=6.540e+02, percent-clipped=0.0 2023-05-16 08:07:42,276 INFO [finetune.py:992] (0/2) Epoch 7, batch 8750, loss[loss=0.2071, simple_loss=0.2921, pruned_loss=0.06107, over 12130.00 frames. ], tot_loss[loss=0.1729, simple_loss=0.2608, pruned_loss=0.04246, over 2381832.16 frames. ], batch size: 38, lr: 4.39e-03, grad_scale: 16.0 2023-05-16 08:08:06,779 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=182186.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 08:08:11,143 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.3609, 3.0751, 4.8252, 2.4223, 2.5879, 3.5737, 3.0271, 3.7061], device='cuda:0'), covar=tensor([0.0511, 0.1274, 0.0377, 0.1266, 0.2045, 0.1637, 0.1426, 0.1227], device='cuda:0'), in_proj_covar=tensor([0.0234, 0.0234, 0.0239, 0.0182, 0.0234, 0.0288, 0.0223, 0.0255], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-16 08:08:18,795 INFO [finetune.py:992] (0/2) Epoch 7, batch 8800, loss[loss=0.1998, simple_loss=0.2833, pruned_loss=0.05813, over 12153.00 frames. ], tot_loss[loss=0.1735, simple_loss=0.2615, pruned_loss=0.04272, over 2379622.75 frames. ], batch size: 34, lr: 4.39e-03, grad_scale: 16.0 2023-05-16 08:08:40,922 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=182234.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 08:08:49,489 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.126e+02 2.888e+02 3.251e+02 4.135e+02 1.544e+03, threshold=6.502e+02, percent-clipped=3.0 2023-05-16 08:08:54,408 INFO [finetune.py:992] (0/2) Epoch 7, batch 8850, loss[loss=0.1365, simple_loss=0.2195, pruned_loss=0.02674, over 12294.00 frames. ], tot_loss[loss=0.1732, simple_loss=0.2617, pruned_loss=0.04241, over 2378801.45 frames. ], batch size: 28, lr: 4.39e-03, grad_scale: 16.0 2023-05-16 08:09:01,738 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([6.1296, 6.1246, 5.8359, 5.2636, 5.1322, 5.9696, 5.6418, 5.4065], device='cuda:0'), covar=tensor([0.0565, 0.0728, 0.0560, 0.1523, 0.0787, 0.0725, 0.1508, 0.1107], device='cuda:0'), in_proj_covar=tensor([0.0590, 0.0528, 0.0494, 0.0610, 0.0401, 0.0691, 0.0748, 0.0550], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0004, 0.0004, 0.0003], device='cuda:0') 2023-05-16 08:09:15,454 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=182281.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 08:09:30,508 INFO [finetune.py:992] (0/2) Epoch 7, batch 8900, loss[loss=0.165, simple_loss=0.2448, pruned_loss=0.04257, over 12348.00 frames. ], tot_loss[loss=0.1729, simple_loss=0.261, pruned_loss=0.04239, over 2378084.24 frames. ], batch size: 31, lr: 4.39e-03, grad_scale: 16.0 2023-05-16 08:09:53,207 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=182334.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 08:09:58,909 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=182342.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 08:10:01,589 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.160e+02 2.876e+02 3.336e+02 4.119e+02 1.220e+03, threshold=6.672e+02, percent-clipped=5.0 2023-05-16 08:10:06,504 INFO [finetune.py:992] (0/2) Epoch 7, batch 8950, loss[loss=0.1823, simple_loss=0.2688, pruned_loss=0.04795, over 10601.00 frames. ], tot_loss[loss=0.1731, simple_loss=0.2608, pruned_loss=0.04264, over 2372878.22 frames. ], batch size: 68, lr: 4.39e-03, grad_scale: 16.0 2023-05-16 08:10:35,463 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.90 vs. limit=2.0 2023-05-16 08:10:42,572 INFO [finetune.py:992] (0/2) Epoch 7, batch 9000, loss[loss=0.1714, simple_loss=0.2613, pruned_loss=0.04074, over 12123.00 frames. ], tot_loss[loss=0.175, simple_loss=0.2623, pruned_loss=0.04381, over 2356845.08 frames. ], batch size: 38, lr: 4.39e-03, grad_scale: 16.0 2023-05-16 08:10:42,573 INFO [finetune.py:1017] (0/2) Computing validation loss 2023-05-16 08:11:01,077 INFO [finetune.py:1026] (0/2) Epoch 7, validation: loss=0.3275, simple_loss=0.4008, pruned_loss=0.1272, over 1020973.00 frames. 2023-05-16 08:11:01,078 INFO [finetune.py:1027] (0/2) Maximum memory allocated so far is 12745MB 2023-05-16 08:11:32,192 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.081e+02 3.005e+02 3.281e+02 4.074e+02 8.769e+02, threshold=6.562e+02, percent-clipped=2.0 2023-05-16 08:11:33,131 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.2864, 2.4305, 3.1562, 4.1787, 2.2353, 4.1706, 4.2453, 4.4008], device='cuda:0'), covar=tensor([0.0097, 0.1232, 0.0453, 0.0120, 0.1294, 0.0240, 0.0160, 0.0082], device='cuda:0'), in_proj_covar=tensor([0.0118, 0.0203, 0.0185, 0.0115, 0.0188, 0.0178, 0.0173, 0.0119], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 08:11:35,866 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.7415, 4.4448, 4.4579, 4.6603, 4.4854, 4.6881, 4.4833, 2.3887], device='cuda:0'), covar=tensor([0.0114, 0.0090, 0.0128, 0.0086, 0.0066, 0.0114, 0.0108, 0.0871], device='cuda:0'), in_proj_covar=tensor([0.0065, 0.0076, 0.0079, 0.0072, 0.0058, 0.0089, 0.0078, 0.0094], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 08:11:37,061 INFO [finetune.py:992] (0/2) Epoch 7, batch 9050, loss[loss=0.1672, simple_loss=0.2442, pruned_loss=0.04516, over 12180.00 frames. ], tot_loss[loss=0.174, simple_loss=0.2614, pruned_loss=0.04328, over 2373588.50 frames. ], batch size: 29, lr: 4.39e-03, grad_scale: 16.0 2023-05-16 08:12:12,736 INFO [finetune.py:992] (0/2) Epoch 7, batch 9100, loss[loss=0.1676, simple_loss=0.2482, pruned_loss=0.04356, over 12174.00 frames. ], tot_loss[loss=0.1739, simple_loss=0.2615, pruned_loss=0.04315, over 2374971.04 frames. ], batch size: 29, lr: 4.39e-03, grad_scale: 16.0 2023-05-16 08:12:20,623 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2023-05-16 08:12:21,228 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=182514.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 08:12:29,993 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-05-16 08:12:43,818 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.814e+02 2.864e+02 3.251e+02 4.199e+02 6.849e+02, threshold=6.503e+02, percent-clipped=1.0 2023-05-16 08:12:48,840 INFO [finetune.py:992] (0/2) Epoch 7, batch 9150, loss[loss=0.188, simple_loss=0.2741, pruned_loss=0.05095, over 12046.00 frames. ], tot_loss[loss=0.1751, simple_loss=0.2625, pruned_loss=0.04381, over 2355874.09 frames. ], batch size: 42, lr: 4.39e-03, grad_scale: 16.0 2023-05-16 08:13:05,464 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=182575.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 08:13:22,614 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.91 vs. limit=2.0 2023-05-16 08:13:25,231 INFO [finetune.py:992] (0/2) Epoch 7, batch 9200, loss[loss=0.1944, simple_loss=0.2842, pruned_loss=0.0523, over 12102.00 frames. ], tot_loss[loss=0.1738, simple_loss=0.2614, pruned_loss=0.04315, over 2365518.80 frames. ], batch size: 38, lr: 4.39e-03, grad_scale: 16.0 2023-05-16 08:13:47,203 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=182634.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 08:13:49,274 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=182637.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 08:13:55,661 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.075e+02 2.773e+02 3.268e+02 3.915e+02 8.909e+02, threshold=6.536e+02, percent-clipped=1.0 2023-05-16 08:14:01,372 INFO [finetune.py:992] (0/2) Epoch 7, batch 9250, loss[loss=0.1471, simple_loss=0.2316, pruned_loss=0.03129, over 12353.00 frames. ], tot_loss[loss=0.1738, simple_loss=0.2618, pruned_loss=0.04294, over 2371049.71 frames. ], batch size: 31, lr: 4.39e-03, grad_scale: 16.0 2023-05-16 08:14:12,326 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.0991, 5.8848, 5.4781, 5.4699, 5.9888, 5.1706, 5.6183, 5.4305], device='cuda:0'), covar=tensor([0.1381, 0.0845, 0.0882, 0.1724, 0.0867, 0.2130, 0.1405, 0.1253], device='cuda:0'), in_proj_covar=tensor([0.0351, 0.0471, 0.0381, 0.0423, 0.0450, 0.0424, 0.0384, 0.0367], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 08:14:20,211 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.3473, 3.5084, 3.2502, 3.6584, 3.4001, 2.5758, 3.2738, 2.9257], device='cuda:0'), covar=tensor([0.0857, 0.0991, 0.1453, 0.0617, 0.1252, 0.1500, 0.1299, 0.2457], device='cuda:0'), in_proj_covar=tensor([0.0294, 0.0366, 0.0348, 0.0276, 0.0355, 0.0262, 0.0336, 0.0358], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 08:14:22,091 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=182682.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 08:14:37,012 INFO [finetune.py:992] (0/2) Epoch 7, batch 9300, loss[loss=0.1807, simple_loss=0.2631, pruned_loss=0.04915, over 10546.00 frames. ], tot_loss[loss=0.1739, simple_loss=0.2619, pruned_loss=0.04292, over 2366873.23 frames. ], batch size: 68, lr: 4.39e-03, grad_scale: 16.0 2023-05-16 08:14:58,759 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=4.43 vs. limit=5.0 2023-05-16 08:14:59,249 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.4855, 2.5098, 3.6093, 4.3803, 3.9226, 4.4158, 3.8202, 3.0654], device='cuda:0'), covar=tensor([0.0042, 0.0360, 0.0150, 0.0045, 0.0098, 0.0075, 0.0105, 0.0339], device='cuda:0'), in_proj_covar=tensor([0.0083, 0.0119, 0.0101, 0.0074, 0.0100, 0.0113, 0.0092, 0.0135], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 08:15:00,827 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.52 vs. limit=2.0 2023-05-16 08:15:08,401 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.716e+02 2.808e+02 3.401e+02 4.148e+02 1.022e+03, threshold=6.801e+02, percent-clipped=3.0 2023-05-16 08:15:11,438 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.4225, 2.6388, 3.1814, 4.3994, 2.1124, 4.2840, 4.4620, 4.5335], device='cuda:0'), covar=tensor([0.0134, 0.1130, 0.0428, 0.0114, 0.1442, 0.0222, 0.0138, 0.0085], device='cuda:0'), in_proj_covar=tensor([0.0120, 0.0205, 0.0187, 0.0116, 0.0189, 0.0180, 0.0175, 0.0121], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 08:15:13,307 INFO [finetune.py:992] (0/2) Epoch 7, batch 9350, loss[loss=0.1768, simple_loss=0.2656, pruned_loss=0.04396, over 12350.00 frames. ], tot_loss[loss=0.1743, simple_loss=0.2623, pruned_loss=0.04315, over 2360648.54 frames. ], batch size: 35, lr: 4.38e-03, grad_scale: 16.0 2023-05-16 08:15:49,515 INFO [finetune.py:992] (0/2) Epoch 7, batch 9400, loss[loss=0.1703, simple_loss=0.2655, pruned_loss=0.03752, over 12305.00 frames. ], tot_loss[loss=0.1744, simple_loss=0.2625, pruned_loss=0.04315, over 2361567.22 frames. ], batch size: 37, lr: 4.38e-03, grad_scale: 16.0 2023-05-16 08:16:08,583 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.56 vs. limit=2.0 2023-05-16 08:16:15,806 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 2023-05-16 08:16:20,422 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.201e+02 2.963e+02 3.390e+02 3.941e+02 6.612e+02, threshold=6.780e+02, percent-clipped=0.0 2023-05-16 08:16:25,538 INFO [finetune.py:992] (0/2) Epoch 7, batch 9450, loss[loss=0.1539, simple_loss=0.2461, pruned_loss=0.03081, over 12290.00 frames. ], tot_loss[loss=0.1735, simple_loss=0.262, pruned_loss=0.04253, over 2370857.63 frames. ], batch size: 33, lr: 4.38e-03, grad_scale: 16.0 2023-05-16 08:16:27,909 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=182856.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 08:16:29,427 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.9454, 3.5338, 5.2523, 2.7862, 2.8982, 3.8561, 3.6049, 3.8893], device='cuda:0'), covar=tensor([0.0373, 0.1060, 0.0272, 0.1194, 0.1833, 0.1504, 0.1066, 0.1159], device='cuda:0'), in_proj_covar=tensor([0.0234, 0.0232, 0.0240, 0.0181, 0.0234, 0.0288, 0.0223, 0.0254], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-16 08:16:32,909 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=182862.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 08:16:34,232 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([6.1071, 6.0954, 5.8998, 5.3257, 5.1957, 6.0010, 5.5894, 5.3695], device='cuda:0'), covar=tensor([0.0664, 0.0799, 0.0583, 0.1496, 0.0608, 0.0606, 0.1263, 0.0968], device='cuda:0'), in_proj_covar=tensor([0.0588, 0.0526, 0.0491, 0.0605, 0.0400, 0.0686, 0.0742, 0.0548], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0004, 0.0004, 0.0003], device='cuda:0') 2023-05-16 08:16:38,348 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=182870.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 08:17:01,687 INFO [finetune.py:992] (0/2) Epoch 7, batch 9500, loss[loss=0.1518, simple_loss=0.2441, pruned_loss=0.02974, over 12026.00 frames. ], tot_loss[loss=0.1734, simple_loss=0.2622, pruned_loss=0.0423, over 2370868.93 frames. ], batch size: 31, lr: 4.38e-03, grad_scale: 16.0 2023-05-16 08:17:04,023 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.7043, 2.8131, 3.3760, 4.5707, 2.6146, 4.5154, 4.7136, 4.8459], device='cuda:0'), covar=tensor([0.0094, 0.1133, 0.0424, 0.0128, 0.1159, 0.0243, 0.0117, 0.0090], device='cuda:0'), in_proj_covar=tensor([0.0118, 0.0202, 0.0184, 0.0115, 0.0186, 0.0178, 0.0173, 0.0119], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 08:17:10,796 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=3.63 vs. limit=5.0 2023-05-16 08:17:11,959 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=182917.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 08:17:16,182 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=182923.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 08:17:26,735 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=182937.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 08:17:31,042 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.2854, 4.7687, 2.9016, 2.7625, 4.0138, 2.6417, 4.0059, 3.1490], device='cuda:0'), covar=tensor([0.0757, 0.0473, 0.1223, 0.1487, 0.0316, 0.1330, 0.0478, 0.0849], device='cuda:0'), in_proj_covar=tensor([0.0188, 0.0252, 0.0176, 0.0198, 0.0139, 0.0179, 0.0196, 0.0173], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 08:17:32,935 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.126e+02 2.717e+02 3.235e+02 4.149e+02 5.868e+02, threshold=6.470e+02, percent-clipped=0.0 2023-05-16 08:17:37,012 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.43 vs. limit=2.0 2023-05-16 08:17:38,015 INFO [finetune.py:992] (0/2) Epoch 7, batch 9550, loss[loss=0.1652, simple_loss=0.2614, pruned_loss=0.03456, over 12366.00 frames. ], tot_loss[loss=0.1723, simple_loss=0.2611, pruned_loss=0.04177, over 2374063.62 frames. ], batch size: 35, lr: 4.38e-03, grad_scale: 16.0 2023-05-16 08:17:44,556 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.1514, 4.8282, 4.9581, 5.0615, 4.8410, 5.0749, 4.9149, 2.5291], device='cuda:0'), covar=tensor([0.0072, 0.0051, 0.0070, 0.0052, 0.0040, 0.0066, 0.0055, 0.0747], device='cuda:0'), in_proj_covar=tensor([0.0064, 0.0075, 0.0078, 0.0071, 0.0058, 0.0088, 0.0077, 0.0092], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 08:18:00,862 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=182985.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 08:18:09,146 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.83 vs. limit=2.0 2023-05-16 08:18:13,725 INFO [finetune.py:992] (0/2) Epoch 7, batch 9600, loss[loss=0.1543, simple_loss=0.2457, pruned_loss=0.0315, over 12109.00 frames. ], tot_loss[loss=0.173, simple_loss=0.2619, pruned_loss=0.04208, over 2379746.50 frames. ], batch size: 33, lr: 4.38e-03, grad_scale: 16.0 2023-05-16 08:18:44,924 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.945e+02 2.671e+02 3.193e+02 3.784e+02 7.294e+02, threshold=6.385e+02, percent-clipped=2.0 2023-05-16 08:18:49,912 INFO [finetune.py:992] (0/2) Epoch 7, batch 9650, loss[loss=0.2879, simple_loss=0.3421, pruned_loss=0.1168, over 7970.00 frames. ], tot_loss[loss=0.1739, simple_loss=0.2627, pruned_loss=0.0426, over 2377622.11 frames. ], batch size: 98, lr: 4.38e-03, grad_scale: 16.0 2023-05-16 08:19:10,790 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 2023-05-16 08:19:16,970 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.6669, 5.1598, 5.6368, 4.9859, 5.2199, 5.0701, 5.6590, 5.1868], device='cuda:0'), covar=tensor([0.0171, 0.0263, 0.0159, 0.0203, 0.0270, 0.0236, 0.0140, 0.0188], device='cuda:0'), in_proj_covar=tensor([0.0242, 0.0249, 0.0271, 0.0246, 0.0245, 0.0244, 0.0220, 0.0205], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 08:19:26,003 INFO [finetune.py:992] (0/2) Epoch 7, batch 9700, loss[loss=0.1805, simple_loss=0.2732, pruned_loss=0.04389, over 12299.00 frames. ], tot_loss[loss=0.1739, simple_loss=0.2625, pruned_loss=0.0427, over 2375177.97 frames. ], batch size: 34, lr: 4.38e-03, grad_scale: 16.0 2023-05-16 08:19:56,664 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.233e+02 2.941e+02 3.488e+02 4.384e+02 8.404e+02, threshold=6.976e+02, percent-clipped=4.0 2023-05-16 08:19:57,617 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.7890, 2.5741, 3.7684, 4.6936, 4.0448, 4.7030, 3.9621, 3.2473], device='cuda:0'), covar=tensor([0.0031, 0.0357, 0.0135, 0.0036, 0.0116, 0.0062, 0.0106, 0.0326], device='cuda:0'), in_proj_covar=tensor([0.0084, 0.0121, 0.0103, 0.0076, 0.0102, 0.0115, 0.0094, 0.0138], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 08:20:01,768 INFO [finetune.py:992] (0/2) Epoch 7, batch 9750, loss[loss=0.1447, simple_loss=0.2275, pruned_loss=0.03091, over 12014.00 frames. ], tot_loss[loss=0.1736, simple_loss=0.2618, pruned_loss=0.04264, over 2375880.73 frames. ], batch size: 28, lr: 4.38e-03, grad_scale: 16.0 2023-05-16 08:20:14,281 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=183170.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 08:20:26,573 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.5181, 2.3668, 3.5880, 4.3836, 3.9595, 4.3713, 3.8611, 2.8859], device='cuda:0'), covar=tensor([0.0030, 0.0348, 0.0144, 0.0034, 0.0094, 0.0074, 0.0100, 0.0369], device='cuda:0'), in_proj_covar=tensor([0.0083, 0.0120, 0.0102, 0.0075, 0.0101, 0.0114, 0.0093, 0.0137], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 08:20:37,922 INFO [finetune.py:992] (0/2) Epoch 7, batch 9800, loss[loss=0.201, simple_loss=0.2976, pruned_loss=0.05221, over 10339.00 frames. ], tot_loss[loss=0.1736, simple_loss=0.262, pruned_loss=0.04261, over 2375314.68 frames. ], batch size: 68, lr: 4.38e-03, grad_scale: 16.0 2023-05-16 08:20:44,401 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=183212.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 08:20:48,530 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=183218.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 08:20:48,545 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=183218.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 08:21:09,309 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.830e+02 2.964e+02 3.446e+02 4.126e+02 1.231e+03, threshold=6.891e+02, percent-clipped=2.0 2023-05-16 08:21:14,336 INFO [finetune.py:992] (0/2) Epoch 7, batch 9850, loss[loss=0.1645, simple_loss=0.2473, pruned_loss=0.04089, over 12085.00 frames. ], tot_loss[loss=0.1736, simple_loss=0.2623, pruned_loss=0.04241, over 2378611.25 frames. ], batch size: 32, lr: 4.38e-03, grad_scale: 16.0 2023-05-16 08:21:50,517 INFO [finetune.py:992] (0/2) Epoch 7, batch 9900, loss[loss=0.1758, simple_loss=0.2611, pruned_loss=0.04518, over 12424.00 frames. ], tot_loss[loss=0.1732, simple_loss=0.2616, pruned_loss=0.0424, over 2381421.04 frames. ], batch size: 32, lr: 4.38e-03, grad_scale: 16.0 2023-05-16 08:22:20,992 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.974e+02 2.943e+02 3.305e+02 4.141e+02 8.788e+02, threshold=6.610e+02, percent-clipped=4.0 2023-05-16 08:22:25,966 INFO [finetune.py:992] (0/2) Epoch 7, batch 9950, loss[loss=0.1571, simple_loss=0.2492, pruned_loss=0.03254, over 12106.00 frames. ], tot_loss[loss=0.1732, simple_loss=0.2613, pruned_loss=0.04257, over 2375023.75 frames. ], batch size: 33, lr: 4.38e-03, grad_scale: 32.0 2023-05-16 08:22:28,690 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.40 vs. limit=2.0 2023-05-16 08:22:36,043 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.4845, 3.6824, 3.2624, 3.1585, 2.8375, 2.6923, 3.6345, 2.1723], device='cuda:0'), covar=tensor([0.0371, 0.0100, 0.0154, 0.0204, 0.0386, 0.0359, 0.0136, 0.0489], device='cuda:0'), in_proj_covar=tensor([0.0184, 0.0154, 0.0150, 0.0175, 0.0195, 0.0190, 0.0160, 0.0195], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 08:22:48,759 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.3302, 4.8174, 2.9211, 2.6479, 4.1410, 2.7017, 4.1106, 3.3086], device='cuda:0'), covar=tensor([0.0669, 0.0378, 0.1077, 0.1443, 0.0231, 0.1166, 0.0404, 0.0725], device='cuda:0'), in_proj_covar=tensor([0.0189, 0.0253, 0.0176, 0.0198, 0.0139, 0.0180, 0.0197, 0.0173], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 08:23:02,360 INFO [finetune.py:992] (0/2) Epoch 7, batch 10000, loss[loss=0.1442, simple_loss=0.2344, pruned_loss=0.02703, over 12338.00 frames. ], tot_loss[loss=0.173, simple_loss=0.2612, pruned_loss=0.04242, over 2379466.46 frames. ], batch size: 30, lr: 4.38e-03, grad_scale: 32.0 2023-05-16 08:23:06,191 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=183408.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 08:23:32,562 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.862e+02 2.814e+02 3.286e+02 3.856e+02 7.018e+02, threshold=6.571e+02, percent-clipped=1.0 2023-05-16 08:23:38,157 INFO [finetune.py:992] (0/2) Epoch 7, batch 10050, loss[loss=0.1658, simple_loss=0.2523, pruned_loss=0.0396, over 12346.00 frames. ], tot_loss[loss=0.1729, simple_loss=0.2611, pruned_loss=0.04239, over 2380287.64 frames. ], batch size: 31, lr: 4.38e-03, grad_scale: 32.0 2023-05-16 08:23:49,658 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=183469.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 08:24:07,408 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.8045, 4.3990, 4.5339, 4.5946, 4.3973, 4.5512, 4.5478, 2.3364], device='cuda:0'), covar=tensor([0.0084, 0.0078, 0.0106, 0.0073, 0.0063, 0.0105, 0.0094, 0.0880], device='cuda:0'), in_proj_covar=tensor([0.0065, 0.0075, 0.0078, 0.0070, 0.0058, 0.0088, 0.0077, 0.0092], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 08:24:14,311 INFO [finetune.py:992] (0/2) Epoch 7, batch 10100, loss[loss=0.134, simple_loss=0.2171, pruned_loss=0.02551, over 12170.00 frames. ], tot_loss[loss=0.1729, simple_loss=0.2608, pruned_loss=0.04249, over 2369181.19 frames. ], batch size: 29, lr: 4.38e-03, grad_scale: 32.0 2023-05-16 08:24:20,712 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=183512.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 08:24:24,867 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=183518.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 08:24:44,689 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.971e+02 2.789e+02 3.208e+02 4.103e+02 9.250e+02, threshold=6.416e+02, percent-clipped=2.0 2023-05-16 08:24:49,698 INFO [finetune.py:992] (0/2) Epoch 7, batch 10150, loss[loss=0.1382, simple_loss=0.2241, pruned_loss=0.02619, over 12118.00 frames. ], tot_loss[loss=0.1733, simple_loss=0.2615, pruned_loss=0.0425, over 2369446.93 frames. ], batch size: 30, lr: 4.38e-03, grad_scale: 32.0 2023-05-16 08:24:54,779 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=183560.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 08:24:59,112 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=183566.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 08:25:05,162 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-05-16 08:25:26,264 INFO [finetune.py:992] (0/2) Epoch 7, batch 10200, loss[loss=0.1666, simple_loss=0.2507, pruned_loss=0.04124, over 12150.00 frames. ], tot_loss[loss=0.1736, simple_loss=0.2618, pruned_loss=0.0427, over 2365576.54 frames. ], batch size: 34, lr: 4.38e-03, grad_scale: 32.0 2023-05-16 08:25:34,299 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=183614.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 08:25:38,807 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.86 vs. limit=2.0 2023-05-16 08:25:43,000 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.3778, 3.3533, 3.1092, 3.0555, 2.7718, 2.5923, 3.3889, 2.1591], device='cuda:0'), covar=tensor([0.0348, 0.0118, 0.0134, 0.0171, 0.0325, 0.0325, 0.0140, 0.0430], device='cuda:0'), in_proj_covar=tensor([0.0186, 0.0155, 0.0152, 0.0177, 0.0197, 0.0192, 0.0162, 0.0197], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 08:25:57,615 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.847e+02 2.740e+02 3.151e+02 3.893e+02 9.370e+02, threshold=6.301e+02, percent-clipped=2.0 2023-05-16 08:26:02,583 INFO [finetune.py:992] (0/2) Epoch 7, batch 10250, loss[loss=0.1867, simple_loss=0.2716, pruned_loss=0.05085, over 12352.00 frames. ], tot_loss[loss=0.1732, simple_loss=0.2613, pruned_loss=0.04254, over 2372401.60 frames. ], batch size: 36, lr: 4.38e-03, grad_scale: 32.0 2023-05-16 08:26:14,073 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.8766, 5.5023, 5.1370, 5.1254, 5.6235, 5.0107, 5.1438, 5.0808], device='cuda:0'), covar=tensor([0.1410, 0.1192, 0.1159, 0.1991, 0.1049, 0.2135, 0.2106, 0.1131], device='cuda:0'), in_proj_covar=tensor([0.0352, 0.0475, 0.0383, 0.0425, 0.0454, 0.0428, 0.0385, 0.0369], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 08:26:18,523 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=183675.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 08:26:37,971 INFO [finetune.py:992] (0/2) Epoch 7, batch 10300, loss[loss=0.1567, simple_loss=0.2377, pruned_loss=0.03785, over 12200.00 frames. ], tot_loss[loss=0.1737, simple_loss=0.2619, pruned_loss=0.04272, over 2384334.41 frames. ], batch size: 31, lr: 4.37e-03, grad_scale: 32.0 2023-05-16 08:27:08,681 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.806e+02 2.995e+02 3.469e+02 4.013e+02 6.943e+02, threshold=6.938e+02, percent-clipped=2.0 2023-05-16 08:27:13,567 INFO [finetune.py:992] (0/2) Epoch 7, batch 10350, loss[loss=0.1711, simple_loss=0.2646, pruned_loss=0.03877, over 12288.00 frames. ], tot_loss[loss=0.1741, simple_loss=0.2623, pruned_loss=0.04293, over 2375870.93 frames. ], batch size: 34, lr: 4.37e-03, grad_scale: 32.0 2023-05-16 08:27:21,463 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=183764.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 08:27:50,108 INFO [finetune.py:992] (0/2) Epoch 7, batch 10400, loss[loss=0.1789, simple_loss=0.2518, pruned_loss=0.05301, over 12090.00 frames. ], tot_loss[loss=0.1739, simple_loss=0.2623, pruned_loss=0.04276, over 2382347.43 frames. ], batch size: 32, lr: 4.37e-03, grad_scale: 32.0 2023-05-16 08:27:56,694 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.4439, 5.2723, 5.3194, 5.4138, 5.0669, 5.0436, 4.8443, 5.3100], device='cuda:0'), covar=tensor([0.0572, 0.0509, 0.0657, 0.0516, 0.1576, 0.1305, 0.0488, 0.0968], device='cuda:0'), in_proj_covar=tensor([0.0513, 0.0671, 0.0572, 0.0609, 0.0820, 0.0715, 0.0530, 0.0479], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0005, 0.0004, 0.0004, 0.0005, 0.0005, 0.0004, 0.0004], device='cuda:0') 2023-05-16 08:28:06,827 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.4903, 4.7376, 4.2010, 5.1650, 4.5175, 3.1232, 4.3670, 3.0351], device='cuda:0'), covar=tensor([0.0756, 0.0857, 0.1483, 0.0392, 0.1142, 0.1587, 0.0977, 0.3479], device='cuda:0'), in_proj_covar=tensor([0.0297, 0.0368, 0.0352, 0.0278, 0.0358, 0.0263, 0.0336, 0.0361], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 08:28:17,788 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=183842.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 08:28:20,391 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.037e+02 3.060e+02 3.451e+02 4.187e+02 8.708e+02, threshold=6.903e+02, percent-clipped=1.0 2023-05-16 08:28:24,132 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.6650, 4.6402, 4.5126, 4.1197, 4.3144, 4.6440, 4.3398, 4.1664], device='cuda:0'), covar=tensor([0.0808, 0.0935, 0.0656, 0.1471, 0.1683, 0.0842, 0.1491, 0.1222], device='cuda:0'), in_proj_covar=tensor([0.0601, 0.0530, 0.0499, 0.0611, 0.0404, 0.0693, 0.0750, 0.0556], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0004, 0.0004, 0.0003], device='cuda:0') 2023-05-16 08:28:25,334 INFO [finetune.py:992] (0/2) Epoch 7, batch 10450, loss[loss=0.1957, simple_loss=0.2946, pruned_loss=0.04846, over 11700.00 frames. ], tot_loss[loss=0.1738, simple_loss=0.2622, pruned_loss=0.04272, over 2372681.14 frames. ], batch size: 48, lr: 4.37e-03, grad_scale: 32.0 2023-05-16 08:28:35,976 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.42 vs. limit=2.0 2023-05-16 08:28:37,168 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.63 vs. limit=2.0 2023-05-16 08:29:01,466 INFO [finetune.py:992] (0/2) Epoch 7, batch 10500, loss[loss=0.1841, simple_loss=0.2709, pruned_loss=0.04866, over 12142.00 frames. ], tot_loss[loss=0.1738, simple_loss=0.2621, pruned_loss=0.04273, over 2372798.96 frames. ], batch size: 36, lr: 4.37e-03, grad_scale: 32.0 2023-05-16 08:29:01,669 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=183903.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 08:29:17,957 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.5998, 2.6432, 4.3718, 4.7142, 3.1019, 2.4825, 2.7163, 1.8894], device='cuda:0'), covar=tensor([0.1356, 0.2938, 0.0457, 0.0346, 0.0949, 0.2037, 0.2741, 0.4111], device='cuda:0'), in_proj_covar=tensor([0.0280, 0.0375, 0.0267, 0.0291, 0.0257, 0.0288, 0.0363, 0.0361], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-16 08:29:22,602 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.0683, 2.3871, 3.4131, 3.8688, 3.7346, 4.0276, 3.5597, 2.8929], device='cuda:0'), covar=tensor([0.0041, 0.0354, 0.0139, 0.0052, 0.0095, 0.0063, 0.0120, 0.0325], device='cuda:0'), in_proj_covar=tensor([0.0083, 0.0120, 0.0101, 0.0075, 0.0102, 0.0114, 0.0094, 0.0137], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 08:29:32,329 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.051e+02 2.934e+02 3.579e+02 4.359e+02 7.289e+02, threshold=7.158e+02, percent-clipped=1.0 2023-05-16 08:29:37,362 INFO [finetune.py:992] (0/2) Epoch 7, batch 10550, loss[loss=0.1868, simple_loss=0.2869, pruned_loss=0.04333, over 12324.00 frames. ], tot_loss[loss=0.1733, simple_loss=0.2617, pruned_loss=0.04246, over 2377852.22 frames. ], batch size: 36, lr: 4.37e-03, grad_scale: 32.0 2023-05-16 08:29:49,467 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=183970.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 08:30:10,847 INFO [checkpoint.py:75] (0/2) Saving checkpoint to pruned_transducer_stateless7/exp_giga_finetune/checkpoint-84000.pt 2023-05-16 08:30:15,428 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.7520, 3.3332, 5.1330, 2.8665, 2.8889, 3.9145, 3.1764, 3.9962], device='cuda:0'), covar=tensor([0.0445, 0.1120, 0.0340, 0.1136, 0.1783, 0.1275, 0.1314, 0.0886], device='cuda:0'), in_proj_covar=tensor([0.0236, 0.0231, 0.0242, 0.0181, 0.0235, 0.0286, 0.0223, 0.0255], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-16 08:30:15,892 INFO [finetune.py:992] (0/2) Epoch 7, batch 10600, loss[loss=0.1859, simple_loss=0.2743, pruned_loss=0.04876, over 12305.00 frames. ], tot_loss[loss=0.175, simple_loss=0.2633, pruned_loss=0.04339, over 2364911.87 frames. ], batch size: 34, lr: 4.37e-03, grad_scale: 32.0 2023-05-16 08:30:19,940 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.42 vs. limit=2.0 2023-05-16 08:30:27,030 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.4828, 5.3349, 5.3917, 5.4622, 5.0546, 5.0797, 4.8671, 5.4022], device='cuda:0'), covar=tensor([0.0591, 0.0553, 0.0777, 0.0494, 0.1890, 0.1357, 0.0513, 0.0946], device='cuda:0'), in_proj_covar=tensor([0.0513, 0.0672, 0.0570, 0.0611, 0.0826, 0.0714, 0.0532, 0.0476], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0005, 0.0004, 0.0004, 0.0005, 0.0005, 0.0004, 0.0003], device='cuda:0') 2023-05-16 08:30:46,414 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 2023-05-16 08:30:46,779 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.156e+02 2.875e+02 3.336e+02 4.059e+02 8.731e+02, threshold=6.672e+02, percent-clipped=2.0 2023-05-16 08:30:49,895 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=184050.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 08:30:51,797 INFO [finetune.py:992] (0/2) Epoch 7, batch 10650, loss[loss=0.1561, simple_loss=0.2398, pruned_loss=0.03616, over 12140.00 frames. ], tot_loss[loss=0.1737, simple_loss=0.2617, pruned_loss=0.04279, over 2370104.62 frames. ], batch size: 30, lr: 4.37e-03, grad_scale: 32.0 2023-05-16 08:31:00,113 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=184064.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 08:31:28,362 INFO [finetune.py:992] (0/2) Epoch 7, batch 10700, loss[loss=0.1837, simple_loss=0.2744, pruned_loss=0.04647, over 12345.00 frames. ], tot_loss[loss=0.1738, simple_loss=0.262, pruned_loss=0.04276, over 2364756.50 frames. ], batch size: 35, lr: 4.37e-03, grad_scale: 32.0 2023-05-16 08:31:34,236 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=184111.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 08:31:34,768 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=184112.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 08:31:58,556 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.949e+02 2.790e+02 3.201e+02 3.867e+02 6.972e+02, threshold=6.402e+02, percent-clipped=2.0 2023-05-16 08:32:03,546 INFO [finetune.py:992] (0/2) Epoch 7, batch 10750, loss[loss=0.1669, simple_loss=0.2605, pruned_loss=0.03663, over 12350.00 frames. ], tot_loss[loss=0.1733, simple_loss=0.2615, pruned_loss=0.04256, over 2367671.88 frames. ], batch size: 36, lr: 4.37e-03, grad_scale: 32.0 2023-05-16 08:32:35,825 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=184198.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 08:32:39,516 INFO [finetune.py:992] (0/2) Epoch 7, batch 10800, loss[loss=0.1686, simple_loss=0.2678, pruned_loss=0.03469, over 12145.00 frames. ], tot_loss[loss=0.1735, simple_loss=0.2615, pruned_loss=0.04274, over 2364368.13 frames. ], batch size: 36, lr: 4.37e-03, grad_scale: 32.0 2023-05-16 08:33:10,725 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.798e+02 2.797e+02 3.306e+02 4.077e+02 6.295e+02, threshold=6.611e+02, percent-clipped=0.0 2023-05-16 08:33:15,686 INFO [finetune.py:992] (0/2) Epoch 7, batch 10850, loss[loss=0.1411, simple_loss=0.2215, pruned_loss=0.03035, over 11772.00 frames. ], tot_loss[loss=0.1731, simple_loss=0.2612, pruned_loss=0.04248, over 2372821.46 frames. ], batch size: 26, lr: 4.37e-03, grad_scale: 32.0 2023-05-16 08:33:25,805 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=3.22 vs. limit=5.0 2023-05-16 08:33:27,599 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=184270.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 08:33:36,696 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=3.90 vs. limit=5.0 2023-05-16 08:33:51,623 INFO [finetune.py:992] (0/2) Epoch 7, batch 10900, loss[loss=0.1589, simple_loss=0.2503, pruned_loss=0.03372, over 12264.00 frames. ], tot_loss[loss=0.1741, simple_loss=0.2618, pruned_loss=0.0432, over 2361719.21 frames. ], batch size: 32, lr: 4.37e-03, grad_scale: 32.0 2023-05-16 08:34:00,540 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.0260, 2.2759, 2.2488, 2.2750, 2.0984, 2.0308, 2.1680, 1.7053], device='cuda:0'), covar=tensor([0.0278, 0.0188, 0.0182, 0.0189, 0.0310, 0.0254, 0.0189, 0.0354], device='cuda:0'), in_proj_covar=tensor([0.0185, 0.0156, 0.0154, 0.0177, 0.0195, 0.0192, 0.0161, 0.0196], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 08:34:02,487 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=184318.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 08:34:22,844 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.205e+02 2.977e+02 3.512e+02 4.014e+02 7.651e+02, threshold=7.023e+02, percent-clipped=3.0 2023-05-16 08:34:27,667 INFO [finetune.py:992] (0/2) Epoch 7, batch 10950, loss[loss=0.1684, simple_loss=0.264, pruned_loss=0.0364, over 12342.00 frames. ], tot_loss[loss=0.175, simple_loss=0.2627, pruned_loss=0.04361, over 2352079.28 frames. ], batch size: 36, lr: 4.37e-03, grad_scale: 32.0 2023-05-16 08:34:32,159 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=184359.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 08:34:35,507 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.8285, 3.6155, 3.6746, 3.7218, 3.4635, 3.8470, 3.8305, 3.9518], device='cuda:0'), covar=tensor([0.0262, 0.0223, 0.0220, 0.0456, 0.0663, 0.0431, 0.0186, 0.0243], device='cuda:0'), in_proj_covar=tensor([0.0186, 0.0186, 0.0184, 0.0236, 0.0237, 0.0207, 0.0169, 0.0217], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0005, 0.0004, 0.0004, 0.0003, 0.0005], device='cuda:0') 2023-05-16 08:35:04,164 INFO [finetune.py:992] (0/2) Epoch 7, batch 11000, loss[loss=0.165, simple_loss=0.25, pruned_loss=0.03998, over 12089.00 frames. ], tot_loss[loss=0.1758, simple_loss=0.2636, pruned_loss=0.04407, over 2341047.40 frames. ], batch size: 32, lr: 4.37e-03, grad_scale: 32.0 2023-05-16 08:35:06,343 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=184406.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 08:35:16,340 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=184420.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 08:35:34,559 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.046e+02 3.155e+02 3.689e+02 4.357e+02 7.918e+02, threshold=7.378e+02, percent-clipped=3.0 2023-05-16 08:35:40,146 INFO [finetune.py:992] (0/2) Epoch 7, batch 11050, loss[loss=0.1703, simple_loss=0.2559, pruned_loss=0.04238, over 12185.00 frames. ], tot_loss[loss=0.1778, simple_loss=0.2655, pruned_loss=0.04501, over 2315925.78 frames. ], batch size: 31, lr: 4.37e-03, grad_scale: 32.0 2023-05-16 08:35:54,179 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.6642, 2.2915, 2.9969, 2.6453, 2.8718, 2.9306, 2.1339, 2.9667], device='cuda:0'), covar=tensor([0.0110, 0.0303, 0.0123, 0.0182, 0.0131, 0.0126, 0.0295, 0.0130], device='cuda:0'), in_proj_covar=tensor([0.0168, 0.0200, 0.0183, 0.0177, 0.0206, 0.0155, 0.0192, 0.0174], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 08:36:03,844 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=184486.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 08:36:12,183 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=184498.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 08:36:15,440 INFO [finetune.py:992] (0/2) Epoch 7, batch 11100, loss[loss=0.2561, simple_loss=0.3202, pruned_loss=0.09601, over 7947.00 frames. ], tot_loss[loss=0.1829, simple_loss=0.2699, pruned_loss=0.048, over 2275796.17 frames. ], batch size: 98, lr: 4.37e-03, grad_scale: 32.0 2023-05-16 08:36:23,673 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=4.63 vs. limit=5.0 2023-05-16 08:36:28,356 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.1872, 3.1016, 4.7807, 2.6091, 2.6259, 3.7378, 2.8964, 3.8494], device='cuda:0'), covar=tensor([0.0638, 0.1258, 0.0222, 0.1158, 0.1929, 0.1156, 0.1503, 0.0828], device='cuda:0'), in_proj_covar=tensor([0.0228, 0.0225, 0.0234, 0.0177, 0.0228, 0.0277, 0.0216, 0.0247], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:0') 2023-05-16 08:36:41,338 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 2023-05-16 08:36:45,886 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.179e+02 3.470e+02 4.055e+02 4.748e+02 9.924e+02, threshold=8.110e+02, percent-clipped=6.0 2023-05-16 08:36:45,988 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=184546.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 08:36:46,802 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=184547.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 08:36:51,510 INFO [finetune.py:992] (0/2) Epoch 7, batch 11150, loss[loss=0.1998, simple_loss=0.2877, pruned_loss=0.05599, over 12366.00 frames. ], tot_loss[loss=0.1913, simple_loss=0.2772, pruned_loss=0.0527, over 2202476.83 frames. ], batch size: 35, lr: 4.37e-03, grad_scale: 32.0 2023-05-16 08:37:25,981 INFO [finetune.py:992] (0/2) Epoch 7, batch 11200, loss[loss=0.3486, simple_loss=0.3952, pruned_loss=0.151, over 6545.00 frames. ], tot_loss[loss=0.1993, simple_loss=0.2844, pruned_loss=0.05716, over 2151354.48 frames. ], batch size: 101, lr: 4.36e-03, grad_scale: 16.0 2023-05-16 08:37:57,915 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.468e+02 3.616e+02 4.361e+02 5.521e+02 1.127e+03, threshold=8.722e+02, percent-clipped=4.0 2023-05-16 08:38:02,376 INFO [finetune.py:992] (0/2) Epoch 7, batch 11250, loss[loss=0.2492, simple_loss=0.3367, pruned_loss=0.0808, over 10496.00 frames. ], tot_loss[loss=0.2085, simple_loss=0.2917, pruned_loss=0.06263, over 2081428.22 frames. ], batch size: 69, lr: 4.36e-03, grad_scale: 16.0 2023-05-16 08:38:09,382 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.5568, 4.7104, 4.2469, 5.0000, 4.6271, 2.9912, 4.3205, 3.0408], device='cuda:0'), covar=tensor([0.0622, 0.0731, 0.1244, 0.0426, 0.1039, 0.1477, 0.0948, 0.3345], device='cuda:0'), in_proj_covar=tensor([0.0290, 0.0360, 0.0343, 0.0270, 0.0349, 0.0259, 0.0328, 0.0352], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 08:38:16,870 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.98 vs. limit=2.0 2023-05-16 08:38:37,122 INFO [finetune.py:992] (0/2) Epoch 7, batch 11300, loss[loss=0.3216, simple_loss=0.3703, pruned_loss=0.1364, over 6518.00 frames. ], tot_loss[loss=0.2171, simple_loss=0.2985, pruned_loss=0.06783, over 1997099.90 frames. ], batch size: 98, lr: 4.36e-03, grad_scale: 16.0 2023-05-16 08:38:39,966 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=184706.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 08:38:46,063 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=184715.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 08:38:46,172 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.6933, 2.1982, 3.3301, 3.4225, 3.5643, 3.5931, 3.4437, 2.5069], device='cuda:0'), covar=tensor([0.0058, 0.0401, 0.0138, 0.0069, 0.0094, 0.0093, 0.0118, 0.0421], device='cuda:0'), in_proj_covar=tensor([0.0081, 0.0117, 0.0099, 0.0074, 0.0098, 0.0111, 0.0091, 0.0133], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 08:38:58,725 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.2425, 2.9440, 3.6195, 2.3865, 2.5739, 3.0749, 2.8904, 3.2407], device='cuda:0'), covar=tensor([0.0569, 0.0930, 0.0297, 0.1148, 0.1653, 0.1159, 0.1068, 0.0772], device='cuda:0'), in_proj_covar=tensor([0.0225, 0.0221, 0.0228, 0.0174, 0.0224, 0.0272, 0.0212, 0.0241], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:0') 2023-05-16 08:39:08,029 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.431e+02 3.560e+02 4.219e+02 5.102e+02 7.690e+02, threshold=8.438e+02, percent-clipped=0.0 2023-05-16 08:39:12,199 INFO [finetune.py:992] (0/2) Epoch 7, batch 11350, loss[loss=0.2771, simple_loss=0.3442, pruned_loss=0.105, over 6925.00 frames. ], tot_loss[loss=0.2225, simple_loss=0.3029, pruned_loss=0.071, over 1937936.45 frames. ], batch size: 102, lr: 4.36e-03, grad_scale: 16.0 2023-05-16 08:39:13,589 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=184754.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 08:39:47,660 INFO [finetune.py:992] (0/2) Epoch 7, batch 11400, loss[loss=0.2608, simple_loss=0.328, pruned_loss=0.09685, over 7020.00 frames. ], tot_loss[loss=0.2279, simple_loss=0.3075, pruned_loss=0.07409, over 1885949.11 frames. ], batch size: 101, lr: 4.36e-03, grad_scale: 16.0 2023-05-16 08:40:15,354 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=184842.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 08:40:18,629 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.447e+02 3.774e+02 4.236e+02 5.048e+02 8.901e+02, threshold=8.472e+02, percent-clipped=1.0 2023-05-16 08:40:22,821 INFO [finetune.py:992] (0/2) Epoch 7, batch 11450, loss[loss=0.243, simple_loss=0.3227, pruned_loss=0.08167, over 6752.00 frames. ], tot_loss[loss=0.233, simple_loss=0.3116, pruned_loss=0.07718, over 1848307.78 frames. ], batch size: 98, lr: 4.36e-03, grad_scale: 16.0 2023-05-16 08:40:31,880 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=184866.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 08:40:57,129 INFO [finetune.py:992] (0/2) Epoch 7, batch 11500, loss[loss=0.2867, simple_loss=0.3496, pruned_loss=0.1119, over 6805.00 frames. ], tot_loss[loss=0.2367, simple_loss=0.3143, pruned_loss=0.07952, over 1813152.62 frames. ], batch size: 98, lr: 4.36e-03, grad_scale: 16.0 2023-05-16 08:41:14,046 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=184927.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 08:41:16,234 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.9490, 2.1338, 2.6771, 3.0512, 2.2525, 3.1814, 3.0312, 3.1726], device='cuda:0'), covar=tensor([0.0162, 0.1098, 0.0452, 0.0155, 0.1024, 0.0250, 0.0308, 0.0130], device='cuda:0'), in_proj_covar=tensor([0.0113, 0.0199, 0.0176, 0.0109, 0.0181, 0.0169, 0.0166, 0.0114], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 08:41:19,542 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.3033, 5.2921, 5.1201, 4.7519, 4.6673, 5.2674, 4.9497, 4.8605], device='cuda:0'), covar=tensor([0.0635, 0.0840, 0.0661, 0.1464, 0.1271, 0.0711, 0.1445, 0.0895], device='cuda:0'), in_proj_covar=tensor([0.0561, 0.0501, 0.0473, 0.0578, 0.0383, 0.0647, 0.0701, 0.0523], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 08:41:22,331 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=184939.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 08:41:28,239 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.564e+02 3.530e+02 3.937e+02 4.658e+02 1.050e+03, threshold=7.874e+02, percent-clipped=1.0 2023-05-16 08:41:32,249 INFO [finetune.py:992] (0/2) Epoch 7, batch 11550, loss[loss=0.2738, simple_loss=0.337, pruned_loss=0.1053, over 7192.00 frames. ], tot_loss[loss=0.2396, simple_loss=0.3166, pruned_loss=0.08131, over 1782060.55 frames. ], batch size: 99, lr: 4.36e-03, grad_scale: 16.0 2023-05-16 08:42:05,494 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=185000.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 08:42:07,362 INFO [finetune.py:992] (0/2) Epoch 7, batch 11600, loss[loss=0.2609, simple_loss=0.3282, pruned_loss=0.09674, over 7066.00 frames. ], tot_loss[loss=0.2409, simple_loss=0.3169, pruned_loss=0.08251, over 1760794.12 frames. ], batch size: 98, lr: 4.36e-03, grad_scale: 16.0 2023-05-16 08:42:16,090 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=185015.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 08:42:31,808 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.8051, 3.0736, 2.3533, 2.1711, 2.7263, 2.2613, 2.9619, 2.5676], device='cuda:0'), covar=tensor([0.0502, 0.0525, 0.0846, 0.1402, 0.0263, 0.1136, 0.0420, 0.0733], device='cuda:0'), in_proj_covar=tensor([0.0177, 0.0235, 0.0166, 0.0188, 0.0132, 0.0172, 0.0183, 0.0165], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 08:42:39,267 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.815e+02 3.850e+02 4.381e+02 5.033e+02 1.342e+03, threshold=8.762e+02, percent-clipped=5.0 2023-05-16 08:42:43,806 INFO [finetune.py:992] (0/2) Epoch 7, batch 11650, loss[loss=0.2789, simple_loss=0.332, pruned_loss=0.1129, over 6449.00 frames. ], tot_loss[loss=0.2408, simple_loss=0.3162, pruned_loss=0.08265, over 1746356.30 frames. ], batch size: 98, lr: 4.36e-03, grad_scale: 16.0 2023-05-16 08:42:49,131 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=185060.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 08:42:51,003 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=185063.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 08:43:05,304 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.0351, 2.2464, 2.7399, 2.8409, 2.9103, 2.9740, 2.8997, 2.4035], device='cuda:0'), covar=tensor([0.0061, 0.0306, 0.0138, 0.0065, 0.0093, 0.0086, 0.0099, 0.0328], device='cuda:0'), in_proj_covar=tensor([0.0079, 0.0116, 0.0097, 0.0072, 0.0096, 0.0109, 0.0088, 0.0130], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 08:43:19,129 INFO [finetune.py:992] (0/2) Epoch 7, batch 11700, loss[loss=0.2704, simple_loss=0.3397, pruned_loss=0.1005, over 6768.00 frames. ], tot_loss[loss=0.2412, simple_loss=0.3161, pruned_loss=0.08313, over 1719164.48 frames. ], batch size: 98, lr: 4.36e-03, grad_scale: 16.0 2023-05-16 08:43:31,359 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=185121.0, num_to_drop=1, layers_to_drop={0} 2023-05-16 08:43:37,472 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=185129.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 08:43:43,732 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.89 vs. limit=2.0 2023-05-16 08:43:45,964 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=185142.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 08:43:49,176 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.257e+02 3.357e+02 4.006e+02 4.852e+02 1.709e+03, threshold=8.012e+02, percent-clipped=2.0 2023-05-16 08:43:50,070 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.9024, 3.0894, 4.4270, 2.5121, 2.5067, 3.4625, 2.9410, 3.6114], device='cuda:0'), covar=tensor([0.0628, 0.1208, 0.0197, 0.1280, 0.2128, 0.1242, 0.1545, 0.0838], device='cuda:0'), in_proj_covar=tensor([0.0223, 0.0221, 0.0224, 0.0174, 0.0225, 0.0270, 0.0213, 0.0238], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:0') 2023-05-16 08:43:53,183 INFO [finetune.py:992] (0/2) Epoch 7, batch 11750, loss[loss=0.2383, simple_loss=0.3217, pruned_loss=0.07742, over 10251.00 frames. ], tot_loss[loss=0.2423, simple_loss=0.3166, pruned_loss=0.08401, over 1711265.31 frames. ], batch size: 69, lr: 4.36e-03, grad_scale: 16.0 2023-05-16 08:44:05,638 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.4768, 4.3569, 4.2872, 4.3464, 3.9548, 4.5040, 4.4919, 4.6013], device='cuda:0'), covar=tensor([0.0203, 0.0131, 0.0181, 0.0286, 0.0707, 0.0273, 0.0141, 0.0174], device='cuda:0'), in_proj_covar=tensor([0.0165, 0.0165, 0.0162, 0.0209, 0.0208, 0.0182, 0.0149, 0.0195], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:0') 2023-05-16 08:44:18,940 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=185190.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 08:44:19,085 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=185190.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 08:44:22,463 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.0230, 2.3828, 3.4788, 3.8658, 3.7914, 4.0132, 3.6862, 2.7351], device='cuda:0'), covar=tensor([0.0047, 0.0395, 0.0123, 0.0053, 0.0084, 0.0068, 0.0096, 0.0395], device='cuda:0'), in_proj_covar=tensor([0.0079, 0.0116, 0.0097, 0.0072, 0.0096, 0.0109, 0.0088, 0.0131], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 08:44:23,367 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=3.28 vs. limit=5.0 2023-05-16 08:44:28,632 INFO [finetune.py:992] (0/2) Epoch 7, batch 11800, loss[loss=0.2422, simple_loss=0.3284, pruned_loss=0.07804, over 10433.00 frames. ], tot_loss[loss=0.2427, simple_loss=0.3176, pruned_loss=0.08395, over 1723562.14 frames. ], batch size: 69, lr: 4.36e-03, grad_scale: 16.0 2023-05-16 08:44:42,318 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=185222.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 08:44:55,581 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.38 vs. limit=2.0 2023-05-16 08:44:59,079 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.628e+02 3.541e+02 4.087e+02 4.932e+02 1.149e+03, threshold=8.173e+02, percent-clipped=3.0 2023-05-16 08:45:03,772 INFO [finetune.py:992] (0/2) Epoch 7, batch 11850, loss[loss=0.244, simple_loss=0.314, pruned_loss=0.08695, over 6834.00 frames. ], tot_loss[loss=0.244, simple_loss=0.319, pruned_loss=0.0845, over 1699114.92 frames. ], batch size: 98, lr: 4.36e-03, grad_scale: 16.0 2023-05-16 08:45:32,662 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=185295.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 08:45:37,874 INFO [finetune.py:992] (0/2) Epoch 7, batch 11900, loss[loss=0.2555, simple_loss=0.3248, pruned_loss=0.09315, over 7385.00 frames. ], tot_loss[loss=0.2425, simple_loss=0.3185, pruned_loss=0.08321, over 1698475.06 frames. ], batch size: 99, lr: 4.36e-03, grad_scale: 16.0 2023-05-16 08:45:49,476 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=4.07 vs. limit=5.0 2023-05-16 08:46:09,046 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.271e+02 3.494e+02 4.092e+02 4.883e+02 7.543e+02, threshold=8.183e+02, percent-clipped=0.0 2023-05-16 08:46:13,133 INFO [finetune.py:992] (0/2) Epoch 7, batch 11950, loss[loss=0.198, simple_loss=0.2899, pruned_loss=0.05302, over 11029.00 frames. ], tot_loss[loss=0.2376, simple_loss=0.3149, pruned_loss=0.08012, over 1693206.74 frames. ], batch size: 55, lr: 4.36e-03, grad_scale: 16.0 2023-05-16 08:46:20,851 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.8370, 4.5083, 4.1801, 4.3041, 4.5622, 4.0409, 4.2833, 4.0402], device='cuda:0'), covar=tensor([0.1751, 0.1081, 0.1200, 0.1696, 0.0983, 0.2118, 0.1444, 0.1123], device='cuda:0'), in_proj_covar=tensor([0.0329, 0.0437, 0.0359, 0.0394, 0.0416, 0.0394, 0.0351, 0.0339], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002], device='cuda:0') 2023-05-16 08:46:38,491 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.37 vs. limit=2.0 2023-05-16 08:46:48,699 INFO [finetune.py:992] (0/2) Epoch 7, batch 12000, loss[loss=0.1969, simple_loss=0.2884, pruned_loss=0.05266, over 10318.00 frames. ], tot_loss[loss=0.2316, simple_loss=0.3103, pruned_loss=0.07642, over 1691064.60 frames. ], batch size: 68, lr: 4.36e-03, grad_scale: 16.0 2023-05-16 08:46:48,700 INFO [finetune.py:1017] (0/2) Computing validation loss 2023-05-16 08:47:04,627 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.1380, 2.7286, 3.4958, 2.2255, 2.4310, 2.9285, 2.7257, 3.1066], device='cuda:0'), covar=tensor([0.0515, 0.1055, 0.0248, 0.1239, 0.1816, 0.1244, 0.1103, 0.0863], device='cuda:0'), in_proj_covar=tensor([0.0218, 0.0216, 0.0217, 0.0170, 0.0220, 0.0262, 0.0208, 0.0232], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:0') 2023-05-16 08:47:06,956 INFO [finetune.py:1026] (0/2) Epoch 7, validation: loss=0.2884, simple_loss=0.3646, pruned_loss=0.106, over 1020973.00 frames. 2023-05-16 08:47:06,957 INFO [finetune.py:1027] (0/2) Maximum memory allocated so far is 12761MB 2023-05-16 08:47:16,079 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=185416.0, num_to_drop=1, layers_to_drop={3} 2023-05-16 08:47:37,285 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.158e+02 3.019e+02 3.476e+02 4.163e+02 9.429e+02, threshold=6.953e+02, percent-clipped=2.0 2023-05-16 08:47:41,308 INFO [finetune.py:992] (0/2) Epoch 7, batch 12050, loss[loss=0.2046, simple_loss=0.2819, pruned_loss=0.06362, over 6961.00 frames. ], tot_loss[loss=0.2272, simple_loss=0.3065, pruned_loss=0.07391, over 1696815.85 frames. ], batch size: 102, lr: 4.36e-03, grad_scale: 16.0 2023-05-16 08:47:51,414 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=185468.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 08:48:01,859 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=185485.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 08:48:13,817 INFO [finetune.py:992] (0/2) Epoch 7, batch 12100, loss[loss=0.2291, simple_loss=0.301, pruned_loss=0.07855, over 6889.00 frames. ], tot_loss[loss=0.2254, simple_loss=0.3051, pruned_loss=0.07286, over 1692479.42 frames. ], batch size: 98, lr: 4.36e-03, grad_scale: 16.0 2023-05-16 08:48:25,989 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=185522.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 08:48:30,528 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=185529.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 08:48:37,226 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=4.85 vs. limit=5.0 2023-05-16 08:48:41,728 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.380e+02 3.250e+02 3.695e+02 4.693e+02 9.793e+02, threshold=7.390e+02, percent-clipped=3.0 2023-05-16 08:48:46,015 INFO [finetune.py:992] (0/2) Epoch 7, batch 12150, loss[loss=0.2571, simple_loss=0.3191, pruned_loss=0.09756, over 7331.00 frames. ], tot_loss[loss=0.2276, simple_loss=0.3065, pruned_loss=0.07435, over 1667838.53 frames. ], batch size: 100, lr: 4.35e-03, grad_scale: 16.0 2023-05-16 08:48:56,663 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=185570.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 08:48:59,168 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.4983, 4.5094, 4.4081, 4.0842, 4.1427, 4.4961, 4.2580, 4.1399], device='cuda:0'), covar=tensor([0.0787, 0.0853, 0.0685, 0.1229, 0.1858, 0.0808, 0.1361, 0.1016], device='cuda:0'), in_proj_covar=tensor([0.0550, 0.0490, 0.0459, 0.0558, 0.0369, 0.0628, 0.0677, 0.0506], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002], device='cuda:0') 2023-05-16 08:49:02,733 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=4.53 vs. limit=5.0 2023-05-16 08:49:12,551 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=185595.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 08:49:17,616 INFO [finetune.py:992] (0/2) Epoch 7, batch 12200, loss[loss=0.244, simple_loss=0.3138, pruned_loss=0.08709, over 7139.00 frames. ], tot_loss[loss=0.2293, simple_loss=0.3076, pruned_loss=0.07552, over 1641028.59 frames. ], batch size: 98, lr: 4.35e-03, grad_scale: 16.0 2023-05-16 08:49:26,487 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.0482, 2.2983, 2.2333, 2.2452, 2.1296, 2.0386, 2.1966, 1.7043], device='cuda:0'), covar=tensor([0.0299, 0.0169, 0.0205, 0.0201, 0.0311, 0.0208, 0.0157, 0.0393], device='cuda:0'), in_proj_covar=tensor([0.0177, 0.0144, 0.0143, 0.0168, 0.0184, 0.0180, 0.0149, 0.0185], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], device='cuda:0') 2023-05-16 08:49:32,817 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=4.71 vs. limit=5.0 2023-05-16 08:49:39,438 INFO [checkpoint.py:75] (0/2) Saving checkpoint to pruned_transducer_stateless7/exp_giga_finetune/epoch-7.pt 2023-05-16 08:50:02,407 INFO [finetune.py:992] (0/2) Epoch 8, batch 0, loss[loss=0.1996, simple_loss=0.2963, pruned_loss=0.05143, over 12259.00 frames. ], tot_loss[loss=0.1996, simple_loss=0.2963, pruned_loss=0.05143, over 12259.00 frames. ], batch size: 37, lr: 4.35e-03, grad_scale: 16.0 2023-05-16 08:50:02,407 INFO [finetune.py:1017] (0/2) Computing validation loss 2023-05-16 08:50:20,129 INFO [finetune.py:1026] (0/2) Epoch 8, validation: loss=0.2881, simple_loss=0.3643, pruned_loss=0.1059, over 1020973.00 frames. 2023-05-16 08:50:20,129 INFO [finetune.py:1027] (0/2) Maximum memory allocated so far is 12761MB 2023-05-16 08:50:24,969 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=185643.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 08:50:27,586 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.234e+02 3.471e+02 4.101e+02 4.868e+02 9.966e+02, threshold=8.203e+02, percent-clipped=2.0 2023-05-16 08:50:55,817 INFO [finetune.py:992] (0/2) Epoch 8, batch 50, loss[loss=0.1935, simple_loss=0.2813, pruned_loss=0.05281, over 12024.00 frames. ], tot_loss[loss=0.1871, simple_loss=0.2756, pruned_loss=0.04929, over 536843.52 frames. ], batch size: 42, lr: 4.35e-03, grad_scale: 16.0 2023-05-16 08:51:16,978 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=185716.0, num_to_drop=1, layers_to_drop={2} 2023-05-16 08:51:31,694 INFO [finetune.py:992] (0/2) Epoch 8, batch 100, loss[loss=0.2059, simple_loss=0.2917, pruned_loss=0.06009, over 12049.00 frames. ], tot_loss[loss=0.1829, simple_loss=0.2709, pruned_loss=0.04748, over 947859.12 frames. ], batch size: 37, lr: 4.35e-03, grad_scale: 16.0 2023-05-16 08:51:38,815 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.875e+02 2.687e+02 3.242e+02 3.856e+02 6.671e+02, threshold=6.484e+02, percent-clipped=0.0 2023-05-16 08:51:47,558 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=185759.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 08:51:47,684 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.7989, 2.7384, 4.6388, 4.8214, 2.8899, 2.6103, 2.8929, 2.0171], device='cuda:0'), covar=tensor([0.1498, 0.3245, 0.0438, 0.0388, 0.1357, 0.2505, 0.2976, 0.4482], device='cuda:0'), in_proj_covar=tensor([0.0274, 0.0365, 0.0258, 0.0281, 0.0251, 0.0283, 0.0356, 0.0357], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-16 08:51:51,019 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=185764.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 08:52:05,514 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=185785.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 08:52:06,843 INFO [finetune.py:992] (0/2) Epoch 8, batch 150, loss[loss=0.1707, simple_loss=0.2511, pruned_loss=0.04517, over 12179.00 frames. ], tot_loss[loss=0.18, simple_loss=0.2682, pruned_loss=0.04587, over 1269276.06 frames. ], batch size: 31, lr: 4.35e-03, grad_scale: 16.0 2023-05-16 08:52:15,901 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=185799.0, num_to_drop=1, layers_to_drop={1} 2023-05-16 08:52:31,144 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=185820.0, num_to_drop=1, layers_to_drop={3} 2023-05-16 08:52:33,803 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=185824.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 08:52:40,219 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=185833.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 08:52:42,950 INFO [finetune.py:992] (0/2) Epoch 8, batch 200, loss[loss=0.1594, simple_loss=0.2508, pruned_loss=0.03395, over 12279.00 frames. ], tot_loss[loss=0.1794, simple_loss=0.2676, pruned_loss=0.04562, over 1514248.63 frames. ], batch size: 33, lr: 4.35e-03, grad_scale: 16.0 2023-05-16 08:52:49,889 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.034e+02 2.956e+02 3.477e+02 4.036e+02 6.994e+02, threshold=6.953e+02, percent-clipped=1.0 2023-05-16 08:52:59,240 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=185860.0, num_to_drop=1, layers_to_drop={1} 2023-05-16 08:53:19,069 INFO [finetune.py:992] (0/2) Epoch 8, batch 250, loss[loss=0.1543, simple_loss=0.2353, pruned_loss=0.03664, over 12365.00 frames. ], tot_loss[loss=0.1776, simple_loss=0.2657, pruned_loss=0.04472, over 1708458.64 frames. ], batch size: 30, lr: 4.35e-03, grad_scale: 16.0 2023-05-16 08:53:35,094 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.5639, 3.5082, 3.1285, 3.2682, 2.7939, 2.6356, 3.4450, 2.2105], device='cuda:0'), covar=tensor([0.0351, 0.0123, 0.0166, 0.0156, 0.0385, 0.0389, 0.0133, 0.0456], device='cuda:0'), in_proj_covar=tensor([0.0180, 0.0147, 0.0145, 0.0172, 0.0188, 0.0184, 0.0153, 0.0189], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], device='cuda:0') 2023-05-16 08:53:52,798 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.91 vs. limit=2.0 2023-05-16 08:53:54,366 INFO [finetune.py:992] (0/2) Epoch 8, batch 300, loss[loss=0.175, simple_loss=0.259, pruned_loss=0.04547, over 12095.00 frames. ], tot_loss[loss=0.1772, simple_loss=0.2652, pruned_loss=0.04461, over 1858934.72 frames. ], batch size: 33, lr: 4.35e-03, grad_scale: 16.0 2023-05-16 08:54:02,109 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.818e+02 2.706e+02 3.258e+02 3.879e+02 6.810e+02, threshold=6.516e+02, percent-clipped=0.0 2023-05-16 08:54:08,025 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-05-16 08:54:30,313 INFO [finetune.py:992] (0/2) Epoch 8, batch 350, loss[loss=0.1894, simple_loss=0.2776, pruned_loss=0.0506, over 11640.00 frames. ], tot_loss[loss=0.1769, simple_loss=0.2653, pruned_loss=0.04427, over 1968768.24 frames. ], batch size: 48, lr: 4.35e-03, grad_scale: 16.0 2023-05-16 08:54:40,036 INFO [checkpoint.py:75] (0/2) Saving checkpoint to pruned_transducer_stateless7/exp_giga_finetune/checkpoint-86000.pt 2023-05-16 08:55:09,930 INFO [finetune.py:992] (0/2) Epoch 8, batch 400, loss[loss=0.1548, simple_loss=0.2379, pruned_loss=0.03583, over 12133.00 frames. ], tot_loss[loss=0.1776, simple_loss=0.2658, pruned_loss=0.04471, over 2053206.07 frames. ], batch size: 30, lr: 4.35e-03, grad_scale: 16.0 2023-05-16 08:55:17,162 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.904e+02 2.883e+02 3.350e+02 4.048e+02 8.627e+02, threshold=6.701e+02, percent-clipped=3.0 2023-05-16 08:55:46,533 INFO [finetune.py:992] (0/2) Epoch 8, batch 450, loss[loss=0.1436, simple_loss=0.2312, pruned_loss=0.028, over 12348.00 frames. ], tot_loss[loss=0.1763, simple_loss=0.264, pruned_loss=0.04433, over 2122255.16 frames. ], batch size: 30, lr: 4.35e-03, grad_scale: 16.0 2023-05-16 08:55:56,007 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=186100.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 08:56:06,448 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=186115.0, num_to_drop=1, layers_to_drop={0} 2023-05-16 08:56:09,614 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-05-16 08:56:12,786 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=186124.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 08:56:16,185 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.6265, 2.5947, 3.6396, 4.4919, 4.1062, 4.5286, 3.9975, 3.0447], device='cuda:0'), covar=tensor([0.0024, 0.0344, 0.0126, 0.0035, 0.0097, 0.0054, 0.0089, 0.0318], device='cuda:0'), in_proj_covar=tensor([0.0081, 0.0119, 0.0100, 0.0073, 0.0098, 0.0112, 0.0090, 0.0133], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 08:56:21,563 INFO [finetune.py:992] (0/2) Epoch 8, batch 500, loss[loss=0.2001, simple_loss=0.288, pruned_loss=0.05614, over 12048.00 frames. ], tot_loss[loss=0.1766, simple_loss=0.2644, pruned_loss=0.04436, over 2174242.20 frames. ], batch size: 40, lr: 4.35e-03, grad_scale: 16.0 2023-05-16 08:56:28,710 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.814e+02 2.866e+02 3.465e+02 4.025e+02 1.738e+03, threshold=6.929e+02, percent-clipped=2.0 2023-05-16 08:56:34,505 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.3801, 4.8479, 5.3549, 4.6308, 4.9430, 4.7579, 5.3998, 5.0247], device='cuda:0'), covar=tensor([0.0222, 0.0381, 0.0257, 0.0274, 0.0384, 0.0302, 0.0183, 0.0225], device='cuda:0'), in_proj_covar=tensor([0.0231, 0.0238, 0.0256, 0.0235, 0.0233, 0.0232, 0.0211, 0.0189], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 08:56:35,131 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=186155.0, num_to_drop=1, layers_to_drop={2} 2023-05-16 08:56:35,980 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.5857, 2.8557, 3.3011, 4.5256, 2.4676, 4.4788, 4.6045, 4.7052], device='cuda:0'), covar=tensor([0.0088, 0.1090, 0.0463, 0.0093, 0.1231, 0.0189, 0.0113, 0.0071], device='cuda:0'), in_proj_covar=tensor([0.0115, 0.0201, 0.0178, 0.0110, 0.0185, 0.0170, 0.0166, 0.0115], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 08:56:37,306 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=186158.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 08:56:39,405 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=186161.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 08:56:46,983 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=186172.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 08:56:53,647 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 2023-05-16 08:56:57,530 INFO [finetune.py:992] (0/2) Epoch 8, batch 550, loss[loss=0.1914, simple_loss=0.2872, pruned_loss=0.04781, over 11015.00 frames. ], tot_loss[loss=0.1763, simple_loss=0.2645, pruned_loss=0.04409, over 2225011.93 frames. ], batch size: 55, lr: 4.35e-03, grad_scale: 16.0 2023-05-16 08:57:20,654 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=186219.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 08:57:32,903 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.2652, 4.6862, 4.1133, 4.9565, 4.6193, 2.9380, 4.2648, 3.1130], device='cuda:0'), covar=tensor([0.0807, 0.0768, 0.1363, 0.0454, 0.0981, 0.1690, 0.1080, 0.3395], device='cuda:0'), in_proj_covar=tensor([0.0293, 0.0364, 0.0348, 0.0269, 0.0354, 0.0265, 0.0329, 0.0359], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 08:57:33,986 INFO [finetune.py:992] (0/2) Epoch 8, batch 600, loss[loss=0.1591, simple_loss=0.2392, pruned_loss=0.03955, over 12033.00 frames. ], tot_loss[loss=0.176, simple_loss=0.2639, pruned_loss=0.04406, over 2252223.00 frames. ], batch size: 31, lr: 4.35e-03, grad_scale: 16.0 2023-05-16 08:57:41,031 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.962e+02 2.802e+02 3.192e+02 3.860e+02 7.758e+02, threshold=6.385e+02, percent-clipped=2.0 2023-05-16 08:57:49,232 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.3455, 4.4113, 4.2139, 4.6672, 3.0867, 4.1438, 2.7764, 4.2446], device='cuda:0'), covar=tensor([0.1654, 0.0570, 0.0812, 0.0540, 0.1086, 0.0562, 0.1668, 0.1353], device='cuda:0'), in_proj_covar=tensor([0.0224, 0.0259, 0.0288, 0.0338, 0.0234, 0.0236, 0.0255, 0.0369], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 08:57:49,878 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=186259.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 08:57:51,180 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.3906, 4.7501, 4.1437, 5.0507, 4.5682, 3.0895, 4.2280, 3.1109], device='cuda:0'), covar=tensor([0.0842, 0.0803, 0.1541, 0.0419, 0.1163, 0.1638, 0.1223, 0.3438], device='cuda:0'), in_proj_covar=tensor([0.0293, 0.0364, 0.0348, 0.0269, 0.0354, 0.0265, 0.0329, 0.0359], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 08:58:08,672 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.4366, 3.5758, 3.2129, 3.3265, 2.9170, 2.7733, 3.7141, 2.5029], device='cuda:0'), covar=tensor([0.0365, 0.0138, 0.0181, 0.0175, 0.0365, 0.0332, 0.0104, 0.0391], device='cuda:0'), in_proj_covar=tensor([0.0183, 0.0151, 0.0150, 0.0174, 0.0192, 0.0188, 0.0156, 0.0192], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], device='cuda:0') 2023-05-16 08:58:09,155 INFO [finetune.py:992] (0/2) Epoch 8, batch 650, loss[loss=0.1633, simple_loss=0.2464, pruned_loss=0.04013, over 12096.00 frames. ], tot_loss[loss=0.1752, simple_loss=0.2635, pruned_loss=0.04346, over 2282468.43 frames. ], batch size: 32, lr: 4.35e-03, grad_scale: 16.0 2023-05-16 08:58:33,274 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=186320.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 08:58:36,279 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.2668, 4.1750, 4.1084, 4.4954, 3.0257, 4.0397, 2.8280, 4.0752], device='cuda:0'), covar=tensor([0.1598, 0.0665, 0.0916, 0.0630, 0.1187, 0.0585, 0.1724, 0.1495], device='cuda:0'), in_proj_covar=tensor([0.0224, 0.0260, 0.0289, 0.0340, 0.0235, 0.0236, 0.0255, 0.0371], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 08:58:45,397 INFO [finetune.py:992] (0/2) Epoch 8, batch 700, loss[loss=0.2149, simple_loss=0.2942, pruned_loss=0.0678, over 8000.00 frames. ], tot_loss[loss=0.1746, simple_loss=0.2631, pruned_loss=0.043, over 2306144.43 frames. ], batch size: 98, lr: 4.34e-03, grad_scale: 16.0 2023-05-16 08:58:52,349 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.920e+02 2.942e+02 3.398e+02 4.313e+02 1.039e+03, threshold=6.795e+02, percent-clipped=2.0 2023-05-16 08:59:05,372 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.8563, 4.4942, 4.5413, 4.6685, 4.4046, 4.7688, 4.5261, 2.4024], device='cuda:0'), covar=tensor([0.0087, 0.0072, 0.0109, 0.0069, 0.0066, 0.0089, 0.0108, 0.0853], device='cuda:0'), in_proj_covar=tensor([0.0063, 0.0074, 0.0078, 0.0069, 0.0057, 0.0087, 0.0076, 0.0094], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-16 08:59:14,721 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.3836, 6.2411, 5.8215, 5.7780, 6.2641, 5.6551, 5.8414, 5.7623], device='cuda:0'), covar=tensor([0.1459, 0.0707, 0.0939, 0.1899, 0.0866, 0.2023, 0.1434, 0.0984], device='cuda:0'), in_proj_covar=tensor([0.0344, 0.0456, 0.0372, 0.0409, 0.0435, 0.0418, 0.0371, 0.0356], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002], device='cuda:0') 2023-05-16 08:59:17,635 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.5355, 4.9858, 5.4537, 4.7422, 5.0154, 4.8163, 5.5367, 5.1261], device='cuda:0'), covar=tensor([0.0246, 0.0373, 0.0244, 0.0276, 0.0379, 0.0306, 0.0192, 0.0197], device='cuda:0'), in_proj_covar=tensor([0.0238, 0.0244, 0.0265, 0.0242, 0.0240, 0.0240, 0.0218, 0.0196], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 08:59:21,703 INFO [finetune.py:992] (0/2) Epoch 8, batch 750, loss[loss=0.1888, simple_loss=0.2709, pruned_loss=0.05332, over 11804.00 frames. ], tot_loss[loss=0.1739, simple_loss=0.2622, pruned_loss=0.04278, over 2330346.63 frames. ], batch size: 44, lr: 4.34e-03, grad_scale: 16.0 2023-05-16 08:59:23,971 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.5172, 2.5971, 3.6006, 4.4340, 3.9906, 4.4867, 3.8664, 3.0973], device='cuda:0'), covar=tensor([0.0038, 0.0381, 0.0140, 0.0036, 0.0102, 0.0065, 0.0106, 0.0339], device='cuda:0'), in_proj_covar=tensor([0.0083, 0.0119, 0.0100, 0.0073, 0.0099, 0.0113, 0.0092, 0.0134], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 08:59:41,934 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=186415.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 08:59:46,367 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=186421.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 08:59:57,632 INFO [finetune.py:992] (0/2) Epoch 8, batch 800, loss[loss=0.1404, simple_loss=0.2226, pruned_loss=0.02912, over 12351.00 frames. ], tot_loss[loss=0.174, simple_loss=0.2625, pruned_loss=0.04271, over 2342491.18 frames. ], batch size: 30, lr: 4.34e-03, grad_scale: 16.0 2023-05-16 09:00:05,528 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.027e+02 2.737e+02 3.197e+02 3.965e+02 6.142e+02, threshold=6.393e+02, percent-clipped=0.0 2023-05-16 09:00:11,177 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=186455.0, num_to_drop=1, layers_to_drop={0} 2023-05-16 09:00:11,862 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=186456.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 09:00:15,339 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.42 vs. limit=2.0 2023-05-16 09:00:15,926 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.96 vs. limit=2.0 2023-05-16 09:00:16,838 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=186463.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 09:00:25,351 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=186475.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 09:00:30,545 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=186482.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 09:00:33,839 INFO [finetune.py:992] (0/2) Epoch 8, batch 850, loss[loss=0.1557, simple_loss=0.2356, pruned_loss=0.03795, over 12373.00 frames. ], tot_loss[loss=0.1729, simple_loss=0.2613, pruned_loss=0.04224, over 2350790.65 frames. ], batch size: 30, lr: 4.34e-03, grad_scale: 16.0 2023-05-16 09:00:45,462 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=186503.0, num_to_drop=1, layers_to_drop={1} 2023-05-16 09:00:52,773 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=186514.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 09:00:55,076 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=186517.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 09:01:09,588 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=186536.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 09:01:10,077 INFO [finetune.py:992] (0/2) Epoch 8, batch 900, loss[loss=0.1701, simple_loss=0.2649, pruned_loss=0.03766, over 12133.00 frames. ], tot_loss[loss=0.1724, simple_loss=0.2607, pruned_loss=0.0421, over 2354818.42 frames. ], batch size: 38, lr: 4.34e-03, grad_scale: 16.0 2023-05-16 09:01:11,714 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=186539.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 09:01:17,272 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.088e+02 2.865e+02 3.396e+02 4.022e+02 7.338e+02, threshold=6.792e+02, percent-clipped=3.0 2023-05-16 09:01:31,840 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=186567.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 09:01:39,792 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=186578.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 09:01:45,266 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=186586.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 09:01:45,745 INFO [finetune.py:992] (0/2) Epoch 8, batch 950, loss[loss=0.1758, simple_loss=0.2692, pruned_loss=0.04117, over 10547.00 frames. ], tot_loss[loss=0.1722, simple_loss=0.2606, pruned_loss=0.04192, over 2359472.84 frames. ], batch size: 68, lr: 4.34e-03, grad_scale: 16.0 2023-05-16 09:01:56,180 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=186600.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 09:02:06,762 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=186615.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 09:02:13,779 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.1804, 6.0475, 5.5855, 5.5984, 6.0626, 5.4682, 5.6305, 5.5802], device='cuda:0'), covar=tensor([0.1351, 0.0746, 0.0922, 0.1479, 0.0770, 0.1812, 0.1678, 0.1109], device='cuda:0'), in_proj_covar=tensor([0.0342, 0.0455, 0.0370, 0.0408, 0.0434, 0.0416, 0.0369, 0.0356], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002], device='cuda:0') 2023-05-16 09:02:15,985 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=186628.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 09:02:16,083 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.6014, 2.6691, 4.3290, 4.5500, 2.8703, 2.4769, 2.8053, 2.1007], device='cuda:0'), covar=tensor([0.1464, 0.3125, 0.0516, 0.0409, 0.1193, 0.2285, 0.2851, 0.3917], device='cuda:0'), in_proj_covar=tensor([0.0278, 0.0371, 0.0262, 0.0288, 0.0253, 0.0287, 0.0362, 0.0361], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-16 09:02:22,394 INFO [finetune.py:992] (0/2) Epoch 8, batch 1000, loss[loss=0.1737, simple_loss=0.2715, pruned_loss=0.03798, over 12199.00 frames. ], tot_loss[loss=0.1723, simple_loss=0.2607, pruned_loss=0.04197, over 2364776.65 frames. ], batch size: 35, lr: 4.34e-03, grad_scale: 32.0 2023-05-16 09:02:29,251 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.887e+02 2.785e+02 3.250e+02 3.932e+02 7.318e+02, threshold=6.500e+02, percent-clipped=1.0 2023-05-16 09:02:29,499 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=186647.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 09:02:58,255 INFO [finetune.py:992] (0/2) Epoch 8, batch 1050, loss[loss=0.1563, simple_loss=0.2373, pruned_loss=0.03766, over 12156.00 frames. ], tot_loss[loss=0.1722, simple_loss=0.2604, pruned_loss=0.04198, over 2363095.63 frames. ], batch size: 29, lr: 4.34e-03, grad_scale: 32.0 2023-05-16 09:03:13,875 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=186709.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 09:03:15,078 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.83 vs. limit=2.0 2023-05-16 09:03:24,878 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-05-16 09:03:33,886 INFO [finetune.py:992] (0/2) Epoch 8, batch 1100, loss[loss=0.1881, simple_loss=0.2783, pruned_loss=0.04897, over 12105.00 frames. ], tot_loss[loss=0.1708, simple_loss=0.259, pruned_loss=0.04131, over 2368883.50 frames. ], batch size: 38, lr: 4.34e-03, grad_scale: 32.0 2023-05-16 09:03:41,716 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.854e+02 2.786e+02 3.230e+02 3.834e+02 9.626e+02, threshold=6.461e+02, percent-clipped=1.0 2023-05-16 09:03:45,767 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.65 vs. limit=2.0 2023-05-16 09:03:48,322 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=186756.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 09:03:58,218 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=186770.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 09:04:03,183 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=186777.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 09:04:10,230 INFO [finetune.py:992] (0/2) Epoch 8, batch 1150, loss[loss=0.1756, simple_loss=0.2632, pruned_loss=0.04402, over 12391.00 frames. ], tot_loss[loss=0.1705, simple_loss=0.2586, pruned_loss=0.04122, over 2379848.98 frames. ], batch size: 38, lr: 4.34e-03, grad_scale: 32.0 2023-05-16 09:04:22,613 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=186804.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 09:04:29,856 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=186814.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 09:04:42,719 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=186831.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 09:04:47,043 INFO [finetune.py:992] (0/2) Epoch 8, batch 1200, loss[loss=0.1902, simple_loss=0.2774, pruned_loss=0.05149, over 10698.00 frames. ], tot_loss[loss=0.1713, simple_loss=0.2594, pruned_loss=0.04164, over 2373321.01 frames. ], batch size: 68, lr: 4.34e-03, grad_scale: 32.0 2023-05-16 09:04:49,396 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=186840.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 09:04:54,250 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.126e+02 3.005e+02 3.396e+02 3.981e+02 6.798e+02, threshold=6.791e+02, percent-clipped=4.0 2023-05-16 09:05:04,860 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=186862.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 09:05:07,104 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.0536, 5.9070, 5.4512, 5.4304, 6.0294, 5.1809, 5.5522, 5.4918], device='cuda:0'), covar=tensor([0.1473, 0.0932, 0.0959, 0.2067, 0.0950, 0.2430, 0.1638, 0.0962], device='cuda:0'), in_proj_covar=tensor([0.0346, 0.0460, 0.0372, 0.0412, 0.0435, 0.0420, 0.0376, 0.0358], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002], device='cuda:0') 2023-05-16 09:05:12,723 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=186873.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 09:05:22,570 INFO [finetune.py:992] (0/2) Epoch 8, batch 1250, loss[loss=0.1565, simple_loss=0.2367, pruned_loss=0.03809, over 12357.00 frames. ], tot_loss[loss=0.1704, simple_loss=0.2587, pruned_loss=0.04106, over 2381044.01 frames. ], batch size: 30, lr: 4.34e-03, grad_scale: 32.0 2023-05-16 09:05:28,319 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=186895.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 09:05:33,384 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=186901.0, num_to_drop=1, layers_to_drop={3} 2023-05-16 09:05:40,361 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=186911.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 09:05:43,248 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=186915.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 09:05:47,163 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.81 vs. limit=2.0 2023-05-16 09:05:49,030 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=186923.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 09:05:58,875 INFO [finetune.py:992] (0/2) Epoch 8, batch 1300, loss[loss=0.1628, simple_loss=0.2513, pruned_loss=0.03716, over 12268.00 frames. ], tot_loss[loss=0.1709, simple_loss=0.2591, pruned_loss=0.0413, over 2382738.14 frames. ], batch size: 28, lr: 4.34e-03, grad_scale: 32.0 2023-05-16 09:06:02,402 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=186942.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 09:06:05,844 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.020e+02 3.030e+02 3.437e+02 3.972e+02 7.661e+02, threshold=6.874e+02, percent-clipped=2.0 2023-05-16 09:06:17,452 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=186963.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 09:06:23,943 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=186972.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 09:06:35,104 INFO [finetune.py:992] (0/2) Epoch 8, batch 1350, loss[loss=0.2047, simple_loss=0.2987, pruned_loss=0.05529, over 10639.00 frames. ], tot_loss[loss=0.171, simple_loss=0.2595, pruned_loss=0.04127, over 2381902.44 frames. ], batch size: 68, lr: 4.34e-03, grad_scale: 32.0 2023-05-16 09:07:02,201 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=187025.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 09:07:10,598 INFO [finetune.py:992] (0/2) Epoch 8, batch 1400, loss[loss=0.1831, simple_loss=0.2735, pruned_loss=0.04634, over 12101.00 frames. ], tot_loss[loss=0.1712, simple_loss=0.2596, pruned_loss=0.04138, over 2376173.36 frames. ], batch size: 39, lr: 4.34e-03, grad_scale: 32.0 2023-05-16 09:07:11,603 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=187038.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 09:07:17,858 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.995e+02 2.675e+02 2.989e+02 3.804e+02 7.426e+02, threshold=5.978e+02, percent-clipped=2.0 2023-05-16 09:07:31,423 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=187065.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 09:07:39,950 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=187077.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 09:07:46,662 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=187086.0, num_to_drop=1, layers_to_drop={3} 2023-05-16 09:07:47,184 INFO [finetune.py:992] (0/2) Epoch 8, batch 1450, loss[loss=0.1746, simple_loss=0.2789, pruned_loss=0.03516, over 12106.00 frames. ], tot_loss[loss=0.1707, simple_loss=0.2592, pruned_loss=0.0411, over 2370789.72 frames. ], batch size: 42, lr: 4.34e-03, grad_scale: 32.0 2023-05-16 09:07:55,790 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=187099.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 09:08:14,291 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=187125.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 09:08:19,223 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=187131.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 09:08:23,488 INFO [finetune.py:992] (0/2) Epoch 8, batch 1500, loss[loss=0.1528, simple_loss=0.2474, pruned_loss=0.02909, over 12280.00 frames. ], tot_loss[loss=0.1709, simple_loss=0.2593, pruned_loss=0.0413, over 2374473.79 frames. ], batch size: 33, lr: 4.34e-03, grad_scale: 16.0 2023-05-16 09:08:26,610 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.4599, 5.0263, 5.4186, 4.7657, 5.0796, 4.8676, 5.4780, 5.1248], device='cuda:0'), covar=tensor([0.0228, 0.0296, 0.0249, 0.0240, 0.0320, 0.0290, 0.0189, 0.0248], device='cuda:0'), in_proj_covar=tensor([0.0245, 0.0250, 0.0271, 0.0247, 0.0246, 0.0244, 0.0222, 0.0202], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 09:08:27,214 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.2339, 5.9917, 5.5302, 5.5302, 6.1060, 5.5269, 5.5252, 5.5579], device='cuda:0'), covar=tensor([0.1365, 0.1031, 0.1095, 0.1948, 0.0786, 0.1990, 0.2067, 0.1039], device='cuda:0'), in_proj_covar=tensor([0.0345, 0.0463, 0.0370, 0.0411, 0.0432, 0.0417, 0.0375, 0.0358], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002], device='cuda:0') 2023-05-16 09:08:31,284 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.046e+02 2.674e+02 3.261e+02 3.971e+02 7.100e+02, threshold=6.522e+02, percent-clipped=2.0 2023-05-16 09:08:33,548 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.9940, 6.0279, 5.7298, 5.2951, 5.1038, 5.8840, 5.5064, 5.3033], device='cuda:0'), covar=tensor([0.0817, 0.0860, 0.0728, 0.1506, 0.0744, 0.0729, 0.1584, 0.1010], device='cuda:0'), in_proj_covar=tensor([0.0588, 0.0525, 0.0497, 0.0605, 0.0398, 0.0680, 0.0738, 0.0551], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0004, 0.0004, 0.0003], device='cuda:0') 2023-05-16 09:08:39,980 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.6073, 3.6220, 3.2159, 3.1242, 2.9532, 2.7526, 3.6945, 2.2670], device='cuda:0'), covar=tensor([0.0323, 0.0120, 0.0227, 0.0194, 0.0296, 0.0331, 0.0105, 0.0436], device='cuda:0'), in_proj_covar=tensor([0.0182, 0.0152, 0.0150, 0.0175, 0.0191, 0.0188, 0.0157, 0.0191], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], device='cuda:0') 2023-05-16 09:08:49,127 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=187173.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 09:08:53,391 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=187179.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 09:08:58,975 INFO [finetune.py:992] (0/2) Epoch 8, batch 1550, loss[loss=0.1947, simple_loss=0.2894, pruned_loss=0.05001, over 12362.00 frames. ], tot_loss[loss=0.1721, simple_loss=0.2605, pruned_loss=0.04179, over 2369023.08 frames. ], batch size: 38, lr: 4.34e-03, grad_scale: 16.0 2023-05-16 09:09:05,462 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=187195.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 09:09:06,101 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=187196.0, num_to_drop=1, layers_to_drop={2} 2023-05-16 09:09:23,864 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=187221.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 09:09:25,306 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=187223.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 09:09:34,987 INFO [finetune.py:992] (0/2) Epoch 8, batch 1600, loss[loss=0.1624, simple_loss=0.2475, pruned_loss=0.0387, over 12249.00 frames. ], tot_loss[loss=0.1724, simple_loss=0.2609, pruned_loss=0.04197, over 2369593.35 frames. ], batch size: 32, lr: 4.34e-03, grad_scale: 16.0 2023-05-16 09:09:38,667 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=187242.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 09:09:39,273 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=187243.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 09:09:42,832 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.027e+02 2.917e+02 3.281e+02 3.882e+02 5.904e+02, threshold=6.562e+02, percent-clipped=0.0 2023-05-16 09:09:55,094 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.6521, 3.5168, 3.1791, 3.1363, 2.9359, 2.6619, 3.5413, 2.3096], device='cuda:0'), covar=tensor([0.0306, 0.0117, 0.0166, 0.0178, 0.0339, 0.0343, 0.0116, 0.0430], device='cuda:0'), in_proj_covar=tensor([0.0184, 0.0153, 0.0151, 0.0177, 0.0193, 0.0190, 0.0159, 0.0193], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 09:09:56,419 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=187267.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 09:09:59,267 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=187271.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 09:10:11,260 INFO [finetune.py:992] (0/2) Epoch 8, batch 1650, loss[loss=0.1496, simple_loss=0.2308, pruned_loss=0.03422, over 12288.00 frames. ], tot_loss[loss=0.1718, simple_loss=0.2602, pruned_loss=0.04172, over 2372248.00 frames. ], batch size: 28, lr: 4.33e-03, grad_scale: 16.0 2023-05-16 09:10:13,352 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=187290.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 09:10:43,353 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.8266, 4.1171, 3.5207, 4.2682, 3.8838, 2.7302, 3.6952, 2.7929], device='cuda:0'), covar=tensor([0.0873, 0.0869, 0.1743, 0.0600, 0.1335, 0.1683, 0.1118, 0.3406], device='cuda:0'), in_proj_covar=tensor([0.0295, 0.0367, 0.0350, 0.0274, 0.0355, 0.0265, 0.0331, 0.0359], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 09:10:46,501 INFO [finetune.py:992] (0/2) Epoch 8, batch 1700, loss[loss=0.1639, simple_loss=0.2451, pruned_loss=0.04139, over 12178.00 frames. ], tot_loss[loss=0.1722, simple_loss=0.2606, pruned_loss=0.04188, over 2372760.50 frames. ], batch size: 31, lr: 4.33e-03, grad_scale: 16.0 2023-05-16 09:10:55,034 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.948e+02 2.737e+02 3.023e+02 3.663e+02 6.780e+02, threshold=6.046e+02, percent-clipped=1.0 2023-05-16 09:11:04,493 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.0161, 4.5505, 4.7231, 4.6850, 4.6632, 4.8462, 4.6599, 2.7987], device='cuda:0'), covar=tensor([0.0092, 0.0079, 0.0096, 0.0076, 0.0046, 0.0092, 0.0099, 0.0719], device='cuda:0'), in_proj_covar=tensor([0.0063, 0.0075, 0.0078, 0.0070, 0.0057, 0.0088, 0.0076, 0.0093], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-16 09:11:07,208 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=187365.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 09:11:09,616 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.94 vs. limit=2.0 2023-05-16 09:11:12,189 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=187372.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 09:11:18,559 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=187381.0, num_to_drop=1, layers_to_drop={0} 2023-05-16 09:11:22,785 INFO [finetune.py:992] (0/2) Epoch 8, batch 1750, loss[loss=0.1898, simple_loss=0.284, pruned_loss=0.04785, over 12303.00 frames. ], tot_loss[loss=0.1721, simple_loss=0.2605, pruned_loss=0.04182, over 2379592.42 frames. ], batch size: 34, lr: 4.33e-03, grad_scale: 16.0 2023-05-16 09:11:27,930 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=187394.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 09:11:41,694 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=187413.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 09:11:57,078 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=187433.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 09:11:59,656 INFO [finetune.py:992] (0/2) Epoch 8, batch 1800, loss[loss=0.1905, simple_loss=0.2757, pruned_loss=0.05259, over 12077.00 frames. ], tot_loss[loss=0.1718, simple_loss=0.26, pruned_loss=0.04179, over 2384217.50 frames. ], batch size: 32, lr: 4.33e-03, grad_scale: 16.0 2023-05-16 09:12:07,494 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.665e+02 2.838e+02 3.392e+02 3.885e+02 9.077e+02, threshold=6.783e+02, percent-clipped=5.0 2023-05-16 09:12:22,831 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.40 vs. limit=2.0 2023-05-16 09:12:27,505 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.1160, 5.0751, 4.9369, 4.9483, 4.6201, 5.0833, 5.0213, 5.3318], device='cuda:0'), covar=tensor([0.0206, 0.0137, 0.0184, 0.0330, 0.0750, 0.0291, 0.0153, 0.0162], device='cuda:0'), in_proj_covar=tensor([0.0184, 0.0186, 0.0182, 0.0233, 0.0229, 0.0204, 0.0166, 0.0212], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0005, 0.0004, 0.0004, 0.0003, 0.0005], device='cuda:0') 2023-05-16 09:12:35,084 INFO [finetune.py:992] (0/2) Epoch 8, batch 1850, loss[loss=0.1737, simple_loss=0.2606, pruned_loss=0.04341, over 12185.00 frames. ], tot_loss[loss=0.1718, simple_loss=0.2601, pruned_loss=0.04176, over 2390311.24 frames. ], batch size: 35, lr: 4.33e-03, grad_scale: 16.0 2023-05-16 09:12:41,533 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=187496.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 09:12:53,477 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=187512.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 09:12:58,433 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=187519.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 09:13:01,329 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=187523.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 09:13:11,124 INFO [finetune.py:992] (0/2) Epoch 8, batch 1900, loss[loss=0.1575, simple_loss=0.25, pruned_loss=0.03251, over 12139.00 frames. ], tot_loss[loss=0.1724, simple_loss=0.2604, pruned_loss=0.04218, over 2382873.45 frames. ], batch size: 30, lr: 4.33e-03, grad_scale: 16.0 2023-05-16 09:13:16,091 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=187544.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 09:13:18,779 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.095e+02 2.883e+02 3.371e+02 4.186e+02 8.354e+02, threshold=6.742e+02, percent-clipped=2.0 2023-05-16 09:13:33,045 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=187567.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 09:13:37,438 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=187573.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 09:13:42,345 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=187580.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 09:13:42,589 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=3.45 vs. limit=5.0 2023-05-16 09:13:45,146 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=187584.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 09:13:47,059 INFO [finetune.py:992] (0/2) Epoch 8, batch 1950, loss[loss=0.1765, simple_loss=0.2724, pruned_loss=0.04029, over 12257.00 frames. ], tot_loss[loss=0.1734, simple_loss=0.2619, pruned_loss=0.04245, over 2386110.96 frames. ], batch size: 37, lr: 4.33e-03, grad_scale: 16.0 2023-05-16 09:14:07,069 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=187615.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 09:14:22,423 INFO [finetune.py:992] (0/2) Epoch 8, batch 2000, loss[loss=0.1558, simple_loss=0.2447, pruned_loss=0.03347, over 12182.00 frames. ], tot_loss[loss=0.1726, simple_loss=0.2606, pruned_loss=0.04231, over 2383805.72 frames. ], batch size: 31, lr: 4.33e-03, grad_scale: 16.0 2023-05-16 09:14:30,412 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.4289, 2.5349, 3.0643, 4.2933, 2.2689, 4.3359, 4.4496, 4.5973], device='cuda:0'), covar=tensor([0.0118, 0.1161, 0.0540, 0.0146, 0.1280, 0.0209, 0.0129, 0.0075], device='cuda:0'), in_proj_covar=tensor([0.0116, 0.0201, 0.0179, 0.0111, 0.0187, 0.0173, 0.0171, 0.0116], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 09:14:30,920 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.725e+02 2.688e+02 3.136e+02 3.958e+02 7.783e+02, threshold=6.272e+02, percent-clipped=1.0 2023-05-16 09:14:54,328 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=187681.0, num_to_drop=1, layers_to_drop={0} 2023-05-16 09:14:58,287 INFO [finetune.py:992] (0/2) Epoch 8, batch 2050, loss[loss=0.1863, simple_loss=0.2805, pruned_loss=0.04609, over 12343.00 frames. ], tot_loss[loss=0.172, simple_loss=0.2599, pruned_loss=0.042, over 2389651.87 frames. ], batch size: 35, lr: 4.33e-03, grad_scale: 16.0 2023-05-16 09:15:03,515 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=187694.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 09:15:28,595 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=187728.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 09:15:29,280 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=187729.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 09:15:34,853 INFO [finetune.py:992] (0/2) Epoch 8, batch 2100, loss[loss=0.1653, simple_loss=0.2383, pruned_loss=0.04616, over 12298.00 frames. ], tot_loss[loss=0.1713, simple_loss=0.2593, pruned_loss=0.04168, over 2396463.53 frames. ], batch size: 28, lr: 4.33e-03, grad_scale: 16.0 2023-05-16 09:15:38,289 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=187742.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 09:15:40,827 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=3.34 vs. limit=5.0 2023-05-16 09:15:42,442 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.059e+02 2.818e+02 3.223e+02 3.937e+02 8.971e+02, threshold=6.446e+02, percent-clipped=3.0 2023-05-16 09:16:09,728 INFO [finetune.py:992] (0/2) Epoch 8, batch 2150, loss[loss=0.1791, simple_loss=0.2593, pruned_loss=0.04945, over 11858.00 frames. ], tot_loss[loss=0.1717, simple_loss=0.2599, pruned_loss=0.0418, over 2392776.93 frames. ], batch size: 26, lr: 4.33e-03, grad_scale: 16.0 2023-05-16 09:16:13,377 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.2183, 6.0015, 5.5359, 5.5610, 6.0948, 5.4367, 5.6003, 5.5976], device='cuda:0'), covar=tensor([0.1518, 0.1003, 0.1232, 0.1989, 0.0906, 0.2302, 0.1856, 0.1230], device='cuda:0'), in_proj_covar=tensor([0.0346, 0.0465, 0.0374, 0.0413, 0.0437, 0.0418, 0.0379, 0.0363], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 09:16:46,218 INFO [finetune.py:992] (0/2) Epoch 8, batch 2200, loss[loss=0.1917, simple_loss=0.2858, pruned_loss=0.04883, over 12138.00 frames. ], tot_loss[loss=0.1711, simple_loss=0.2592, pruned_loss=0.04151, over 2389269.15 frames. ], batch size: 39, lr: 4.33e-03, grad_scale: 16.0 2023-05-16 09:16:53,917 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.976e+02 2.731e+02 3.100e+02 3.737e+02 9.100e+02, threshold=6.201e+02, percent-clipped=2.0 2023-05-16 09:17:01,508 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=5.06 vs. limit=5.0 2023-05-16 09:17:08,795 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=187868.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 09:17:13,591 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=187875.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 09:17:16,476 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=187879.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 09:17:22,269 INFO [finetune.py:992] (0/2) Epoch 8, batch 2250, loss[loss=0.1675, simple_loss=0.2555, pruned_loss=0.03981, over 11799.00 frames. ], tot_loss[loss=0.1712, simple_loss=0.2593, pruned_loss=0.04157, over 2389799.13 frames. ], batch size: 44, lr: 4.33e-03, grad_scale: 16.0 2023-05-16 09:17:57,924 INFO [finetune.py:992] (0/2) Epoch 8, batch 2300, loss[loss=0.1796, simple_loss=0.2525, pruned_loss=0.05335, over 12182.00 frames. ], tot_loss[loss=0.1714, simple_loss=0.2596, pruned_loss=0.0416, over 2382004.42 frames. ], batch size: 29, lr: 4.33e-03, grad_scale: 16.0 2023-05-16 09:18:01,460 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.3090, 5.1899, 5.3194, 5.3368, 5.0115, 4.9507, 4.7517, 5.2866], device='cuda:0'), covar=tensor([0.0648, 0.0513, 0.0673, 0.0519, 0.1712, 0.1255, 0.0532, 0.0934], device='cuda:0'), in_proj_covar=tensor([0.0498, 0.0652, 0.0560, 0.0589, 0.0793, 0.0688, 0.0512, 0.0455], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0005, 0.0005, 0.0003, 0.0003], device='cuda:0') 2023-05-16 09:18:06,256 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.959e+02 2.922e+02 3.527e+02 4.326e+02 1.992e+03, threshold=7.054e+02, percent-clipped=7.0 2023-05-16 09:18:33,745 INFO [finetune.py:992] (0/2) Epoch 8, batch 2350, loss[loss=0.1441, simple_loss=0.2339, pruned_loss=0.02716, over 12385.00 frames. ], tot_loss[loss=0.1715, simple_loss=0.2595, pruned_loss=0.04173, over 2381252.76 frames. ], batch size: 32, lr: 4.33e-03, grad_scale: 16.0 2023-05-16 09:18:43,205 INFO [checkpoint.py:75] (0/2) Saving checkpoint to pruned_transducer_stateless7/exp_giga_finetune/checkpoint-88000.pt 2023-05-16 09:19:03,603 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-05-16 09:19:06,845 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=188028.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 09:19:12,932 INFO [finetune.py:992] (0/2) Epoch 8, batch 2400, loss[loss=0.192, simple_loss=0.2756, pruned_loss=0.05416, over 8473.00 frames. ], tot_loss[loss=0.1722, simple_loss=0.2602, pruned_loss=0.04213, over 2369429.08 frames. ], batch size: 98, lr: 4.33e-03, grad_scale: 16.0 2023-05-16 09:19:20,875 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.821e+02 2.762e+02 3.244e+02 3.839e+02 6.809e+02, threshold=6.487e+02, percent-clipped=0.0 2023-05-16 09:19:40,451 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=188076.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 09:19:48,619 INFO [finetune.py:992] (0/2) Epoch 8, batch 2450, loss[loss=0.2529, simple_loss=0.3231, pruned_loss=0.09132, over 8233.00 frames. ], tot_loss[loss=0.1721, simple_loss=0.2599, pruned_loss=0.04215, over 2371034.76 frames. ], batch size: 99, lr: 4.33e-03, grad_scale: 16.0 2023-05-16 09:20:24,118 INFO [finetune.py:992] (0/2) Epoch 8, batch 2500, loss[loss=0.1657, simple_loss=0.2619, pruned_loss=0.03476, over 12140.00 frames. ], tot_loss[loss=0.1727, simple_loss=0.2608, pruned_loss=0.0423, over 2371286.42 frames. ], batch size: 36, lr: 4.33e-03, grad_scale: 16.0 2023-05-16 09:20:32,508 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.088e+02 2.770e+02 3.184e+02 4.099e+02 5.447e+02, threshold=6.368e+02, percent-clipped=0.0 2023-05-16 09:20:39,030 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.1561, 4.8007, 4.8294, 4.9898, 4.7101, 4.9385, 4.8696, 2.7009], device='cuda:0'), covar=tensor([0.0088, 0.0059, 0.0086, 0.0058, 0.0044, 0.0086, 0.0076, 0.0756], device='cuda:0'), in_proj_covar=tensor([0.0064, 0.0075, 0.0078, 0.0070, 0.0057, 0.0088, 0.0076, 0.0094], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-16 09:20:46,813 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=188168.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 09:20:51,888 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=188175.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 09:20:54,747 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=188179.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 09:21:00,285 INFO [finetune.py:992] (0/2) Epoch 8, batch 2550, loss[loss=0.175, simple_loss=0.2721, pruned_loss=0.03893, over 12122.00 frames. ], tot_loss[loss=0.173, simple_loss=0.2614, pruned_loss=0.04235, over 2372518.30 frames. ], batch size: 39, lr: 4.32e-03, grad_scale: 16.0 2023-05-16 09:21:21,641 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=188216.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 09:21:26,606 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=188223.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 09:21:30,192 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=188227.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 09:21:37,123 INFO [finetune.py:992] (0/2) Epoch 8, batch 2600, loss[loss=0.2093, simple_loss=0.3002, pruned_loss=0.05919, over 12141.00 frames. ], tot_loss[loss=0.1719, simple_loss=0.2602, pruned_loss=0.04176, over 2380488.16 frames. ], batch size: 34, lr: 4.32e-03, grad_scale: 16.0 2023-05-16 09:21:39,432 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.5714, 2.4894, 3.5084, 4.5028, 4.0053, 4.4845, 3.8314, 3.1761], device='cuda:0'), covar=tensor([0.0032, 0.0385, 0.0144, 0.0033, 0.0112, 0.0056, 0.0120, 0.0329], device='cuda:0'), in_proj_covar=tensor([0.0084, 0.0120, 0.0102, 0.0074, 0.0099, 0.0114, 0.0093, 0.0135], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 09:21:44,686 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.096e+02 2.791e+02 3.137e+02 3.810e+02 9.850e+02, threshold=6.275e+02, percent-clipped=2.0 2023-05-16 09:22:09,857 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=188283.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 09:22:12,632 INFO [finetune.py:992] (0/2) Epoch 8, batch 2650, loss[loss=0.1547, simple_loss=0.2492, pruned_loss=0.03008, over 12297.00 frames. ], tot_loss[loss=0.1718, simple_loss=0.2602, pruned_loss=0.04174, over 2384442.36 frames. ], batch size: 34, lr: 4.32e-03, grad_scale: 16.0 2023-05-16 09:22:32,186 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.1549, 2.6700, 3.7984, 3.3443, 3.6645, 3.3290, 2.7482, 3.7320], device='cuda:0'), covar=tensor([0.0121, 0.0326, 0.0125, 0.0177, 0.0135, 0.0160, 0.0334, 0.0118], device='cuda:0'), in_proj_covar=tensor([0.0165, 0.0198, 0.0177, 0.0176, 0.0203, 0.0153, 0.0190, 0.0174], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 09:22:42,620 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.2655, 3.3770, 3.4319, 3.9979, 2.6513, 3.4630, 2.3179, 3.5442], device='cuda:0'), covar=tensor([0.1467, 0.1026, 0.1240, 0.0840, 0.1289, 0.0776, 0.2039, 0.1153], device='cuda:0'), in_proj_covar=tensor([0.0223, 0.0260, 0.0292, 0.0347, 0.0234, 0.0236, 0.0255, 0.0372], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 09:22:48,945 INFO [finetune.py:992] (0/2) Epoch 8, batch 2700, loss[loss=0.1655, simple_loss=0.2595, pruned_loss=0.03581, over 12293.00 frames. ], tot_loss[loss=0.1711, simple_loss=0.2593, pruned_loss=0.04149, over 2382432.74 frames. ], batch size: 34, lr: 4.32e-03, grad_scale: 16.0 2023-05-16 09:22:54,250 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=188344.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 09:22:56,969 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.644e+02 2.755e+02 3.326e+02 3.869e+02 7.495e+02, threshold=6.653e+02, percent-clipped=1.0 2023-05-16 09:23:02,985 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.8637, 2.9725, 4.6136, 4.8330, 2.8814, 2.7427, 2.9570, 2.2772], device='cuda:0'), covar=tensor([0.1363, 0.2961, 0.0450, 0.0372, 0.1251, 0.2113, 0.2602, 0.3936], device='cuda:0'), in_proj_covar=tensor([0.0281, 0.0376, 0.0268, 0.0294, 0.0259, 0.0289, 0.0364, 0.0363], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-16 09:23:24,729 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.9457, 2.3503, 3.5579, 3.0104, 3.3086, 3.0871, 2.4820, 3.3946], device='cuda:0'), covar=tensor([0.0140, 0.0369, 0.0147, 0.0224, 0.0157, 0.0165, 0.0362, 0.0151], device='cuda:0'), in_proj_covar=tensor([0.0165, 0.0198, 0.0177, 0.0176, 0.0203, 0.0153, 0.0190, 0.0174], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 09:23:25,228 INFO [finetune.py:992] (0/2) Epoch 8, batch 2750, loss[loss=0.1815, simple_loss=0.2729, pruned_loss=0.04501, over 12128.00 frames. ], tot_loss[loss=0.1716, simple_loss=0.26, pruned_loss=0.04156, over 2382902.52 frames. ], batch size: 39, lr: 4.32e-03, grad_scale: 16.0 2023-05-16 09:23:35,037 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.6505, 2.7758, 3.5276, 4.5412, 4.0552, 4.5854, 3.8846, 3.0594], device='cuda:0'), covar=tensor([0.0030, 0.0340, 0.0152, 0.0038, 0.0099, 0.0060, 0.0111, 0.0353], device='cuda:0'), in_proj_covar=tensor([0.0085, 0.0121, 0.0103, 0.0075, 0.0100, 0.0115, 0.0093, 0.0136], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 09:23:48,712 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.7546, 2.9470, 4.7412, 4.9889, 2.9499, 2.6910, 3.0352, 2.2511], device='cuda:0'), covar=tensor([0.1442, 0.3075, 0.0387, 0.0332, 0.1253, 0.2150, 0.2615, 0.3863], device='cuda:0'), in_proj_covar=tensor([0.0281, 0.0375, 0.0267, 0.0293, 0.0259, 0.0289, 0.0363, 0.0362], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-16 09:24:01,865 INFO [finetune.py:992] (0/2) Epoch 8, batch 2800, loss[loss=0.1725, simple_loss=0.2648, pruned_loss=0.04008, over 11426.00 frames. ], tot_loss[loss=0.1718, simple_loss=0.2602, pruned_loss=0.04169, over 2373772.88 frames. ], batch size: 48, lr: 4.32e-03, grad_scale: 16.0 2023-05-16 09:24:04,142 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.1517, 5.9605, 5.5038, 5.4945, 6.0545, 5.3927, 5.6206, 5.6093], device='cuda:0'), covar=tensor([0.1423, 0.1009, 0.0915, 0.1871, 0.0980, 0.2053, 0.1701, 0.1185], device='cuda:0'), in_proj_covar=tensor([0.0353, 0.0472, 0.0377, 0.0420, 0.0445, 0.0427, 0.0386, 0.0367], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 09:24:09,719 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.770e+02 2.626e+02 3.087e+02 3.593e+02 6.799e+02, threshold=6.173e+02, percent-clipped=1.0 2023-05-16 09:24:12,931 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2023-05-16 09:24:33,563 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=188481.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 09:24:37,600 INFO [finetune.py:992] (0/2) Epoch 8, batch 2850, loss[loss=0.1725, simple_loss=0.2502, pruned_loss=0.04737, over 12304.00 frames. ], tot_loss[loss=0.1719, simple_loss=0.2603, pruned_loss=0.04177, over 2373690.00 frames. ], batch size: 33, lr: 4.32e-03, grad_scale: 16.0 2023-05-16 09:24:52,693 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.3257, 4.8666, 5.3179, 4.6225, 4.9412, 4.6954, 5.3544, 5.0406], device='cuda:0'), covar=tensor([0.0241, 0.0359, 0.0248, 0.0261, 0.0361, 0.0302, 0.0195, 0.0294], device='cuda:0'), in_proj_covar=tensor([0.0246, 0.0249, 0.0275, 0.0249, 0.0248, 0.0247, 0.0224, 0.0206], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 09:25:13,834 INFO [finetune.py:992] (0/2) Epoch 8, batch 2900, loss[loss=0.1642, simple_loss=0.2568, pruned_loss=0.03573, over 12360.00 frames. ], tot_loss[loss=0.1705, simple_loss=0.2586, pruned_loss=0.04114, over 2368913.58 frames. ], batch size: 35, lr: 4.32e-03, grad_scale: 16.0 2023-05-16 09:25:17,527 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=188542.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 09:25:21,489 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.114e+02 2.935e+02 3.309e+02 3.834e+02 6.179e+02, threshold=6.618e+02, percent-clipped=1.0 2023-05-16 09:25:35,460 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2023-05-16 09:25:50,035 INFO [finetune.py:992] (0/2) Epoch 8, batch 2950, loss[loss=0.2023, simple_loss=0.2864, pruned_loss=0.05911, over 12150.00 frames. ], tot_loss[loss=0.1706, simple_loss=0.2585, pruned_loss=0.0413, over 2376154.65 frames. ], batch size: 36, lr: 4.32e-03, grad_scale: 16.0 2023-05-16 09:25:53,086 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.3864, 2.8910, 2.8086, 2.8126, 2.5387, 2.4026, 2.8673, 2.1068], device='cuda:0'), covar=tensor([0.0328, 0.0163, 0.0176, 0.0164, 0.0350, 0.0302, 0.0166, 0.0388], device='cuda:0'), in_proj_covar=tensor([0.0194, 0.0162, 0.0158, 0.0184, 0.0204, 0.0199, 0.0167, 0.0202], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 09:26:07,688 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.2339, 5.1893, 5.0221, 5.1061, 4.6838, 5.2363, 5.1432, 5.5246], device='cuda:0'), covar=tensor([0.0220, 0.0142, 0.0189, 0.0307, 0.0772, 0.0260, 0.0154, 0.0126], device='cuda:0'), in_proj_covar=tensor([0.0190, 0.0193, 0.0188, 0.0241, 0.0238, 0.0212, 0.0170, 0.0218], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0005, 0.0004, 0.0005, 0.0003, 0.0005], device='cuda:0') 2023-05-16 09:26:16,139 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.1209, 2.3753, 3.6803, 3.0750, 3.4548, 3.2232, 2.5227, 3.4565], device='cuda:0'), covar=tensor([0.0129, 0.0358, 0.0134, 0.0250, 0.0163, 0.0152, 0.0362, 0.0142], device='cuda:0'), in_proj_covar=tensor([0.0166, 0.0198, 0.0178, 0.0178, 0.0204, 0.0153, 0.0191, 0.0175], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 09:26:24,051 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.4911, 4.2364, 4.1315, 4.5211, 2.8961, 3.9638, 2.6605, 4.1900], device='cuda:0'), covar=tensor([0.1394, 0.0584, 0.0949, 0.0592, 0.1190, 0.0611, 0.1735, 0.1255], device='cuda:0'), in_proj_covar=tensor([0.0222, 0.0259, 0.0290, 0.0345, 0.0234, 0.0235, 0.0254, 0.0369], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 09:26:25,893 INFO [finetune.py:992] (0/2) Epoch 8, batch 3000, loss[loss=0.1949, simple_loss=0.2812, pruned_loss=0.0543, over 12356.00 frames. ], tot_loss[loss=0.1702, simple_loss=0.2581, pruned_loss=0.04118, over 2383279.16 frames. ], batch size: 38, lr: 4.32e-03, grad_scale: 16.0 2023-05-16 09:26:25,894 INFO [finetune.py:1017] (0/2) Computing validation loss 2023-05-16 09:26:44,282 INFO [finetune.py:1026] (0/2) Epoch 8, validation: loss=0.3187, simple_loss=0.3968, pruned_loss=0.1203, over 1020973.00 frames. 2023-05-16 09:26:44,282 INFO [finetune.py:1027] (0/2) Maximum memory allocated so far is 12761MB 2023-05-16 09:26:45,897 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=188639.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 09:26:48,230 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.8415, 2.8812, 4.6379, 4.9541, 2.9368, 2.7146, 3.0086, 2.1804], device='cuda:0'), covar=tensor([0.1384, 0.2839, 0.0408, 0.0338, 0.1193, 0.2166, 0.2610, 0.3892], device='cuda:0'), in_proj_covar=tensor([0.0282, 0.0377, 0.0268, 0.0294, 0.0260, 0.0291, 0.0364, 0.0364], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-16 09:26:52,300 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.755e+02 2.728e+02 3.186e+02 3.763e+02 6.061e+02, threshold=6.372e+02, percent-clipped=0.0 2023-05-16 09:26:54,113 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.34 vs. limit=2.0 2023-05-16 09:27:20,032 INFO [finetune.py:992] (0/2) Epoch 8, batch 3050, loss[loss=0.1615, simple_loss=0.2459, pruned_loss=0.03859, over 12194.00 frames. ], tot_loss[loss=0.1704, simple_loss=0.2584, pruned_loss=0.04122, over 2375286.66 frames. ], batch size: 31, lr: 4.32e-03, grad_scale: 16.0 2023-05-16 09:27:56,173 INFO [finetune.py:992] (0/2) Epoch 8, batch 3100, loss[loss=0.168, simple_loss=0.2578, pruned_loss=0.03908, over 12089.00 frames. ], tot_loss[loss=0.1711, simple_loss=0.2593, pruned_loss=0.04143, over 2377976.10 frames. ], batch size: 32, lr: 4.32e-03, grad_scale: 16.0 2023-05-16 09:28:03,966 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.878e+02 2.640e+02 3.122e+02 3.982e+02 7.416e+02, threshold=6.244e+02, percent-clipped=3.0 2023-05-16 09:28:31,443 INFO [finetune.py:992] (0/2) Epoch 8, batch 3150, loss[loss=0.1525, simple_loss=0.2427, pruned_loss=0.0311, over 12278.00 frames. ], tot_loss[loss=0.1711, simple_loss=0.2593, pruned_loss=0.04144, over 2374948.65 frames. ], batch size: 33, lr: 4.32e-03, grad_scale: 16.0 2023-05-16 09:28:31,719 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.6555, 2.5579, 4.5042, 4.9607, 3.3289, 2.6892, 3.0190, 1.9307], device='cuda:0'), covar=tensor([0.1555, 0.3589, 0.0454, 0.0310, 0.0980, 0.2258, 0.2760, 0.4971], device='cuda:0'), in_proj_covar=tensor([0.0282, 0.0376, 0.0267, 0.0293, 0.0259, 0.0290, 0.0362, 0.0362], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-16 09:29:08,910 INFO [finetune.py:992] (0/2) Epoch 8, batch 3200, loss[loss=0.1603, simple_loss=0.2574, pruned_loss=0.03161, over 12179.00 frames. ], tot_loss[loss=0.1709, simple_loss=0.2591, pruned_loss=0.04141, over 2367592.58 frames. ], batch size: 35, lr: 4.32e-03, grad_scale: 16.0 2023-05-16 09:29:08,992 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=188837.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 09:29:16,724 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.984e+02 2.669e+02 3.069e+02 3.720e+02 5.873e+02, threshold=6.139e+02, percent-clipped=0.0 2023-05-16 09:29:33,869 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.9176, 3.5540, 5.2762, 2.7395, 2.9900, 3.9359, 3.4842, 3.9908], device='cuda:0'), covar=tensor([0.0402, 0.1037, 0.0277, 0.1204, 0.1774, 0.1458, 0.1192, 0.1153], device='cuda:0'), in_proj_covar=tensor([0.0234, 0.0231, 0.0238, 0.0180, 0.0235, 0.0283, 0.0220, 0.0253], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:0') 2023-05-16 09:29:44,263 INFO [finetune.py:992] (0/2) Epoch 8, batch 3250, loss[loss=0.1623, simple_loss=0.2518, pruned_loss=0.03637, over 12261.00 frames. ], tot_loss[loss=0.1713, simple_loss=0.2592, pruned_loss=0.04171, over 2362375.41 frames. ], batch size: 32, lr: 4.32e-03, grad_scale: 16.0 2023-05-16 09:30:10,314 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.3089, 5.0465, 5.4151, 5.2704, 4.4444, 4.6506, 4.8250, 5.0905], device='cuda:0'), covar=tensor([0.1101, 0.1235, 0.0836, 0.1031, 0.3498, 0.2096, 0.0644, 0.1665], device='cuda:0'), in_proj_covar=tensor([0.0510, 0.0663, 0.0573, 0.0604, 0.0820, 0.0704, 0.0528, 0.0467], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0005, 0.0004, 0.0004, 0.0005, 0.0005, 0.0004, 0.0003], device='cuda:0') 2023-05-16 09:30:13,750 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=188929.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 09:30:15,841 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=188931.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 09:30:19,946 INFO [finetune.py:992] (0/2) Epoch 8, batch 3300, loss[loss=0.1554, simple_loss=0.2458, pruned_loss=0.03254, over 12186.00 frames. ], tot_loss[loss=0.1713, simple_loss=0.2592, pruned_loss=0.04166, over 2367401.95 frames. ], batch size: 31, lr: 4.32e-03, grad_scale: 16.0 2023-05-16 09:30:21,554 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=188939.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 09:30:27,769 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.954e+02 2.793e+02 3.385e+02 3.964e+02 1.560e+03, threshold=6.770e+02, percent-clipped=7.0 2023-05-16 09:30:28,213 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2023-05-16 09:30:40,635 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.5019, 5.2936, 5.4208, 5.4394, 5.0341, 5.1039, 4.8831, 5.3449], device='cuda:0'), covar=tensor([0.0673, 0.0610, 0.0761, 0.0558, 0.1669, 0.1226, 0.0562, 0.1045], device='cuda:0'), in_proj_covar=tensor([0.0509, 0.0664, 0.0573, 0.0603, 0.0819, 0.0703, 0.0527, 0.0467], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0005, 0.0004, 0.0004, 0.0005, 0.0005, 0.0004, 0.0003], device='cuda:0') 2023-05-16 09:30:44,889 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.69 vs. limit=2.0 2023-05-16 09:30:54,632 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.2183, 5.0664, 5.0198, 4.9338, 4.6380, 5.1106, 5.2011, 5.3672], device='cuda:0'), covar=tensor([0.0165, 0.0143, 0.0147, 0.0329, 0.0696, 0.0269, 0.0115, 0.0141], device='cuda:0'), in_proj_covar=tensor([0.0188, 0.0191, 0.0186, 0.0240, 0.0236, 0.0210, 0.0169, 0.0217], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0005, 0.0004, 0.0005, 0.0003, 0.0005], device='cuda:0') 2023-05-16 09:30:55,906 INFO [finetune.py:992] (0/2) Epoch 8, batch 3350, loss[loss=0.1632, simple_loss=0.2511, pruned_loss=0.03767, over 12155.00 frames. ], tot_loss[loss=0.1719, simple_loss=0.2597, pruned_loss=0.042, over 2370958.51 frames. ], batch size: 36, lr: 4.32e-03, grad_scale: 16.0 2023-05-16 09:30:55,978 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=188987.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 09:30:58,342 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=188990.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 09:30:59,807 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=188992.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 09:31:26,906 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=189030.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 09:31:31,704 INFO [finetune.py:992] (0/2) Epoch 8, batch 3400, loss[loss=0.1645, simple_loss=0.2542, pruned_loss=0.03743, over 12108.00 frames. ], tot_loss[loss=0.1722, simple_loss=0.26, pruned_loss=0.04216, over 2367967.83 frames. ], batch size: 33, lr: 4.32e-03, grad_scale: 16.0 2023-05-16 09:31:39,752 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.971e+02 2.760e+02 3.345e+02 4.036e+02 1.056e+03, threshold=6.689e+02, percent-clipped=2.0 2023-05-16 09:31:51,494 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=189064.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 09:32:02,767 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=189079.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 09:32:08,140 INFO [finetune.py:992] (0/2) Epoch 8, batch 3450, loss[loss=0.163, simple_loss=0.2518, pruned_loss=0.03707, over 12289.00 frames. ], tot_loss[loss=0.172, simple_loss=0.2599, pruned_loss=0.04205, over 2371476.24 frames. ], batch size: 33, lr: 4.32e-03, grad_scale: 16.0 2023-05-16 09:32:11,344 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=189091.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 09:32:21,301 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.3965, 3.1668, 3.0430, 3.0338, 2.7600, 2.6031, 3.2961, 2.0664], device='cuda:0'), covar=tensor([0.0362, 0.0182, 0.0189, 0.0202, 0.0370, 0.0355, 0.0138, 0.0490], device='cuda:0'), in_proj_covar=tensor([0.0190, 0.0160, 0.0156, 0.0181, 0.0201, 0.0196, 0.0164, 0.0200], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 09:32:36,045 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=189125.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 09:32:44,435 INFO [finetune.py:992] (0/2) Epoch 8, batch 3500, loss[loss=0.1729, simple_loss=0.2609, pruned_loss=0.04249, over 12284.00 frames. ], tot_loss[loss=0.1715, simple_loss=0.2593, pruned_loss=0.04188, over 2375157.28 frames. ], batch size: 33, lr: 4.31e-03, grad_scale: 32.0 2023-05-16 09:32:44,536 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=189137.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 09:32:46,035 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.5857, 3.6962, 3.2946, 3.3133, 3.0552, 2.8772, 3.7537, 2.2316], device='cuda:0'), covar=tensor([0.0363, 0.0123, 0.0201, 0.0170, 0.0373, 0.0328, 0.0111, 0.0501], device='cuda:0'), in_proj_covar=tensor([0.0189, 0.0159, 0.0156, 0.0180, 0.0200, 0.0194, 0.0163, 0.0199], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 09:32:46,721 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=189140.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 09:32:52,137 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.832e+02 2.761e+02 3.216e+02 3.789e+02 1.190e+03, threshold=6.432e+02, percent-clipped=2.0 2023-05-16 09:32:55,235 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.6314, 3.0506, 5.0133, 2.4585, 2.7565, 3.6873, 3.1299, 3.6461], device='cuda:0'), covar=tensor([0.0421, 0.1252, 0.0254, 0.1366, 0.1968, 0.1523, 0.1370, 0.1279], device='cuda:0'), in_proj_covar=tensor([0.0233, 0.0230, 0.0236, 0.0179, 0.0233, 0.0282, 0.0219, 0.0252], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:0') 2023-05-16 09:33:18,285 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=189185.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 09:33:19,655 INFO [finetune.py:992] (0/2) Epoch 8, batch 3550, loss[loss=0.1822, simple_loss=0.2797, pruned_loss=0.04235, over 11583.00 frames. ], tot_loss[loss=0.1709, simple_loss=0.2586, pruned_loss=0.04162, over 2374945.21 frames. ], batch size: 48, lr: 4.31e-03, grad_scale: 32.0 2023-05-16 09:33:23,527 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.0980, 2.1126, 2.7121, 3.1275, 2.1909, 3.2478, 3.1063, 3.3088], device='cuda:0'), covar=tensor([0.0164, 0.1089, 0.0444, 0.0158, 0.1106, 0.0301, 0.0295, 0.0116], device='cuda:0'), in_proj_covar=tensor([0.0119, 0.0199, 0.0178, 0.0111, 0.0187, 0.0173, 0.0169, 0.0116], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 09:33:56,126 INFO [finetune.py:992] (0/2) Epoch 8, batch 3600, loss[loss=0.1858, simple_loss=0.2578, pruned_loss=0.05689, over 8685.00 frames. ], tot_loss[loss=0.1703, simple_loss=0.2583, pruned_loss=0.04117, over 2370338.47 frames. ], batch size: 98, lr: 4.31e-03, grad_scale: 32.0 2023-05-16 09:34:03,901 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.891e+02 2.606e+02 3.237e+02 3.912e+02 7.810e+02, threshold=6.473e+02, percent-clipped=1.0 2023-05-16 09:34:04,091 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=189248.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 09:34:30,955 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=189285.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 09:34:32,255 INFO [finetune.py:992] (0/2) Epoch 8, batch 3650, loss[loss=0.176, simple_loss=0.2708, pruned_loss=0.04062, over 11786.00 frames. ], tot_loss[loss=0.1703, simple_loss=0.2581, pruned_loss=0.04124, over 2373161.80 frames. ], batch size: 44, lr: 4.31e-03, grad_scale: 32.0 2023-05-16 09:34:32,333 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=189287.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 09:34:48,146 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=189309.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 09:35:07,879 INFO [finetune.py:992] (0/2) Epoch 8, batch 3700, loss[loss=0.1672, simple_loss=0.2653, pruned_loss=0.03457, over 12134.00 frames. ], tot_loss[loss=0.1702, simple_loss=0.2581, pruned_loss=0.04113, over 2369630.25 frames. ], batch size: 38, lr: 4.31e-03, grad_scale: 32.0 2023-05-16 09:35:14,700 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.7093, 3.1197, 5.0005, 2.6554, 2.7244, 3.7878, 3.0753, 3.8283], device='cuda:0'), covar=tensor([0.0360, 0.1170, 0.0276, 0.1111, 0.1924, 0.1236, 0.1353, 0.1153], device='cuda:0'), in_proj_covar=tensor([0.0234, 0.0231, 0.0238, 0.0180, 0.0235, 0.0284, 0.0221, 0.0254], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:0') 2023-05-16 09:35:15,326 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=189347.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 09:35:15,880 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.867e+02 2.651e+02 3.048e+02 3.558e+02 6.280e+02, threshold=6.096e+02, percent-clipped=0.0 2023-05-16 09:35:22,897 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.4763, 4.9503, 5.4347, 4.7532, 5.0291, 4.8288, 5.4862, 5.1494], device='cuda:0'), covar=tensor([0.0217, 0.0315, 0.0233, 0.0243, 0.0308, 0.0296, 0.0183, 0.0205], device='cuda:0'), in_proj_covar=tensor([0.0246, 0.0249, 0.0275, 0.0249, 0.0249, 0.0248, 0.0226, 0.0204], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 09:35:23,490 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.4290, 5.1676, 4.6252, 4.6560, 5.2372, 4.5358, 4.7647, 4.5995], device='cuda:0'), covar=tensor([0.1600, 0.0984, 0.1164, 0.2129, 0.0983, 0.2178, 0.1823, 0.1418], device='cuda:0'), in_proj_covar=tensor([0.0346, 0.0466, 0.0375, 0.0414, 0.0440, 0.0420, 0.0378, 0.0366], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 09:35:43,116 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=189386.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 09:35:43,753 INFO [finetune.py:992] (0/2) Epoch 8, batch 3750, loss[loss=0.1727, simple_loss=0.2695, pruned_loss=0.03792, over 12350.00 frames. ], tot_loss[loss=0.1706, simple_loss=0.2587, pruned_loss=0.04127, over 2372684.62 frames. ], batch size: 35, lr: 4.31e-03, grad_scale: 16.0 2023-05-16 09:35:59,337 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=189408.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 09:36:01,053 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.31 vs. limit=2.0 2023-05-16 09:36:08,265 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=189420.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 09:36:16,929 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.7544, 3.9704, 3.5528, 4.2551, 3.8773, 2.6321, 3.6395, 2.8730], device='cuda:0'), covar=tensor([0.0938, 0.0990, 0.1553, 0.0614, 0.1256, 0.1700, 0.1161, 0.3145], device='cuda:0'), in_proj_covar=tensor([0.0298, 0.0375, 0.0352, 0.0280, 0.0359, 0.0266, 0.0336, 0.0361], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 09:36:18,851 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=189435.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 09:36:20,208 INFO [finetune.py:992] (0/2) Epoch 8, batch 3800, loss[loss=0.1689, simple_loss=0.2588, pruned_loss=0.03947, over 12247.00 frames. ], tot_loss[loss=0.1709, simple_loss=0.2587, pruned_loss=0.0415, over 2367147.04 frames. ], batch size: 32, lr: 4.31e-03, grad_scale: 16.0 2023-05-16 09:36:24,611 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.0793, 4.7270, 4.8873, 4.9342, 4.7573, 4.9409, 4.8935, 2.6308], device='cuda:0'), covar=tensor([0.0094, 0.0075, 0.0081, 0.0066, 0.0048, 0.0081, 0.0075, 0.0789], device='cuda:0'), in_proj_covar=tensor([0.0065, 0.0076, 0.0079, 0.0072, 0.0058, 0.0089, 0.0078, 0.0094], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 09:36:28,643 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.003e+02 2.751e+02 3.207e+02 3.788e+02 9.973e+02, threshold=6.415e+02, percent-clipped=4.0 2023-05-16 09:36:43,781 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=4.62 vs. limit=5.0 2023-05-16 09:36:55,375 INFO [finetune.py:992] (0/2) Epoch 8, batch 3850, loss[loss=0.171, simple_loss=0.2639, pruned_loss=0.03902, over 12148.00 frames. ], tot_loss[loss=0.1705, simple_loss=0.2585, pruned_loss=0.04128, over 2378743.10 frames. ], batch size: 36, lr: 4.31e-03, grad_scale: 16.0 2023-05-16 09:37:05,557 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=189501.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 09:37:31,830 INFO [finetune.py:992] (0/2) Epoch 8, batch 3900, loss[loss=0.1897, simple_loss=0.2861, pruned_loss=0.04664, over 12008.00 frames. ], tot_loss[loss=0.1707, simple_loss=0.2587, pruned_loss=0.04133, over 2370397.20 frames. ], batch size: 40, lr: 4.31e-03, grad_scale: 16.0 2023-05-16 09:37:40,308 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.932e+02 2.778e+02 3.186e+02 3.785e+02 9.683e+02, threshold=6.372e+02, percent-clipped=2.0 2023-05-16 09:37:49,713 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=189562.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 09:38:02,539 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=3.03 vs. limit=5.0 2023-05-16 09:38:06,498 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=189585.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 09:38:07,775 INFO [finetune.py:992] (0/2) Epoch 8, batch 3950, loss[loss=0.1715, simple_loss=0.2606, pruned_loss=0.04123, over 12129.00 frames. ], tot_loss[loss=0.1715, simple_loss=0.2595, pruned_loss=0.04175, over 2364723.79 frames. ], batch size: 38, lr: 4.31e-03, grad_scale: 16.0 2023-05-16 09:38:07,938 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=189587.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 09:38:15,098 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.4699, 5.3042, 5.3573, 5.4429, 5.0484, 5.0990, 4.8926, 5.3412], device='cuda:0'), covar=tensor([0.0604, 0.0541, 0.0763, 0.0547, 0.1686, 0.1364, 0.0532, 0.1028], device='cuda:0'), in_proj_covar=tensor([0.0514, 0.0666, 0.0578, 0.0600, 0.0820, 0.0708, 0.0529, 0.0468], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0005, 0.0004, 0.0004, 0.0005, 0.0005, 0.0004, 0.0003], device='cuda:0') 2023-05-16 09:38:20,367 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=189604.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 09:38:28,865 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 2023-05-16 09:38:40,566 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=189633.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 09:38:41,902 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=189635.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 09:38:43,250 INFO [finetune.py:992] (0/2) Epoch 8, batch 4000, loss[loss=0.1673, simple_loss=0.2631, pruned_loss=0.03575, over 12204.00 frames. ], tot_loss[loss=0.1715, simple_loss=0.2599, pruned_loss=0.04161, over 2367940.87 frames. ], batch size: 35, lr: 4.31e-03, grad_scale: 16.0 2023-05-16 09:38:44,451 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=4.64 vs. limit=5.0 2023-05-16 09:38:45,540 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=189640.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 09:38:51,552 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.582e+02 2.907e+02 3.282e+02 4.224e+02 1.207e+03, threshold=6.563e+02, percent-clipped=1.0 2023-05-16 09:38:55,977 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.2788, 4.0984, 4.3095, 4.5512, 2.9879, 3.9113, 2.8378, 4.2130], device='cuda:0'), covar=tensor([0.1528, 0.0754, 0.0801, 0.0607, 0.1163, 0.0637, 0.1645, 0.1314], device='cuda:0'), in_proj_covar=tensor([0.0226, 0.0263, 0.0296, 0.0352, 0.0239, 0.0238, 0.0257, 0.0377], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 09:39:18,618 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=189686.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 09:39:19,187 INFO [finetune.py:992] (0/2) Epoch 8, batch 4050, loss[loss=0.1887, simple_loss=0.2797, pruned_loss=0.04887, over 11202.00 frames. ], tot_loss[loss=0.1721, simple_loss=0.2604, pruned_loss=0.04195, over 2375637.57 frames. ], batch size: 55, lr: 4.31e-03, grad_scale: 16.0 2023-05-16 09:39:29,345 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=189701.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 09:39:30,688 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=189703.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 09:39:31,574 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=189704.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 09:39:43,432 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=189720.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 09:39:48,906 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=189728.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 09:39:53,088 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=189734.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 09:39:53,868 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=189735.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 09:39:55,219 INFO [finetune.py:992] (0/2) Epoch 8, batch 4100, loss[loss=0.1481, simple_loss=0.231, pruned_loss=0.03259, over 11787.00 frames. ], tot_loss[loss=0.1722, simple_loss=0.2605, pruned_loss=0.04195, over 2380973.49 frames. ], batch size: 26, lr: 4.31e-03, grad_scale: 16.0 2023-05-16 09:40:03,530 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.851e+02 2.768e+02 3.461e+02 4.135e+02 7.228e+02, threshold=6.921e+02, percent-clipped=1.0 2023-05-16 09:40:15,247 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=189765.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 09:40:17,254 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=189768.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 09:40:27,848 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=189783.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 09:40:30,598 INFO [finetune.py:992] (0/2) Epoch 8, batch 4150, loss[loss=0.1801, simple_loss=0.2741, pruned_loss=0.04304, over 11752.00 frames. ], tot_loss[loss=0.1715, simple_loss=0.2601, pruned_loss=0.04147, over 2385128.07 frames. ], batch size: 44, lr: 4.31e-03, grad_scale: 16.0 2023-05-16 09:40:32,141 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=189789.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 09:40:36,528 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.8328, 2.2042, 3.1832, 3.7145, 3.3686, 3.7554, 3.4025, 2.7613], device='cuda:0'), covar=tensor([0.0047, 0.0402, 0.0150, 0.0049, 0.0152, 0.0072, 0.0119, 0.0332], device='cuda:0'), in_proj_covar=tensor([0.0084, 0.0120, 0.0102, 0.0075, 0.0100, 0.0114, 0.0092, 0.0136], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 09:40:53,745 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.2110, 5.1688, 4.9996, 5.0426, 4.7034, 5.1866, 5.1644, 5.4113], device='cuda:0'), covar=tensor([0.0200, 0.0136, 0.0196, 0.0299, 0.0692, 0.0261, 0.0142, 0.0141], device='cuda:0'), in_proj_covar=tensor([0.0189, 0.0192, 0.0187, 0.0242, 0.0237, 0.0210, 0.0171, 0.0217], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0005, 0.0004, 0.0005, 0.0003, 0.0005], device='cuda:0') 2023-05-16 09:41:07,504 INFO [finetune.py:992] (0/2) Epoch 8, batch 4200, loss[loss=0.1831, simple_loss=0.272, pruned_loss=0.04704, over 12125.00 frames. ], tot_loss[loss=0.1717, simple_loss=0.2604, pruned_loss=0.04157, over 2384461.74 frames. ], batch size: 38, lr: 4.31e-03, grad_scale: 16.0 2023-05-16 09:41:16,040 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.847e+02 2.753e+02 3.083e+02 3.674e+02 8.354e+02, threshold=6.165e+02, percent-clipped=1.0 2023-05-16 09:41:22,634 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=189857.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 09:41:31,997 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 2023-05-16 09:41:43,593 INFO [finetune.py:992] (0/2) Epoch 8, batch 4250, loss[loss=0.1854, simple_loss=0.2748, pruned_loss=0.04803, over 12045.00 frames. ], tot_loss[loss=0.1709, simple_loss=0.2596, pruned_loss=0.0411, over 2379718.36 frames. ], batch size: 37, lr: 4.31e-03, grad_scale: 16.0 2023-05-16 09:41:55,930 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=189904.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 09:42:19,165 INFO [finetune.py:992] (0/2) Epoch 8, batch 4300, loss[loss=0.1708, simple_loss=0.2649, pruned_loss=0.03835, over 12371.00 frames. ], tot_loss[loss=0.171, simple_loss=0.2597, pruned_loss=0.04116, over 2389249.58 frames. ], batch size: 36, lr: 4.31e-03, grad_scale: 16.0 2023-05-16 09:42:27,505 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.720e+02 2.709e+02 3.186e+02 3.885e+02 8.568e+02, threshold=6.372e+02, percent-clipped=2.0 2023-05-16 09:42:29,794 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=189952.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 09:42:55,447 INFO [finetune.py:992] (0/2) Epoch 8, batch 4350, loss[loss=0.1602, simple_loss=0.249, pruned_loss=0.03563, over 12298.00 frames. ], tot_loss[loss=0.1706, simple_loss=0.2588, pruned_loss=0.04118, over 2387977.68 frames. ], batch size: 33, lr: 4.31e-03, grad_scale: 16.0 2023-05-16 09:43:01,619 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=189996.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 09:43:05,325 INFO [checkpoint.py:75] (0/2) Saving checkpoint to pruned_transducer_stateless7/exp_giga_finetune/checkpoint-90000.pt 2023-05-16 09:43:10,610 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=190003.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 09:43:20,598 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.2457, 4.8013, 5.2447, 4.5782, 4.8496, 4.6735, 5.2965, 4.9776], device='cuda:0'), covar=tensor([0.0256, 0.0337, 0.0231, 0.0244, 0.0387, 0.0268, 0.0190, 0.0279], device='cuda:0'), in_proj_covar=tensor([0.0247, 0.0250, 0.0275, 0.0250, 0.0250, 0.0247, 0.0226, 0.0205], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 09:43:34,432 INFO [finetune.py:992] (0/2) Epoch 8, batch 4400, loss[loss=0.1494, simple_loss=0.2278, pruned_loss=0.03546, over 12003.00 frames. ], tot_loss[loss=0.1704, simple_loss=0.2583, pruned_loss=0.0413, over 2390177.61 frames. ], batch size: 28, lr: 4.31e-03, grad_scale: 16.0 2023-05-16 09:43:41,011 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.1670, 4.5812, 4.1350, 5.0040, 4.5521, 2.9707, 4.2956, 3.0632], device='cuda:0'), covar=tensor([0.0935, 0.0809, 0.1440, 0.0419, 0.1127, 0.1572, 0.1021, 0.3034], device='cuda:0'), in_proj_covar=tensor([0.0299, 0.0376, 0.0353, 0.0281, 0.0360, 0.0266, 0.0337, 0.0361], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 09:43:42,847 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.004e+02 2.848e+02 3.319e+02 3.981e+02 8.022e+02, threshold=6.638e+02, percent-clipped=3.0 2023-05-16 09:43:44,381 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=190051.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 09:43:50,659 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=190060.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 09:44:03,222 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=4.63 vs. limit=5.0 2023-05-16 09:44:07,608 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=190084.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 09:44:09,647 INFO [finetune.py:992] (0/2) Epoch 8, batch 4450, loss[loss=0.1708, simple_loss=0.2649, pruned_loss=0.03841, over 10372.00 frames. ], tot_loss[loss=0.1707, simple_loss=0.2586, pruned_loss=0.04139, over 2382004.51 frames. ], batch size: 68, lr: 4.30e-03, grad_scale: 16.0 2023-05-16 09:44:45,600 INFO [finetune.py:992] (0/2) Epoch 8, batch 4500, loss[loss=0.1893, simple_loss=0.27, pruned_loss=0.05426, over 10690.00 frames. ], tot_loss[loss=0.1704, simple_loss=0.2581, pruned_loss=0.04141, over 2378745.67 frames. ], batch size: 68, lr: 4.30e-03, grad_scale: 8.0 2023-05-16 09:44:47,199 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=190139.0, num_to_drop=1, layers_to_drop={0} 2023-05-16 09:44:55,474 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.964e+02 2.655e+02 3.257e+02 3.664e+02 9.318e+02, threshold=6.515e+02, percent-clipped=2.0 2023-05-16 09:44:56,459 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=190151.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 09:45:00,656 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=190157.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 09:45:21,873 INFO [finetune.py:992] (0/2) Epoch 8, batch 4550, loss[loss=0.1995, simple_loss=0.2851, pruned_loss=0.05692, over 12371.00 frames. ], tot_loss[loss=0.1699, simple_loss=0.2575, pruned_loss=0.04119, over 2379680.08 frames. ], batch size: 38, lr: 4.30e-03, grad_scale: 8.0 2023-05-16 09:45:27,053 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.0760, 2.3111, 3.2878, 3.9834, 3.5616, 3.9043, 3.5337, 2.7652], device='cuda:0'), covar=tensor([0.0045, 0.0393, 0.0168, 0.0042, 0.0128, 0.0076, 0.0118, 0.0376], device='cuda:0'), in_proj_covar=tensor([0.0083, 0.0119, 0.0101, 0.0074, 0.0099, 0.0113, 0.0091, 0.0134], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 09:45:31,675 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=190200.0, num_to_drop=1, layers_to_drop={3} 2023-05-16 09:45:35,095 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=190205.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 09:45:40,065 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=190212.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 09:45:54,554 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.87 vs. limit=2.0 2023-05-16 09:45:57,589 INFO [finetune.py:992] (0/2) Epoch 8, batch 4600, loss[loss=0.2233, simple_loss=0.3004, pruned_loss=0.07313, over 7535.00 frames. ], tot_loss[loss=0.1704, simple_loss=0.2582, pruned_loss=0.04127, over 2381961.07 frames. ], batch size: 98, lr: 4.30e-03, grad_scale: 8.0 2023-05-16 09:46:06,861 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.941e+02 2.690e+02 3.227e+02 3.913e+02 5.489e+02, threshold=6.454e+02, percent-clipped=0.0 2023-05-16 09:46:34,221 INFO [finetune.py:992] (0/2) Epoch 8, batch 4650, loss[loss=0.1823, simple_loss=0.2687, pruned_loss=0.04791, over 10345.00 frames. ], tot_loss[loss=0.1711, simple_loss=0.2591, pruned_loss=0.04154, over 2376299.15 frames. ], batch size: 68, lr: 4.30e-03, grad_scale: 8.0 2023-05-16 09:46:38,036 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.9344, 4.9016, 4.8239, 4.8132, 4.3271, 4.9814, 5.0126, 5.1906], device='cuda:0'), covar=tensor([0.0237, 0.0160, 0.0169, 0.0327, 0.0903, 0.0251, 0.0143, 0.0160], device='cuda:0'), in_proj_covar=tensor([0.0192, 0.0194, 0.0189, 0.0244, 0.0241, 0.0213, 0.0173, 0.0220], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0005, 0.0004, 0.0005, 0.0004, 0.0005], device='cuda:0') 2023-05-16 09:46:40,449 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.95 vs. limit=2.0 2023-05-16 09:46:40,876 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=190296.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 09:46:55,176 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=190316.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 09:47:10,130 INFO [finetune.py:992] (0/2) Epoch 8, batch 4700, loss[loss=0.1829, simple_loss=0.2704, pruned_loss=0.04768, over 12041.00 frames. ], tot_loss[loss=0.1719, simple_loss=0.2597, pruned_loss=0.04204, over 2367184.12 frames. ], batch size: 37, lr: 4.30e-03, grad_scale: 8.0 2023-05-16 09:47:15,207 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=190344.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 09:47:19,551 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.997e+02 2.904e+02 3.405e+02 3.927e+02 7.391e+02, threshold=6.810e+02, percent-clipped=1.0 2023-05-16 09:47:24,551 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.6273, 3.4061, 4.9882, 2.2002, 2.6149, 3.7091, 3.2393, 3.8533], device='cuda:0'), covar=tensor([0.0323, 0.0947, 0.0226, 0.1228, 0.1816, 0.1210, 0.1125, 0.0897], device='cuda:0'), in_proj_covar=tensor([0.0234, 0.0230, 0.0237, 0.0179, 0.0233, 0.0283, 0.0220, 0.0253], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:0') 2023-05-16 09:47:26,579 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=190360.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 09:47:34,620 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-05-16 09:47:38,837 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=190377.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 09:47:42,377 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.3298, 5.1737, 5.2742, 5.3101, 4.9415, 5.0013, 4.8042, 5.2214], device='cuda:0'), covar=tensor([0.0682, 0.0576, 0.0778, 0.0530, 0.1997, 0.1163, 0.0520, 0.1038], device='cuda:0'), in_proj_covar=tensor([0.0516, 0.0669, 0.0578, 0.0599, 0.0830, 0.0709, 0.0531, 0.0467], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0005, 0.0004, 0.0004, 0.0005, 0.0005, 0.0004, 0.0003], device='cuda:0') 2023-05-16 09:47:43,801 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=190384.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 09:47:45,814 INFO [finetune.py:992] (0/2) Epoch 8, batch 4750, loss[loss=0.1769, simple_loss=0.2618, pruned_loss=0.04598, over 12039.00 frames. ], tot_loss[loss=0.1726, simple_loss=0.2605, pruned_loss=0.04238, over 2365956.58 frames. ], batch size: 40, lr: 4.30e-03, grad_scale: 8.0 2023-05-16 09:47:55,135 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=190399.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 09:48:01,845 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=190408.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 09:48:04,156 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2023-05-16 09:48:19,622 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=190432.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 09:48:23,114 INFO [finetune.py:992] (0/2) Epoch 8, batch 4800, loss[loss=0.1809, simple_loss=0.2574, pruned_loss=0.05226, over 11800.00 frames. ], tot_loss[loss=0.173, simple_loss=0.2611, pruned_loss=0.04247, over 2372027.19 frames. ], batch size: 26, lr: 4.30e-03, grad_scale: 8.0 2023-05-16 09:48:32,413 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.074e+02 2.843e+02 3.222e+02 3.876e+02 1.576e+03, threshold=6.444e+02, percent-clipped=3.0 2023-05-16 09:48:39,802 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=190460.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 09:48:58,823 INFO [finetune.py:992] (0/2) Epoch 8, batch 4850, loss[loss=0.1639, simple_loss=0.2516, pruned_loss=0.03808, over 12284.00 frames. ], tot_loss[loss=0.1731, simple_loss=0.2614, pruned_loss=0.04237, over 2376234.65 frames. ], batch size: 33, lr: 4.30e-03, grad_scale: 8.0 2023-05-16 09:49:01,294 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.7726, 3.4792, 5.0930, 2.5079, 2.7227, 3.6814, 3.2745, 3.7732], device='cuda:0'), covar=tensor([0.0357, 0.0979, 0.0327, 0.1210, 0.1877, 0.1538, 0.1229, 0.1161], device='cuda:0'), in_proj_covar=tensor([0.0234, 0.0230, 0.0238, 0.0179, 0.0233, 0.0283, 0.0220, 0.0253], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:0') 2023-05-16 09:49:04,639 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=190495.0, num_to_drop=1, layers_to_drop={0} 2023-05-16 09:49:13,106 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=190507.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 09:49:34,165 INFO [finetune.py:992] (0/2) Epoch 8, batch 4900, loss[loss=0.1797, simple_loss=0.271, pruned_loss=0.04416, over 12345.00 frames. ], tot_loss[loss=0.1723, simple_loss=0.2604, pruned_loss=0.04205, over 2372570.29 frames. ], batch size: 36, lr: 4.30e-03, grad_scale: 8.0 2023-05-16 09:49:44,299 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.174e+02 2.711e+02 3.144e+02 3.905e+02 9.316e+02, threshold=6.287e+02, percent-clipped=4.0 2023-05-16 09:50:11,197 INFO [finetune.py:992] (0/2) Epoch 8, batch 4950, loss[loss=0.1446, simple_loss=0.2308, pruned_loss=0.02924, over 12338.00 frames. ], tot_loss[loss=0.1716, simple_loss=0.2597, pruned_loss=0.04173, over 2377548.34 frames. ], batch size: 31, lr: 4.30e-03, grad_scale: 8.0 2023-05-16 09:50:14,230 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=190591.0, num_to_drop=1, layers_to_drop={1} 2023-05-16 09:50:27,354 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=190609.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 09:50:47,350 INFO [finetune.py:992] (0/2) Epoch 8, batch 5000, loss[loss=0.2163, simple_loss=0.3004, pruned_loss=0.06614, over 12026.00 frames. ], tot_loss[loss=0.1712, simple_loss=0.2592, pruned_loss=0.04158, over 2372876.74 frames. ], batch size: 40, lr: 4.30e-03, grad_scale: 8.0 2023-05-16 09:50:56,665 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.804e+02 2.729e+02 3.218e+02 4.194e+02 6.963e+02, threshold=6.436e+02, percent-clipped=1.0 2023-05-16 09:50:58,217 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=190652.0, num_to_drop=1, layers_to_drop={1} 2023-05-16 09:51:10,972 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=190670.0, num_to_drop=1, layers_to_drop={0} 2023-05-16 09:51:12,313 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=190672.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 09:51:14,557 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=190675.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 09:51:23,625 INFO [finetune.py:992] (0/2) Epoch 8, batch 5050, loss[loss=0.1832, simple_loss=0.2808, pruned_loss=0.0428, over 12368.00 frames. ], tot_loss[loss=0.1723, simple_loss=0.2604, pruned_loss=0.04216, over 2372336.48 frames. ], batch size: 35, lr: 4.30e-03, grad_scale: 8.0 2023-05-16 09:51:27,361 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.3416, 5.0996, 5.2830, 5.2910, 4.8966, 4.9691, 4.7596, 5.2315], device='cuda:0'), covar=tensor([0.0650, 0.0597, 0.0734, 0.0543, 0.1933, 0.1251, 0.0534, 0.1013], device='cuda:0'), in_proj_covar=tensor([0.0518, 0.0676, 0.0579, 0.0603, 0.0834, 0.0711, 0.0532, 0.0470], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0005, 0.0004, 0.0004, 0.0005, 0.0005, 0.0004, 0.0003], device='cuda:0') 2023-05-16 09:51:28,880 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.5050, 2.4721, 3.3308, 4.3580, 2.6531, 4.5274, 4.4412, 4.7082], device='cuda:0'), covar=tensor([0.0121, 0.1217, 0.0409, 0.0169, 0.1086, 0.0187, 0.0149, 0.0071], device='cuda:0'), in_proj_covar=tensor([0.0119, 0.0200, 0.0179, 0.0114, 0.0188, 0.0173, 0.0171, 0.0117], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 09:51:32,441 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.9078, 4.4887, 4.6873, 4.7147, 4.5469, 4.8050, 4.6559, 2.4236], device='cuda:0'), covar=tensor([0.0128, 0.0087, 0.0106, 0.0086, 0.0071, 0.0113, 0.0100, 0.0957], device='cuda:0'), in_proj_covar=tensor([0.0065, 0.0076, 0.0079, 0.0071, 0.0058, 0.0088, 0.0077, 0.0094], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 09:51:33,338 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=3.96 vs. limit=5.0 2023-05-16 09:51:59,600 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=190736.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 09:52:00,155 INFO [finetune.py:992] (0/2) Epoch 8, batch 5100, loss[loss=0.1639, simple_loss=0.2485, pruned_loss=0.03967, over 12029.00 frames. ], tot_loss[loss=0.172, simple_loss=0.2601, pruned_loss=0.04195, over 2376827.63 frames. ], batch size: 31, lr: 4.30e-03, grad_scale: 8.0 2023-05-16 09:52:09,422 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.961e+02 2.722e+02 3.204e+02 3.972e+02 7.518e+02, threshold=6.407e+02, percent-clipped=2.0 2023-05-16 09:52:13,107 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=190755.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 09:52:35,753 INFO [finetune.py:992] (0/2) Epoch 8, batch 5150, loss[loss=0.1489, simple_loss=0.2306, pruned_loss=0.03363, over 12340.00 frames. ], tot_loss[loss=0.1711, simple_loss=0.2594, pruned_loss=0.04142, over 2378435.85 frames. ], batch size: 31, lr: 4.30e-03, grad_scale: 8.0 2023-05-16 09:52:41,581 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=190795.0, num_to_drop=1, layers_to_drop={0} 2023-05-16 09:52:50,231 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=190807.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 09:53:05,657 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 2023-05-16 09:53:12,291 INFO [finetune.py:992] (0/2) Epoch 8, batch 5200, loss[loss=0.1921, simple_loss=0.2911, pruned_loss=0.04651, over 12278.00 frames. ], tot_loss[loss=0.1709, simple_loss=0.2593, pruned_loss=0.0413, over 2380728.68 frames. ], batch size: 37, lr: 4.30e-03, grad_scale: 8.0 2023-05-16 09:53:16,729 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=190843.0, num_to_drop=1, layers_to_drop={1} 2023-05-16 09:53:21,685 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.026e+02 2.723e+02 3.198e+02 3.819e+02 9.971e+02, threshold=6.395e+02, percent-clipped=4.0 2023-05-16 09:53:25,439 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=190855.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 09:53:49,104 INFO [finetune.py:992] (0/2) Epoch 8, batch 5250, loss[loss=0.1755, simple_loss=0.2599, pruned_loss=0.04556, over 12096.00 frames. ], tot_loss[loss=0.1706, simple_loss=0.2589, pruned_loss=0.04114, over 2378377.43 frames. ], batch size: 32, lr: 4.30e-03, grad_scale: 8.0 2023-05-16 09:54:03,007 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.66 vs. limit=2.0 2023-05-16 09:54:06,593 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.57 vs. limit=2.0 2023-05-16 09:54:24,894 INFO [finetune.py:992] (0/2) Epoch 8, batch 5300, loss[loss=0.1765, simple_loss=0.2504, pruned_loss=0.05128, over 12001.00 frames. ], tot_loss[loss=0.1708, simple_loss=0.2588, pruned_loss=0.04138, over 2372818.50 frames. ], batch size: 28, lr: 4.30e-03, grad_scale: 8.0 2023-05-16 09:54:25,142 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.6990, 2.4299, 3.9966, 4.2606, 3.1064, 2.6219, 2.7311, 2.1283], device='cuda:0'), covar=tensor([0.1389, 0.3122, 0.0521, 0.0394, 0.0904, 0.2103, 0.2654, 0.4192], device='cuda:0'), in_proj_covar=tensor([0.0285, 0.0381, 0.0270, 0.0296, 0.0263, 0.0293, 0.0366, 0.0366], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-16 09:54:31,970 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=190947.0, num_to_drop=1, layers_to_drop={3} 2023-05-16 09:54:33,933 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.953e+02 2.779e+02 3.150e+02 3.652e+02 5.658e+02, threshold=6.299e+02, percent-clipped=0.0 2023-05-16 09:54:44,366 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=190965.0, num_to_drop=1, layers_to_drop={2} 2023-05-16 09:54:50,131 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=190972.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 09:54:56,465 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.5156, 2.1379, 3.7051, 4.3286, 3.9559, 4.3631, 4.0238, 2.9626], device='cuda:0'), covar=tensor([0.0041, 0.0542, 0.0139, 0.0051, 0.0117, 0.0071, 0.0112, 0.0394], device='cuda:0'), in_proj_covar=tensor([0.0082, 0.0117, 0.0101, 0.0074, 0.0097, 0.0112, 0.0089, 0.0133], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 09:55:00,572 INFO [finetune.py:992] (0/2) Epoch 8, batch 5350, loss[loss=0.1554, simple_loss=0.2441, pruned_loss=0.03333, over 12178.00 frames. ], tot_loss[loss=0.1709, simple_loss=0.2592, pruned_loss=0.0413, over 2378179.70 frames. ], batch size: 31, lr: 4.29e-03, grad_scale: 8.0 2023-05-16 09:55:23,844 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.1798, 4.8032, 5.0243, 4.9153, 4.7143, 5.0033, 4.8151, 3.0529], device='cuda:0'), covar=tensor([0.0082, 0.0076, 0.0080, 0.0071, 0.0063, 0.0090, 0.0088, 0.0641], device='cuda:0'), in_proj_covar=tensor([0.0065, 0.0076, 0.0078, 0.0071, 0.0058, 0.0088, 0.0078, 0.0094], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 09:55:25,216 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=191020.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 09:55:33,120 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=191031.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 09:55:37,377 INFO [finetune.py:992] (0/2) Epoch 8, batch 5400, loss[loss=0.171, simple_loss=0.2613, pruned_loss=0.04034, over 12292.00 frames. ], tot_loss[loss=0.1723, simple_loss=0.2601, pruned_loss=0.04226, over 2351302.43 frames. ], batch size: 33, lr: 4.29e-03, grad_scale: 8.0 2023-05-16 09:55:46,720 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.869e+02 3.027e+02 3.387e+02 3.939e+02 1.797e+03, threshold=6.773e+02, percent-clipped=3.0 2023-05-16 09:55:50,506 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=191055.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 09:55:59,744 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=191068.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 09:56:13,100 INFO [finetune.py:992] (0/2) Epoch 8, batch 5450, loss[loss=0.1626, simple_loss=0.2499, pruned_loss=0.03761, over 12145.00 frames. ], tot_loss[loss=0.1718, simple_loss=0.2599, pruned_loss=0.04186, over 2356884.00 frames. ], batch size: 30, lr: 4.29e-03, grad_scale: 8.0 2023-05-16 09:56:24,221 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=191103.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 09:56:43,113 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=191129.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 09:56:45,131 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=191132.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 09:56:48,521 INFO [finetune.py:992] (0/2) Epoch 8, batch 5500, loss[loss=0.1725, simple_loss=0.2651, pruned_loss=0.03998, over 12061.00 frames. ], tot_loss[loss=0.1723, simple_loss=0.2603, pruned_loss=0.04213, over 2357698.02 frames. ], batch size: 37, lr: 4.29e-03, grad_scale: 8.0 2023-05-16 09:56:58,291 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.004e+02 2.739e+02 3.385e+02 4.146e+02 3.019e+03, threshold=6.771e+02, percent-clipped=2.0 2023-05-16 09:57:17,035 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.3617, 4.8170, 2.9222, 2.6031, 4.1577, 2.7711, 4.0784, 3.5238], device='cuda:0'), covar=tensor([0.0676, 0.0510, 0.1034, 0.1451, 0.0266, 0.1168, 0.0431, 0.0667], device='cuda:0'), in_proj_covar=tensor([0.0188, 0.0252, 0.0174, 0.0197, 0.0140, 0.0179, 0.0192, 0.0174], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 09:57:23,246 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=191185.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 09:57:24,465 INFO [finetune.py:992] (0/2) Epoch 8, batch 5550, loss[loss=0.1522, simple_loss=0.2378, pruned_loss=0.03327, over 12248.00 frames. ], tot_loss[loss=0.1721, simple_loss=0.2601, pruned_loss=0.042, over 2366510.26 frames. ], batch size: 32, lr: 4.29e-03, grad_scale: 8.0 2023-05-16 09:57:28,976 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=191193.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 09:57:30,461 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.1907, 4.6305, 4.0753, 4.9647, 4.4528, 2.9440, 4.2560, 3.0976], device='cuda:0'), covar=tensor([0.0787, 0.0724, 0.1314, 0.0378, 0.0992, 0.1575, 0.0946, 0.3054], device='cuda:0'), in_proj_covar=tensor([0.0301, 0.0381, 0.0357, 0.0284, 0.0365, 0.0268, 0.0340, 0.0364], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 09:58:00,382 INFO [finetune.py:992] (0/2) Epoch 8, batch 5600, loss[loss=0.1427, simple_loss=0.2232, pruned_loss=0.03103, over 11809.00 frames. ], tot_loss[loss=0.1713, simple_loss=0.2595, pruned_loss=0.04154, over 2375316.63 frames. ], batch size: 26, lr: 4.29e-03, grad_scale: 8.0 2023-05-16 09:58:02,610 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.3159, 6.0364, 5.5252, 5.6071, 6.1308, 5.5158, 5.7361, 5.6556], device='cuda:0'), covar=tensor([0.1263, 0.0894, 0.1044, 0.1820, 0.0789, 0.1761, 0.1410, 0.1142], device='cuda:0'), in_proj_covar=tensor([0.0343, 0.0462, 0.0372, 0.0414, 0.0442, 0.0418, 0.0375, 0.0363], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 09:58:06,884 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=191246.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 09:58:07,498 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=191247.0, num_to_drop=1, layers_to_drop={1} 2023-05-16 09:58:09,411 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.949e+02 2.817e+02 3.355e+02 3.946e+02 6.887e+02, threshold=6.710e+02, percent-clipped=1.0 2023-05-16 09:58:20,901 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=191265.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 09:58:36,572 INFO [finetune.py:992] (0/2) Epoch 8, batch 5650, loss[loss=0.1804, simple_loss=0.2691, pruned_loss=0.04585, over 12099.00 frames. ], tot_loss[loss=0.1718, simple_loss=0.2602, pruned_loss=0.04169, over 2373167.25 frames. ], batch size: 38, lr: 4.29e-03, grad_scale: 8.0 2023-05-16 09:58:42,821 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=191295.0, num_to_drop=1, layers_to_drop={1} 2023-05-16 09:58:55,519 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=191313.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 09:59:06,988 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.2821, 4.7125, 2.8917, 2.5326, 4.0481, 2.5079, 4.1087, 3.3916], device='cuda:0'), covar=tensor([0.0778, 0.0459, 0.1149, 0.1608, 0.0294, 0.1545, 0.0384, 0.0812], device='cuda:0'), in_proj_covar=tensor([0.0186, 0.0251, 0.0173, 0.0196, 0.0139, 0.0178, 0.0191, 0.0173], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 09:59:08,223 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=191331.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 09:59:12,419 INFO [finetune.py:992] (0/2) Epoch 8, batch 5700, loss[loss=0.1705, simple_loss=0.2676, pruned_loss=0.03676, over 12053.00 frames. ], tot_loss[loss=0.1724, simple_loss=0.2607, pruned_loss=0.04208, over 2369568.08 frames. ], batch size: 40, lr: 4.29e-03, grad_scale: 8.0 2023-05-16 09:59:21,699 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.937e+02 2.701e+02 3.181e+02 3.913e+02 6.720e+02, threshold=6.362e+02, percent-clipped=1.0 2023-05-16 09:59:29,122 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.4860, 2.7107, 5.1695, 2.3785, 2.5495, 3.9811, 3.2339, 3.8792], device='cuda:0'), covar=tensor([0.0556, 0.1712, 0.0245, 0.1605, 0.2237, 0.1441, 0.1575, 0.1118], device='cuda:0'), in_proj_covar=tensor([0.0235, 0.0232, 0.0238, 0.0180, 0.0233, 0.0284, 0.0220, 0.0254], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:0') 2023-05-16 09:59:32,585 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=191365.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 09:59:34,927 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.57 vs. limit=2.0 2023-05-16 09:59:37,790 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.88 vs. limit=2.0 2023-05-16 09:59:42,444 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=191379.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 09:59:48,024 INFO [finetune.py:992] (0/2) Epoch 8, batch 5750, loss[loss=0.1628, simple_loss=0.2424, pruned_loss=0.04161, over 11787.00 frames. ], tot_loss[loss=0.1727, simple_loss=0.2612, pruned_loss=0.04209, over 2370509.64 frames. ], batch size: 26, lr: 4.29e-03, grad_scale: 8.0 2023-05-16 10:00:01,198 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.51 vs. limit=2.0 2023-05-16 10:00:15,209 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=191424.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 10:00:16,766 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=191426.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 10:00:25,231 INFO [finetune.py:992] (0/2) Epoch 8, batch 5800, loss[loss=0.1597, simple_loss=0.2329, pruned_loss=0.04319, over 12348.00 frames. ], tot_loss[loss=0.1719, simple_loss=0.2604, pruned_loss=0.04174, over 2376250.32 frames. ], batch size: 30, lr: 4.29e-03, grad_scale: 8.0 2023-05-16 10:00:34,273 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.981e+02 2.692e+02 3.071e+02 3.732e+02 6.496e+02, threshold=6.141e+02, percent-clipped=1.0 2023-05-16 10:01:00,493 INFO [finetune.py:992] (0/2) Epoch 8, batch 5850, loss[loss=0.1473, simple_loss=0.2304, pruned_loss=0.03207, over 12199.00 frames. ], tot_loss[loss=0.1714, simple_loss=0.2598, pruned_loss=0.04153, over 2376907.10 frames. ], batch size: 29, lr: 4.29e-03, grad_scale: 8.0 2023-05-16 10:01:01,250 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=191488.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 10:01:19,261 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.2755, 5.0698, 5.2205, 5.2581, 4.8786, 4.8851, 4.6943, 5.1687], device='cuda:0'), covar=tensor([0.0736, 0.0622, 0.0779, 0.0553, 0.1781, 0.1268, 0.0539, 0.1123], device='cuda:0'), in_proj_covar=tensor([0.0515, 0.0674, 0.0578, 0.0599, 0.0822, 0.0709, 0.0530, 0.0471], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0005, 0.0004, 0.0004, 0.0005, 0.0005, 0.0004, 0.0003], device='cuda:0') 2023-05-16 10:01:19,576 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=3.17 vs. limit=5.0 2023-05-16 10:01:36,101 INFO [finetune.py:992] (0/2) Epoch 8, batch 5900, loss[loss=0.156, simple_loss=0.2473, pruned_loss=0.03241, over 12103.00 frames. ], tot_loss[loss=0.1706, simple_loss=0.2588, pruned_loss=0.04124, over 2386690.82 frames. ], batch size: 33, lr: 4.29e-03, grad_scale: 8.0 2023-05-16 10:01:39,023 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=191541.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 10:01:46,011 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.085e+02 2.786e+02 3.233e+02 3.888e+02 1.854e+03, threshold=6.466e+02, percent-clipped=4.0 2023-05-16 10:01:49,842 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.2438, 3.2424, 3.0854, 3.0017, 2.7527, 2.4896, 3.2561, 2.0913], device='cuda:0'), covar=tensor([0.0432, 0.0134, 0.0163, 0.0179, 0.0353, 0.0365, 0.0132, 0.0479], device='cuda:0'), in_proj_covar=tensor([0.0189, 0.0160, 0.0154, 0.0181, 0.0200, 0.0196, 0.0164, 0.0197], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 10:02:12,294 INFO [finetune.py:992] (0/2) Epoch 8, batch 5950, loss[loss=0.1587, simple_loss=0.2444, pruned_loss=0.03655, over 12282.00 frames. ], tot_loss[loss=0.1712, simple_loss=0.2595, pruned_loss=0.04148, over 2381270.00 frames. ], batch size: 33, lr: 4.29e-03, grad_scale: 8.0 2023-05-16 10:02:32,930 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.7071, 2.6515, 3.6560, 4.7033, 3.9927, 4.6471, 4.0840, 3.3369], device='cuda:0'), covar=tensor([0.0029, 0.0361, 0.0143, 0.0025, 0.0108, 0.0067, 0.0091, 0.0309], device='cuda:0'), in_proj_covar=tensor([0.0083, 0.0118, 0.0101, 0.0074, 0.0098, 0.0114, 0.0090, 0.0134], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 10:02:48,863 INFO [finetune.py:992] (0/2) Epoch 8, batch 6000, loss[loss=0.1401, simple_loss=0.2206, pruned_loss=0.02982, over 12004.00 frames. ], tot_loss[loss=0.1707, simple_loss=0.2588, pruned_loss=0.04134, over 2379732.27 frames. ], batch size: 28, lr: 4.29e-03, grad_scale: 8.0 2023-05-16 10:02:48,864 INFO [finetune.py:1017] (0/2) Computing validation loss 2023-05-16 10:03:07,070 INFO [finetune.py:1026] (0/2) Epoch 8, validation: loss=0.3192, simple_loss=0.397, pruned_loss=0.1207, over 1020973.00 frames. 2023-05-16 10:03:07,071 INFO [finetune.py:1027] (0/2) Maximum memory allocated so far is 12761MB 2023-05-16 10:03:16,320 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.692e+02 2.721e+02 3.249e+02 3.799e+02 7.618e+02, threshold=6.498e+02, percent-clipped=2.0 2023-05-16 10:03:43,552 INFO [finetune.py:992] (0/2) Epoch 8, batch 6050, loss[loss=0.187, simple_loss=0.2847, pruned_loss=0.04458, over 12148.00 frames. ], tot_loss[loss=0.1716, simple_loss=0.2594, pruned_loss=0.04193, over 2369896.32 frames. ], batch size: 36, lr: 4.29e-03, grad_scale: 8.0 2023-05-16 10:03:52,202 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.5324, 4.7861, 4.2491, 5.1097, 4.7027, 2.9897, 4.4554, 3.0524], device='cuda:0'), covar=tensor([0.0750, 0.0739, 0.1392, 0.0411, 0.1118, 0.1479, 0.0799, 0.3058], device='cuda:0'), in_proj_covar=tensor([0.0296, 0.0376, 0.0352, 0.0280, 0.0360, 0.0264, 0.0336, 0.0357], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 10:04:07,727 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=191721.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 10:04:09,966 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=191724.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 10:04:10,319 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=4.86 vs. limit=5.0 2023-05-16 10:04:19,016 INFO [finetune.py:992] (0/2) Epoch 8, batch 6100, loss[loss=0.199, simple_loss=0.2843, pruned_loss=0.05682, over 10528.00 frames. ], tot_loss[loss=0.1725, simple_loss=0.2602, pruned_loss=0.04241, over 2363320.44 frames. ], batch size: 68, lr: 4.29e-03, grad_scale: 8.0 2023-05-16 10:04:28,200 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.776e+02 2.691e+02 3.267e+02 4.002e+02 6.127e+02, threshold=6.534e+02, percent-clipped=0.0 2023-05-16 10:04:36,491 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.4666, 4.7200, 4.2613, 5.1299, 4.6693, 3.0490, 4.4784, 3.1155], device='cuda:0'), covar=tensor([0.0731, 0.0792, 0.1321, 0.0363, 0.1013, 0.1499, 0.0984, 0.3050], device='cuda:0'), in_proj_covar=tensor([0.0296, 0.0375, 0.0352, 0.0279, 0.0359, 0.0263, 0.0335, 0.0356], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 10:04:43,750 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=191772.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 10:04:48,143 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.7226, 5.7289, 5.4887, 5.1283, 4.9156, 5.6224, 5.1815, 5.0706], device='cuda:0'), covar=tensor([0.0838, 0.1017, 0.0744, 0.1557, 0.0849, 0.0821, 0.1903, 0.1109], device='cuda:0'), in_proj_covar=tensor([0.0602, 0.0540, 0.0501, 0.0624, 0.0409, 0.0701, 0.0763, 0.0564], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0004, 0.0004, 0.0003], device='cuda:0') 2023-05-16 10:04:54,424 INFO [finetune.py:992] (0/2) Epoch 8, batch 6150, loss[loss=0.1679, simple_loss=0.266, pruned_loss=0.03492, over 12359.00 frames. ], tot_loss[loss=0.1726, simple_loss=0.2608, pruned_loss=0.04218, over 2367356.06 frames. ], batch size: 36, lr: 4.29e-03, grad_scale: 8.0 2023-05-16 10:04:55,280 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=191788.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 10:05:30,660 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=191836.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 10:05:31,301 INFO [finetune.py:992] (0/2) Epoch 8, batch 6200, loss[loss=0.1855, simple_loss=0.2729, pruned_loss=0.04904, over 12120.00 frames. ], tot_loss[loss=0.1728, simple_loss=0.2612, pruned_loss=0.04223, over 2368222.51 frames. ], batch size: 38, lr: 4.29e-03, grad_scale: 8.0 2023-05-16 10:05:34,421 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=191841.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 10:05:40,550 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.830e+02 2.958e+02 3.387e+02 4.141e+02 7.725e+02, threshold=6.774e+02, percent-clipped=2.0 2023-05-16 10:06:00,446 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.7364, 3.4312, 5.0337, 2.5981, 2.8085, 3.7591, 3.3849, 3.8854], device='cuda:0'), covar=tensor([0.0425, 0.1036, 0.0387, 0.1198, 0.1868, 0.1473, 0.1210, 0.1087], device='cuda:0'), in_proj_covar=tensor([0.0233, 0.0230, 0.0237, 0.0178, 0.0233, 0.0283, 0.0220, 0.0252], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:0') 2023-05-16 10:06:06,571 INFO [finetune.py:992] (0/2) Epoch 8, batch 6250, loss[loss=0.1499, simple_loss=0.2359, pruned_loss=0.0319, over 12344.00 frames. ], tot_loss[loss=0.1732, simple_loss=0.2615, pruned_loss=0.04242, over 2364637.22 frames. ], batch size: 31, lr: 4.29e-03, grad_scale: 8.0 2023-05-16 10:06:08,074 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=191889.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 10:06:16,554 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.9683, 5.8177, 5.4264, 5.3740, 5.9164, 5.3086, 5.4823, 5.4238], device='cuda:0'), covar=tensor([0.1525, 0.0922, 0.1051, 0.1662, 0.0963, 0.1977, 0.1784, 0.1239], device='cuda:0'), in_proj_covar=tensor([0.0348, 0.0465, 0.0374, 0.0415, 0.0443, 0.0423, 0.0379, 0.0369], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 10:06:41,759 INFO [finetune.py:992] (0/2) Epoch 8, batch 6300, loss[loss=0.1758, simple_loss=0.258, pruned_loss=0.04682, over 12350.00 frames. ], tot_loss[loss=0.1742, simple_loss=0.2626, pruned_loss=0.04294, over 2364224.31 frames. ], batch size: 31, lr: 4.28e-03, grad_scale: 8.0 2023-05-16 10:06:51,670 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.004e+02 2.808e+02 3.454e+02 4.208e+02 8.526e+02, threshold=6.909e+02, percent-clipped=2.0 2023-05-16 10:07:06,046 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.0211, 2.2316, 3.0866, 3.9246, 2.2017, 4.0642, 3.9236, 4.1346], device='cuda:0'), covar=tensor([0.0139, 0.1270, 0.0429, 0.0135, 0.1222, 0.0231, 0.0205, 0.0091], device='cuda:0'), in_proj_covar=tensor([0.0120, 0.0201, 0.0179, 0.0113, 0.0188, 0.0173, 0.0171, 0.0118], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 10:07:08,411 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.51 vs. limit=2.0 2023-05-16 10:07:18,489 INFO [finetune.py:992] (0/2) Epoch 8, batch 6350, loss[loss=0.2057, simple_loss=0.2943, pruned_loss=0.0585, over 11342.00 frames. ], tot_loss[loss=0.1739, simple_loss=0.2624, pruned_loss=0.04272, over 2369419.21 frames. ], batch size: 55, lr: 4.28e-03, grad_scale: 8.0 2023-05-16 10:07:28,001 INFO [checkpoint.py:75] (0/2) Saving checkpoint to pruned_transducer_stateless7/exp_giga_finetune/checkpoint-92000.pt 2023-05-16 10:07:45,993 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=192021.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 10:07:47,446 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.1902, 2.1326, 2.7270, 3.0530, 3.0123, 3.1517, 2.8935, 2.2756], device='cuda:0'), covar=tensor([0.0058, 0.0329, 0.0180, 0.0067, 0.0112, 0.0090, 0.0110, 0.0340], device='cuda:0'), in_proj_covar=tensor([0.0083, 0.0118, 0.0101, 0.0074, 0.0098, 0.0114, 0.0089, 0.0134], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 10:07:57,027 INFO [finetune.py:992] (0/2) Epoch 8, batch 6400, loss[loss=0.1656, simple_loss=0.259, pruned_loss=0.0361, over 12354.00 frames. ], tot_loss[loss=0.174, simple_loss=0.2624, pruned_loss=0.04278, over 2370688.39 frames. ], batch size: 36, lr: 4.28e-03, grad_scale: 8.0 2023-05-16 10:08:06,314 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.610e+02 2.917e+02 3.513e+02 4.206e+02 9.694e+02, threshold=7.025e+02, percent-clipped=2.0 2023-05-16 10:08:09,529 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.7593, 3.5126, 5.1449, 2.7237, 2.7335, 3.8000, 3.2442, 3.9281], device='cuda:0'), covar=tensor([0.0410, 0.1012, 0.0256, 0.1083, 0.1906, 0.1475, 0.1291, 0.0964], device='cuda:0'), in_proj_covar=tensor([0.0233, 0.0230, 0.0237, 0.0178, 0.0232, 0.0282, 0.0219, 0.0251], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:0') 2023-05-16 10:08:20,246 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=192069.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 10:08:30,376 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.4564, 3.8361, 3.8398, 4.3420, 3.2084, 3.8522, 2.6572, 4.2516], device='cuda:0'), covar=tensor([0.1224, 0.0756, 0.1138, 0.0795, 0.0870, 0.0583, 0.1517, 0.1045], device='cuda:0'), in_proj_covar=tensor([0.0228, 0.0266, 0.0293, 0.0353, 0.0238, 0.0240, 0.0260, 0.0380], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 10:08:33,715 INFO [finetune.py:992] (0/2) Epoch 8, batch 6450, loss[loss=0.1776, simple_loss=0.2627, pruned_loss=0.04622, over 12083.00 frames. ], tot_loss[loss=0.1729, simple_loss=0.2617, pruned_loss=0.04205, over 2372415.63 frames. ], batch size: 40, lr: 4.28e-03, grad_scale: 8.0 2023-05-16 10:08:44,582 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=192102.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 10:08:55,928 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.1292, 2.3286, 3.6318, 3.0868, 3.4503, 3.1952, 2.4364, 3.5110], device='cuda:0'), covar=tensor([0.0118, 0.0375, 0.0151, 0.0213, 0.0164, 0.0173, 0.0409, 0.0131], device='cuda:0'), in_proj_covar=tensor([0.0171, 0.0203, 0.0182, 0.0180, 0.0210, 0.0156, 0.0195, 0.0180], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 10:09:05,795 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.1469, 3.8899, 3.9688, 4.3351, 2.9018, 3.8652, 2.4742, 4.1589], device='cuda:0'), covar=tensor([0.1671, 0.0804, 0.0915, 0.0643, 0.1214, 0.0658, 0.1937, 0.1018], device='cuda:0'), in_proj_covar=tensor([0.0227, 0.0265, 0.0293, 0.0353, 0.0237, 0.0239, 0.0259, 0.0380], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 10:09:09,806 INFO [finetune.py:992] (0/2) Epoch 8, batch 6500, loss[loss=0.1978, simple_loss=0.2919, pruned_loss=0.05184, over 12047.00 frames. ], tot_loss[loss=0.173, simple_loss=0.2616, pruned_loss=0.04224, over 2373863.63 frames. ], batch size: 42, lr: 4.28e-03, grad_scale: 16.0 2023-05-16 10:09:19,205 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.892e+02 2.680e+02 3.163e+02 3.527e+02 5.515e+02, threshold=6.326e+02, percent-clipped=0.0 2023-05-16 10:09:22,921 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=192155.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 10:09:28,505 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=192163.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 10:09:44,957 INFO [finetune.py:992] (0/2) Epoch 8, batch 6550, loss[loss=0.2191, simple_loss=0.2951, pruned_loss=0.07155, over 12337.00 frames. ], tot_loss[loss=0.173, simple_loss=0.2614, pruned_loss=0.04232, over 2376121.17 frames. ], batch size: 38, lr: 4.28e-03, grad_scale: 16.0 2023-05-16 10:09:59,096 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.2518, 2.6247, 3.7224, 3.1629, 3.5927, 3.2592, 2.6869, 3.6908], device='cuda:0'), covar=tensor([0.0116, 0.0332, 0.0155, 0.0207, 0.0125, 0.0171, 0.0339, 0.0096], device='cuda:0'), in_proj_covar=tensor([0.0173, 0.0205, 0.0183, 0.0181, 0.0211, 0.0157, 0.0196, 0.0181], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 10:10:05,912 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=192216.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 10:10:07,972 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=192219.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 10:10:09,430 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=192221.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 10:10:21,131 INFO [finetune.py:992] (0/2) Epoch 8, batch 6600, loss[loss=0.154, simple_loss=0.2465, pruned_loss=0.03075, over 12087.00 frames. ], tot_loss[loss=0.174, simple_loss=0.2624, pruned_loss=0.04284, over 2376304.07 frames. ], batch size: 33, lr: 4.28e-03, grad_scale: 16.0 2023-05-16 10:10:30,378 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.044e+02 2.904e+02 3.471e+02 4.736e+02 9.167e+02, threshold=6.942e+02, percent-clipped=6.0 2023-05-16 10:10:52,768 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=192280.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 10:10:54,202 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=192282.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 10:10:57,648 INFO [finetune.py:992] (0/2) Epoch 8, batch 6650, loss[loss=0.2021, simple_loss=0.2813, pruned_loss=0.06148, over 12068.00 frames. ], tot_loss[loss=0.1739, simple_loss=0.262, pruned_loss=0.04285, over 2376851.87 frames. ], batch size: 40, lr: 4.28e-03, grad_scale: 16.0 2023-05-16 10:11:31,961 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.2827, 5.2098, 5.1265, 5.1579, 4.8096, 5.2743, 5.2476, 5.5180], device='cuda:0'), covar=tensor([0.0200, 0.0130, 0.0151, 0.0257, 0.0650, 0.0253, 0.0141, 0.0147], device='cuda:0'), in_proj_covar=tensor([0.0188, 0.0190, 0.0186, 0.0239, 0.0237, 0.0209, 0.0170, 0.0218], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0005, 0.0004, 0.0004, 0.0003, 0.0005], device='cuda:0') 2023-05-16 10:11:33,302 INFO [finetune.py:992] (0/2) Epoch 8, batch 6700, loss[loss=0.1687, simple_loss=0.2536, pruned_loss=0.04189, over 12094.00 frames. ], tot_loss[loss=0.1729, simple_loss=0.2608, pruned_loss=0.04244, over 2377470.19 frames. ], batch size: 32, lr: 4.28e-03, grad_scale: 16.0 2023-05-16 10:11:40,313 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=4.72 vs. limit=5.0 2023-05-16 10:11:42,648 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.829e+02 2.670e+02 3.043e+02 3.525e+02 1.080e+03, threshold=6.085e+02, percent-clipped=1.0 2023-05-16 10:12:09,494 INFO [finetune.py:992] (0/2) Epoch 8, batch 6750, loss[loss=0.1447, simple_loss=0.2246, pruned_loss=0.03236, over 12015.00 frames. ], tot_loss[loss=0.1726, simple_loss=0.2608, pruned_loss=0.04221, over 2378090.59 frames. ], batch size: 28, lr: 4.28e-03, grad_scale: 16.0 2023-05-16 10:12:13,176 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=192392.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 10:12:18,028 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=192399.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 10:12:27,321 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.6836, 2.2931, 3.0010, 2.6442, 2.8891, 2.8127, 2.2297, 2.9404], device='cuda:0'), covar=tensor([0.0102, 0.0314, 0.0158, 0.0215, 0.0160, 0.0168, 0.0302, 0.0158], device='cuda:0'), in_proj_covar=tensor([0.0171, 0.0205, 0.0184, 0.0181, 0.0211, 0.0157, 0.0196, 0.0181], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 10:12:39,289 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2023-05-16 10:12:42,465 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=192433.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 10:12:45,158 INFO [finetune.py:992] (0/2) Epoch 8, batch 6800, loss[loss=0.1767, simple_loss=0.2583, pruned_loss=0.04759, over 12178.00 frames. ], tot_loss[loss=0.173, simple_loss=0.2612, pruned_loss=0.04233, over 2382432.12 frames. ], batch size: 31, lr: 4.28e-03, grad_scale: 16.0 2023-05-16 10:12:54,370 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.996e+02 2.674e+02 3.537e+02 4.460e+02 1.477e+03, threshold=7.075e+02, percent-clipped=4.0 2023-05-16 10:12:56,685 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=192453.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 10:13:00,142 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=192458.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 10:13:01,728 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=192460.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 10:13:20,842 INFO [finetune.py:992] (0/2) Epoch 8, batch 6850, loss[loss=0.1495, simple_loss=0.2269, pruned_loss=0.03606, over 12306.00 frames. ], tot_loss[loss=0.1725, simple_loss=0.2608, pruned_loss=0.04212, over 2384395.08 frames. ], batch size: 28, lr: 4.28e-03, grad_scale: 16.0 2023-05-16 10:13:25,984 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=192494.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 10:13:26,815 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.0981, 2.5106, 3.7246, 3.0563, 3.4487, 3.2552, 2.5066, 3.5386], device='cuda:0'), covar=tensor([0.0126, 0.0372, 0.0157, 0.0270, 0.0147, 0.0170, 0.0372, 0.0137], device='cuda:0'), in_proj_covar=tensor([0.0172, 0.0206, 0.0184, 0.0183, 0.0212, 0.0158, 0.0198, 0.0182], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 10:13:38,262 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=192511.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 10:13:57,439 INFO [finetune.py:992] (0/2) Epoch 8, batch 6900, loss[loss=0.1523, simple_loss=0.2512, pruned_loss=0.02665, over 12193.00 frames. ], tot_loss[loss=0.1728, simple_loss=0.2612, pruned_loss=0.04224, over 2384504.54 frames. ], batch size: 35, lr: 4.28e-03, grad_scale: 16.0 2023-05-16 10:14:07,467 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.737e+02 2.847e+02 3.151e+02 3.808e+02 6.868e+02, threshold=6.302e+02, percent-clipped=0.0 2023-05-16 10:14:07,838 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=2.46 vs. limit=5.0 2023-05-16 10:14:18,324 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.6510, 3.6677, 3.3446, 3.2395, 2.8725, 2.8477, 3.6158, 2.2766], device='cuda:0'), covar=tensor([0.0340, 0.0154, 0.0164, 0.0201, 0.0362, 0.0321, 0.0129, 0.0488], device='cuda:0'), in_proj_covar=tensor([0.0192, 0.0163, 0.0155, 0.0184, 0.0201, 0.0200, 0.0167, 0.0200], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 10:14:25,154 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=192575.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 10:14:26,603 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=192577.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 10:14:33,742 INFO [finetune.py:992] (0/2) Epoch 8, batch 6950, loss[loss=0.1947, simple_loss=0.2807, pruned_loss=0.05436, over 12122.00 frames. ], tot_loss[loss=0.1721, simple_loss=0.2602, pruned_loss=0.04198, over 2380782.10 frames. ], batch size: 42, lr: 4.28e-03, grad_scale: 16.0 2023-05-16 10:14:47,110 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.7672, 2.3665, 3.2016, 2.8127, 3.1183, 2.9023, 2.3051, 3.2033], device='cuda:0'), covar=tensor([0.0127, 0.0332, 0.0169, 0.0226, 0.0139, 0.0169, 0.0345, 0.0127], device='cuda:0'), in_proj_covar=tensor([0.0171, 0.0205, 0.0184, 0.0182, 0.0211, 0.0158, 0.0197, 0.0181], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 10:14:54,869 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=192616.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 10:15:09,837 INFO [finetune.py:992] (0/2) Epoch 8, batch 7000, loss[loss=0.1426, simple_loss=0.2234, pruned_loss=0.03095, over 12184.00 frames. ], tot_loss[loss=0.1719, simple_loss=0.2601, pruned_loss=0.04187, over 2387555.14 frames. ], batch size: 29, lr: 4.28e-03, grad_scale: 16.0 2023-05-16 10:15:19,101 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.884e+02 2.679e+02 3.248e+02 3.813e+02 6.575e+02, threshold=6.495e+02, percent-clipped=2.0 2023-05-16 10:15:36,635 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=192673.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 10:15:39,508 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=192677.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 10:15:43,176 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2023-05-16 10:15:46,244 INFO [finetune.py:992] (0/2) Epoch 8, batch 7050, loss[loss=0.1788, simple_loss=0.273, pruned_loss=0.04236, over 12161.00 frames. ], tot_loss[loss=0.1722, simple_loss=0.2607, pruned_loss=0.04185, over 2373208.70 frames. ], batch size: 36, lr: 4.28e-03, grad_scale: 16.0 2023-05-16 10:16:20,701 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=192734.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 10:16:21,405 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=192735.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 10:16:22,656 INFO [finetune.py:992] (0/2) Epoch 8, batch 7100, loss[loss=0.1674, simple_loss=0.2539, pruned_loss=0.04042, over 12301.00 frames. ], tot_loss[loss=0.1712, simple_loss=0.2595, pruned_loss=0.04141, over 2379887.23 frames. ], batch size: 34, lr: 4.28e-03, grad_scale: 16.0 2023-05-16 10:16:30,566 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=192748.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 10:16:31,938 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.740e+02 2.754e+02 3.372e+02 4.025e+02 1.084e+03, threshold=6.745e+02, percent-clipped=3.0 2023-05-16 10:16:35,375 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=192755.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 10:16:37,569 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=192758.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 10:16:58,087 INFO [finetune.py:992] (0/2) Epoch 8, batch 7150, loss[loss=0.1847, simple_loss=0.2758, pruned_loss=0.04681, over 12352.00 frames. ], tot_loss[loss=0.1715, simple_loss=0.26, pruned_loss=0.04152, over 2376298.76 frames. ], batch size: 35, lr: 4.28e-03, grad_scale: 16.0 2023-05-16 10:16:59,584 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=192789.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 10:17:04,704 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=192796.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 10:17:07,829 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.3040, 2.5764, 3.9018, 3.2432, 3.5832, 3.3226, 2.7291, 3.7454], device='cuda:0'), covar=tensor([0.0107, 0.0339, 0.0115, 0.0210, 0.0147, 0.0156, 0.0335, 0.0108], device='cuda:0'), in_proj_covar=tensor([0.0172, 0.0205, 0.0184, 0.0182, 0.0212, 0.0158, 0.0197, 0.0181], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 10:17:12,023 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=192806.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 10:17:15,786 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=192811.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 10:17:24,731 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.6091, 2.8320, 3.5898, 4.7007, 3.9456, 4.7568, 4.1182, 3.2259], device='cuda:0'), covar=tensor([0.0039, 0.0353, 0.0160, 0.0033, 0.0124, 0.0055, 0.0094, 0.0359], device='cuda:0'), in_proj_covar=tensor([0.0084, 0.0119, 0.0103, 0.0075, 0.0099, 0.0115, 0.0091, 0.0136], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 10:17:30,382 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=192831.0, num_to_drop=1, layers_to_drop={0} 2023-05-16 10:17:34,986 INFO [finetune.py:992] (0/2) Epoch 8, batch 7200, loss[loss=0.2011, simple_loss=0.2863, pruned_loss=0.05788, over 12088.00 frames. ], tot_loss[loss=0.1711, simple_loss=0.2593, pruned_loss=0.04149, over 2380177.77 frames. ], batch size: 42, lr: 4.28e-03, grad_scale: 16.0 2023-05-16 10:17:44,260 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.889e+02 2.677e+02 3.312e+02 4.462e+02 1.222e+03, threshold=6.625e+02, percent-clipped=7.0 2023-05-16 10:17:50,856 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=192859.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 10:18:02,106 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=192875.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 10:18:03,635 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=192877.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 10:18:10,459 INFO [finetune.py:992] (0/2) Epoch 8, batch 7250, loss[loss=0.1583, simple_loss=0.2402, pruned_loss=0.03825, over 11759.00 frames. ], tot_loss[loss=0.1718, simple_loss=0.2601, pruned_loss=0.0418, over 2384207.65 frames. ], batch size: 26, lr: 4.27e-03, grad_scale: 16.0 2023-05-16 10:18:14,221 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=192892.0, num_to_drop=1, layers_to_drop={2} 2023-05-16 10:18:35,947 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=192923.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 10:18:37,308 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=192925.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 10:18:45,916 INFO [finetune.py:992] (0/2) Epoch 8, batch 7300, loss[loss=0.1623, simple_loss=0.2546, pruned_loss=0.03503, over 12284.00 frames. ], tot_loss[loss=0.1714, simple_loss=0.2598, pruned_loss=0.04157, over 2384496.42 frames. ], batch size: 33, lr: 4.27e-03, grad_scale: 16.0 2023-05-16 10:18:55,089 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.871e+02 2.770e+02 3.180e+02 3.835e+02 6.912e+02, threshold=6.361e+02, percent-clipped=1.0 2023-05-16 10:19:11,492 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=192972.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 10:19:12,355 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.7859, 2.5730, 3.5311, 3.6725, 3.0260, 2.7388, 2.6262, 2.4323], device='cuda:0'), covar=tensor([0.1094, 0.2349, 0.0606, 0.0472, 0.0793, 0.1672, 0.2250, 0.2921], device='cuda:0'), in_proj_covar=tensor([0.0282, 0.0378, 0.0266, 0.0295, 0.0260, 0.0291, 0.0362, 0.0362], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-16 10:19:22,580 INFO [finetune.py:992] (0/2) Epoch 8, batch 7350, loss[loss=0.2003, simple_loss=0.2932, pruned_loss=0.05374, over 12368.00 frames. ], tot_loss[loss=0.1715, simple_loss=0.2596, pruned_loss=0.04172, over 2378135.63 frames. ], batch size: 36, lr: 4.27e-03, grad_scale: 16.0 2023-05-16 10:19:44,493 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.1158, 4.4705, 3.8996, 4.7974, 4.2187, 2.7660, 4.0637, 2.8620], device='cuda:0'), covar=tensor([0.0838, 0.0827, 0.1434, 0.0525, 0.1273, 0.1712, 0.1078, 0.3434], device='cuda:0'), in_proj_covar=tensor([0.0293, 0.0373, 0.0351, 0.0278, 0.0358, 0.0263, 0.0333, 0.0355], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 10:19:45,909 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=2.53 vs. limit=5.0 2023-05-16 10:19:52,923 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=193029.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 10:19:58,429 INFO [finetune.py:992] (0/2) Epoch 8, batch 7400, loss[loss=0.1559, simple_loss=0.2406, pruned_loss=0.03559, over 12115.00 frames. ], tot_loss[loss=0.1726, simple_loss=0.2606, pruned_loss=0.04227, over 2368845.47 frames. ], batch size: 30, lr: 4.27e-03, grad_scale: 16.0 2023-05-16 10:20:00,998 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.81 vs. limit=2.0 2023-05-16 10:20:01,549 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=193041.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 10:20:06,466 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=193048.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 10:20:07,828 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.950e+02 2.810e+02 3.225e+02 3.988e+02 1.116e+03, threshold=6.451e+02, percent-clipped=6.0 2023-05-16 10:20:11,571 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=193055.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 10:20:22,325 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.5423, 5.0706, 5.5061, 4.7709, 5.1275, 4.8839, 5.5416, 5.1555], device='cuda:0'), covar=tensor([0.0183, 0.0297, 0.0210, 0.0215, 0.0314, 0.0255, 0.0173, 0.0248], device='cuda:0'), in_proj_covar=tensor([0.0248, 0.0256, 0.0276, 0.0252, 0.0250, 0.0248, 0.0230, 0.0208], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 10:20:29,053 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.37 vs. limit=2.0 2023-05-16 10:20:29,425 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.7133, 2.0753, 3.1196, 3.7057, 3.4046, 3.7495, 3.4367, 2.4624], device='cuda:0'), covar=tensor([0.0060, 0.0427, 0.0182, 0.0052, 0.0107, 0.0081, 0.0107, 0.0432], device='cuda:0'), in_proj_covar=tensor([0.0084, 0.0118, 0.0102, 0.0074, 0.0098, 0.0114, 0.0090, 0.0134], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 10:20:34,154 INFO [finetune.py:992] (0/2) Epoch 8, batch 7450, loss[loss=0.1631, simple_loss=0.2554, pruned_loss=0.03539, over 12106.00 frames. ], tot_loss[loss=0.1723, simple_loss=0.2605, pruned_loss=0.04205, over 2373160.80 frames. ], batch size: 33, lr: 4.27e-03, grad_scale: 16.0 2023-05-16 10:20:35,797 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=193089.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 10:20:37,177 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=193091.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 10:20:38,019 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.2103, 2.5711, 3.7620, 3.1593, 3.5202, 3.3167, 2.6612, 3.6518], device='cuda:0'), covar=tensor([0.0123, 0.0311, 0.0172, 0.0236, 0.0160, 0.0145, 0.0302, 0.0125], device='cuda:0'), in_proj_covar=tensor([0.0170, 0.0203, 0.0183, 0.0180, 0.0210, 0.0156, 0.0195, 0.0179], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 10:20:40,800 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=193096.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 10:20:45,897 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=193102.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 10:20:46,449 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=193103.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 10:20:57,659 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.38 vs. limit=2.0 2023-05-16 10:21:11,261 INFO [finetune.py:992] (0/2) Epoch 8, batch 7500, loss[loss=0.1735, simple_loss=0.2643, pruned_loss=0.04142, over 12113.00 frames. ], tot_loss[loss=0.1722, simple_loss=0.2607, pruned_loss=0.0419, over 2373276.37 frames. ], batch size: 39, lr: 4.27e-03, grad_scale: 16.0 2023-05-16 10:21:11,337 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=193137.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 10:21:20,457 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.770e+02 2.655e+02 3.373e+02 4.142e+02 9.369e+02, threshold=6.746e+02, percent-clipped=3.0 2023-05-16 10:21:46,382 INFO [finetune.py:992] (0/2) Epoch 8, batch 7550, loss[loss=0.1854, simple_loss=0.2707, pruned_loss=0.05006, over 11179.00 frames. ], tot_loss[loss=0.1729, simple_loss=0.2611, pruned_loss=0.04236, over 2373551.95 frames. ], batch size: 55, lr: 4.27e-03, grad_scale: 16.0 2023-05-16 10:21:46,460 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=193187.0, num_to_drop=1, layers_to_drop={1} 2023-05-16 10:22:03,593 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=193211.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 10:22:21,697 INFO [finetune.py:992] (0/2) Epoch 8, batch 7600, loss[loss=0.1526, simple_loss=0.2441, pruned_loss=0.03051, over 12278.00 frames. ], tot_loss[loss=0.1732, simple_loss=0.2615, pruned_loss=0.0424, over 2374524.89 frames. ], batch size: 33, lr: 4.27e-03, grad_scale: 16.0 2023-05-16 10:22:31,569 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.059e+02 2.941e+02 3.431e+02 4.021e+02 6.757e+02, threshold=6.862e+02, percent-clipped=1.0 2023-05-16 10:22:48,062 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=193272.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 10:22:48,096 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=193272.0, num_to_drop=1, layers_to_drop={0} 2023-05-16 10:22:56,658 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=193284.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 10:22:58,691 INFO [finetune.py:992] (0/2) Epoch 8, batch 7650, loss[loss=0.1445, simple_loss=0.2273, pruned_loss=0.03085, over 11993.00 frames. ], tot_loss[loss=0.1723, simple_loss=0.2607, pruned_loss=0.04192, over 2376669.53 frames. ], batch size: 28, lr: 4.27e-03, grad_scale: 16.0 2023-05-16 10:23:21,978 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=193320.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 10:23:28,385 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=193329.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 10:23:33,880 INFO [finetune.py:992] (0/2) Epoch 8, batch 7700, loss[loss=0.1709, simple_loss=0.2658, pruned_loss=0.03805, over 12149.00 frames. ], tot_loss[loss=0.1725, simple_loss=0.2609, pruned_loss=0.04207, over 2381496.13 frames. ], batch size: 34, lr: 4.27e-03, grad_scale: 16.0 2023-05-16 10:23:39,687 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=193345.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 10:23:43,034 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.152e+02 2.912e+02 3.447e+02 4.156e+02 7.107e+02, threshold=6.894e+02, percent-clipped=1.0 2023-05-16 10:24:02,192 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=193377.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 10:24:09,243 INFO [finetune.py:992] (0/2) Epoch 8, batch 7750, loss[loss=0.1835, simple_loss=0.2699, pruned_loss=0.04854, over 11676.00 frames. ], tot_loss[loss=0.172, simple_loss=0.2602, pruned_loss=0.04194, over 2380947.88 frames. ], batch size: 48, lr: 4.27e-03, grad_scale: 16.0 2023-05-16 10:24:12,359 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=193391.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 10:24:17,248 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=193397.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 10:24:34,652 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.4405, 4.2160, 4.1850, 4.5720, 3.2417, 3.9961, 2.8189, 4.3583], device='cuda:0'), covar=tensor([0.1477, 0.0628, 0.0868, 0.0566, 0.1019, 0.0557, 0.1635, 0.1143], device='cuda:0'), in_proj_covar=tensor([0.0226, 0.0264, 0.0295, 0.0351, 0.0236, 0.0237, 0.0258, 0.0380], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 10:24:35,327 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=193421.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 10:24:36,271 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=2.00 vs. limit=2.0 2023-05-16 10:24:46,559 INFO [finetune.py:992] (0/2) Epoch 8, batch 7800, loss[loss=0.1573, simple_loss=0.2442, pruned_loss=0.03524, over 12239.00 frames. ], tot_loss[loss=0.1709, simple_loss=0.259, pruned_loss=0.04137, over 2386714.40 frames. ], batch size: 32, lr: 4.27e-03, grad_scale: 16.0 2023-05-16 10:24:48,014 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=193439.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 10:24:55,873 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.973e+02 2.834e+02 3.445e+02 3.968e+02 8.886e+02, threshold=6.891e+02, percent-clipped=1.0 2023-05-16 10:24:58,957 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=193454.0, num_to_drop=1, layers_to_drop={1} 2023-05-16 10:25:18,939 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=193482.0, num_to_drop=1, layers_to_drop={2} 2023-05-16 10:25:18,988 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.7987, 2.5765, 4.8870, 5.0991, 3.2254, 2.7809, 2.9546, 2.0664], device='cuda:0'), covar=tensor([0.1653, 0.3939, 0.0387, 0.0399, 0.1111, 0.2383, 0.3008, 0.5221], device='cuda:0'), in_proj_covar=tensor([0.0285, 0.0384, 0.0271, 0.0298, 0.0264, 0.0296, 0.0367, 0.0368], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-16 10:25:22,203 INFO [finetune.py:992] (0/2) Epoch 8, batch 7850, loss[loss=0.1733, simple_loss=0.2596, pruned_loss=0.04347, over 12084.00 frames. ], tot_loss[loss=0.1723, simple_loss=0.2604, pruned_loss=0.04217, over 2374263.43 frames. ], batch size: 32, lr: 4.27e-03, grad_scale: 16.0 2023-05-16 10:25:22,397 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=193487.0, num_to_drop=1, layers_to_drop={0} 2023-05-16 10:25:31,631 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.0557, 4.6604, 4.7015, 4.9350, 4.6302, 4.8517, 4.7300, 2.3784], device='cuda:0'), covar=tensor([0.0093, 0.0071, 0.0100, 0.0064, 0.0061, 0.0114, 0.0095, 0.0796], device='cuda:0'), in_proj_covar=tensor([0.0065, 0.0077, 0.0079, 0.0071, 0.0058, 0.0088, 0.0078, 0.0094], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 10:25:42,433 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=193515.0, num_to_drop=1, layers_to_drop={0} 2023-05-16 10:25:57,111 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=193535.0, num_to_drop=1, layers_to_drop={0} 2023-05-16 10:25:58,441 INFO [finetune.py:992] (0/2) Epoch 8, batch 7900, loss[loss=0.1877, simple_loss=0.2831, pruned_loss=0.04613, over 12368.00 frames. ], tot_loss[loss=0.1733, simple_loss=0.2609, pruned_loss=0.04287, over 2364888.31 frames. ], batch size: 38, lr: 4.27e-03, grad_scale: 16.0 2023-05-16 10:26:08,194 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.590e+02 2.779e+02 3.308e+02 3.802e+02 1.440e+03, threshold=6.616e+02, percent-clipped=1.0 2023-05-16 10:26:20,253 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=193567.0, num_to_drop=1, layers_to_drop={2} 2023-05-16 10:26:28,740 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.7421, 2.8783, 4.3728, 4.5156, 2.9072, 2.6811, 2.8505, 2.0655], device='cuda:0'), covar=tensor([0.1408, 0.2641, 0.0454, 0.0428, 0.1214, 0.2171, 0.2489, 0.3865], device='cuda:0'), in_proj_covar=tensor([0.0283, 0.0380, 0.0268, 0.0295, 0.0261, 0.0293, 0.0364, 0.0364], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-16 10:26:34,101 INFO [finetune.py:992] (0/2) Epoch 8, batch 7950, loss[loss=0.1757, simple_loss=0.2713, pruned_loss=0.04012, over 12139.00 frames. ], tot_loss[loss=0.1734, simple_loss=0.2614, pruned_loss=0.04275, over 2367002.03 frames. ], batch size: 34, lr: 4.27e-03, grad_scale: 16.0 2023-05-16 10:26:46,677 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.5053, 4.8915, 4.2795, 5.0789, 4.6854, 3.1516, 4.3607, 3.2821], device='cuda:0'), covar=tensor([0.0677, 0.0608, 0.1284, 0.0404, 0.1023, 0.1440, 0.1015, 0.2799], device='cuda:0'), in_proj_covar=tensor([0.0298, 0.0376, 0.0355, 0.0283, 0.0362, 0.0266, 0.0338, 0.0361], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 10:26:54,269 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=193615.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 10:27:09,222 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=2.00 vs. limit=2.0 2023-05-16 10:27:09,454 INFO [finetune.py:992] (0/2) Epoch 8, batch 8000, loss[loss=0.217, simple_loss=0.2946, pruned_loss=0.06971, over 7927.00 frames. ], tot_loss[loss=0.1743, simple_loss=0.2619, pruned_loss=0.04336, over 2358428.58 frames. ], batch size: 98, lr: 4.27e-03, grad_scale: 16.0 2023-05-16 10:27:11,720 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=193640.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 10:27:18,679 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.127e+02 2.742e+02 3.203e+02 3.870e+02 1.111e+03, threshold=6.406e+02, percent-clipped=2.0 2023-05-16 10:27:37,306 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=193676.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 10:27:46,191 INFO [finetune.py:992] (0/2) Epoch 8, batch 8050, loss[loss=0.1641, simple_loss=0.2469, pruned_loss=0.0406, over 12168.00 frames. ], tot_loss[loss=0.1737, simple_loss=0.2612, pruned_loss=0.04307, over 2364849.57 frames. ], batch size: 29, lr: 4.27e-03, grad_scale: 16.0 2023-05-16 10:27:53,494 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=193697.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 10:28:21,688 INFO [finetune.py:992] (0/2) Epoch 8, batch 8100, loss[loss=0.1824, simple_loss=0.272, pruned_loss=0.0464, over 12160.00 frames. ], tot_loss[loss=0.1744, simple_loss=0.2621, pruned_loss=0.04339, over 2362047.21 frames. ], batch size: 34, lr: 4.27e-03, grad_scale: 16.0 2023-05-16 10:28:27,493 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=193745.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 10:28:30,941 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.047e+02 2.724e+02 3.178e+02 4.113e+02 8.569e+02, threshold=6.357e+02, percent-clipped=4.0 2023-05-16 10:28:38,977 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.2306, 6.1178, 5.6430, 5.6675, 6.1425, 5.5085, 5.7925, 5.6913], device='cuda:0'), covar=tensor([0.1419, 0.0849, 0.1057, 0.1743, 0.1060, 0.2132, 0.1605, 0.1062], device='cuda:0'), in_proj_covar=tensor([0.0351, 0.0471, 0.0379, 0.0418, 0.0449, 0.0426, 0.0378, 0.0373], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 10:28:50,270 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=193777.0, num_to_drop=1, layers_to_drop={2} 2023-05-16 10:28:53,638 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 2023-05-16 10:28:57,380 INFO [finetune.py:992] (0/2) Epoch 8, batch 8150, loss[loss=0.1694, simple_loss=0.2603, pruned_loss=0.03926, over 12085.00 frames. ], tot_loss[loss=0.1743, simple_loss=0.2623, pruned_loss=0.04316, over 2367212.05 frames. ], batch size: 32, lr: 4.27e-03, grad_scale: 16.0 2023-05-16 10:29:14,044 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=193810.0, num_to_drop=1, layers_to_drop={0} 2023-05-16 10:29:33,671 INFO [finetune.py:992] (0/2) Epoch 8, batch 8200, loss[loss=0.1589, simple_loss=0.2465, pruned_loss=0.03571, over 12098.00 frames. ], tot_loss[loss=0.1738, simple_loss=0.2619, pruned_loss=0.04283, over 2381334.63 frames. ], batch size: 32, lr: 4.26e-03, grad_scale: 16.0 2023-05-16 10:29:42,870 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.208e+02 2.824e+02 3.337e+02 3.924e+02 1.543e+03, threshold=6.675e+02, percent-clipped=2.0 2023-05-16 10:29:54,973 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=193867.0, num_to_drop=1, layers_to_drop={0} 2023-05-16 10:30:06,830 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=193884.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 10:30:08,858 INFO [finetune.py:992] (0/2) Epoch 8, batch 8250, loss[loss=0.1879, simple_loss=0.2782, pruned_loss=0.04882, over 12113.00 frames. ], tot_loss[loss=0.1732, simple_loss=0.2616, pruned_loss=0.04244, over 2387092.62 frames. ], batch size: 42, lr: 4.26e-03, grad_scale: 16.0 2023-05-16 10:30:29,185 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=193915.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 10:30:42,078 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.5451, 2.5452, 3.4047, 4.5553, 2.2051, 4.5997, 4.5071, 4.6945], device='cuda:0'), covar=tensor([0.0124, 0.1122, 0.0349, 0.0114, 0.1268, 0.0171, 0.0134, 0.0084], device='cuda:0'), in_proj_covar=tensor([0.0119, 0.0200, 0.0178, 0.0114, 0.0187, 0.0173, 0.0171, 0.0117], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 10:30:44,562 INFO [finetune.py:992] (0/2) Epoch 8, batch 8300, loss[loss=0.2119, simple_loss=0.3044, pruned_loss=0.05971, over 12077.00 frames. ], tot_loss[loss=0.1729, simple_loss=0.2612, pruned_loss=0.04227, over 2384802.26 frames. ], batch size: 40, lr: 4.26e-03, grad_scale: 16.0 2023-05-16 10:30:45,422 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.2446, 4.8969, 5.2321, 4.4329, 4.9174, 4.5039, 5.1974, 4.9983], device='cuda:0'), covar=tensor([0.0304, 0.0369, 0.0348, 0.0335, 0.0406, 0.0372, 0.0333, 0.0312], device='cuda:0'), in_proj_covar=tensor([0.0247, 0.0256, 0.0278, 0.0253, 0.0251, 0.0251, 0.0230, 0.0208], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 10:30:46,658 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=193940.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 10:30:50,298 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=193945.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 10:30:53,666 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.139e+02 2.837e+02 3.419e+02 4.022e+02 1.405e+03, threshold=6.839e+02, percent-clipped=4.0 2023-05-16 10:30:53,944 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.5945, 3.6172, 3.2984, 3.2956, 2.8808, 2.8078, 3.6861, 2.4385], device='cuda:0'), covar=tensor([0.0330, 0.0131, 0.0157, 0.0162, 0.0381, 0.0314, 0.0122, 0.0405], device='cuda:0'), in_proj_covar=tensor([0.0189, 0.0160, 0.0154, 0.0181, 0.0200, 0.0197, 0.0166, 0.0197], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 10:31:08,585 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=193971.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 10:31:21,239 INFO [finetune.py:992] (0/2) Epoch 8, batch 8350, loss[loss=0.1677, simple_loss=0.2622, pruned_loss=0.03663, over 12181.00 frames. ], tot_loss[loss=0.1729, simple_loss=0.2611, pruned_loss=0.04234, over 2380171.72 frames. ], batch size: 35, lr: 4.26e-03, grad_scale: 16.0 2023-05-16 10:31:22,036 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=193988.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 10:31:30,981 INFO [checkpoint.py:75] (0/2) Saving checkpoint to pruned_transducer_stateless7/exp_giga_finetune/checkpoint-94000.pt 2023-05-16 10:31:51,293 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.2675, 4.8087, 5.0880, 5.1026, 4.8815, 5.1135, 4.9638, 2.6189], device='cuda:0'), covar=tensor([0.0092, 0.0071, 0.0066, 0.0066, 0.0046, 0.0097, 0.0081, 0.0744], device='cuda:0'), in_proj_covar=tensor([0.0066, 0.0077, 0.0079, 0.0072, 0.0058, 0.0089, 0.0078, 0.0095], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 10:31:59,570 INFO [finetune.py:992] (0/2) Epoch 8, batch 8400, loss[loss=0.1859, simple_loss=0.268, pruned_loss=0.05187, over 11482.00 frames. ], tot_loss[loss=0.1722, simple_loss=0.2604, pruned_loss=0.04202, over 2383302.57 frames. ], batch size: 25, lr: 4.26e-03, grad_scale: 16.0 2023-05-16 10:32:08,007 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.85 vs. limit=2.0 2023-05-16 10:32:08,740 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.068e+02 2.838e+02 3.415e+02 4.165e+02 1.425e+03, threshold=6.831e+02, percent-clipped=3.0 2023-05-16 10:32:24,550 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=194072.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 10:32:28,006 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=194077.0, num_to_drop=1, layers_to_drop={2} 2023-05-16 10:32:34,900 INFO [finetune.py:992] (0/2) Epoch 8, batch 8450, loss[loss=0.156, simple_loss=0.2391, pruned_loss=0.0364, over 12166.00 frames. ], tot_loss[loss=0.172, simple_loss=0.2605, pruned_loss=0.04173, over 2386186.81 frames. ], batch size: 29, lr: 4.26e-03, grad_scale: 16.0 2023-05-16 10:32:51,282 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=194110.0, num_to_drop=1, layers_to_drop={2} 2023-05-16 10:33:03,252 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=194125.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 10:33:09,213 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=194133.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 10:33:11,805 INFO [finetune.py:992] (0/2) Epoch 8, batch 8500, loss[loss=0.1687, simple_loss=0.2585, pruned_loss=0.03947, over 12276.00 frames. ], tot_loss[loss=0.1713, simple_loss=0.2597, pruned_loss=0.04147, over 2384215.34 frames. ], batch size: 37, lr: 4.26e-03, grad_scale: 32.0 2023-05-16 10:33:20,968 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.640e+02 2.571e+02 3.045e+02 3.848e+02 1.070e+03, threshold=6.091e+02, percent-clipped=1.0 2023-05-16 10:33:26,802 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=194158.0, num_to_drop=1, layers_to_drop={1} 2023-05-16 10:33:46,928 INFO [finetune.py:992] (0/2) Epoch 8, batch 8550, loss[loss=0.184, simple_loss=0.2759, pruned_loss=0.04607, over 11874.00 frames. ], tot_loss[loss=0.172, simple_loss=0.2601, pruned_loss=0.04194, over 2370240.44 frames. ], batch size: 44, lr: 4.26e-03, grad_scale: 32.0 2023-05-16 10:33:57,785 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=4.95 vs. limit=5.0 2023-05-16 10:34:21,778 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.6253, 2.7756, 3.7107, 4.7264, 4.0020, 4.6312, 4.0288, 3.2767], device='cuda:0'), covar=tensor([0.0033, 0.0342, 0.0128, 0.0029, 0.0100, 0.0079, 0.0100, 0.0348], device='cuda:0'), in_proj_covar=tensor([0.0086, 0.0121, 0.0104, 0.0076, 0.0100, 0.0118, 0.0092, 0.0139], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 10:34:21,806 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.4682, 4.8677, 3.0952, 2.9131, 4.1768, 2.5603, 4.1620, 3.3379], device='cuda:0'), covar=tensor([0.0605, 0.0475, 0.1033, 0.1237, 0.0248, 0.1320, 0.0468, 0.0723], device='cuda:0'), in_proj_covar=tensor([0.0188, 0.0252, 0.0174, 0.0197, 0.0139, 0.0178, 0.0196, 0.0174], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 10:34:23,073 INFO [finetune.py:992] (0/2) Epoch 8, batch 8600, loss[loss=0.1579, simple_loss=0.2511, pruned_loss=0.03233, over 12357.00 frames. ], tot_loss[loss=0.1707, simple_loss=0.2589, pruned_loss=0.04125, over 2376195.95 frames. ], batch size: 35, lr: 4.26e-03, grad_scale: 16.0 2023-05-16 10:34:25,252 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=194240.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 10:34:32,991 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.886e+02 2.656e+02 3.272e+02 4.250e+02 7.840e+02, threshold=6.543e+02, percent-clipped=3.0 2023-05-16 10:34:48,810 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=194271.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 10:34:51,819 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.8098, 2.3338, 3.2680, 2.8117, 3.1708, 2.9649, 2.3235, 3.2654], device='cuda:0'), covar=tensor([0.0126, 0.0367, 0.0203, 0.0204, 0.0142, 0.0176, 0.0349, 0.0132], device='cuda:0'), in_proj_covar=tensor([0.0170, 0.0201, 0.0182, 0.0180, 0.0208, 0.0157, 0.0193, 0.0177], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 10:35:00,259 INFO [finetune.py:992] (0/2) Epoch 8, batch 8650, loss[loss=0.1735, simple_loss=0.2712, pruned_loss=0.03794, over 12046.00 frames. ], tot_loss[loss=0.1708, simple_loss=0.2592, pruned_loss=0.04114, over 2375689.51 frames. ], batch size: 40, lr: 4.26e-03, grad_scale: 16.0 2023-05-16 10:35:21,998 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=194318.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 10:35:22,593 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=194319.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 10:35:35,220 INFO [finetune.py:992] (0/2) Epoch 8, batch 8700, loss[loss=0.1443, simple_loss=0.2371, pruned_loss=0.02577, over 12104.00 frames. ], tot_loss[loss=0.1716, simple_loss=0.2599, pruned_loss=0.04164, over 2380034.49 frames. ], batch size: 32, lr: 4.26e-03, grad_scale: 16.0 2023-05-16 10:35:45,189 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.978e+02 2.865e+02 3.268e+02 3.963e+02 7.088e+02, threshold=6.536e+02, percent-clipped=1.0 2023-05-16 10:35:46,814 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=194353.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 10:35:53,383 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=194362.0, num_to_drop=1, layers_to_drop={0} 2023-05-16 10:36:05,528 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=194379.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 10:36:11,049 INFO [finetune.py:992] (0/2) Epoch 8, batch 8750, loss[loss=0.1955, simple_loss=0.286, pruned_loss=0.05253, over 12074.00 frames. ], tot_loss[loss=0.1719, simple_loss=0.2602, pruned_loss=0.04186, over 2378495.09 frames. ], batch size: 42, lr: 4.26e-03, grad_scale: 16.0 2023-05-16 10:36:22,688 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.9611, 5.9713, 5.7393, 5.1580, 5.0524, 5.8160, 5.4591, 5.2244], device='cuda:0'), covar=tensor([0.0756, 0.0855, 0.0655, 0.1699, 0.0749, 0.0714, 0.1478, 0.1044], device='cuda:0'), in_proj_covar=tensor([0.0592, 0.0533, 0.0498, 0.0618, 0.0402, 0.0695, 0.0750, 0.0554], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0004, 0.0004, 0.0003], device='cuda:0') 2023-05-16 10:36:29,459 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=4.66 vs. limit=5.0 2023-05-16 10:36:32,063 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=194414.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 10:36:38,606 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=194423.0, num_to_drop=1, layers_to_drop={3} 2023-05-16 10:36:42,124 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=194428.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 10:36:48,327 INFO [finetune.py:992] (0/2) Epoch 8, batch 8800, loss[loss=0.275, simple_loss=0.3394, pruned_loss=0.1053, over 8296.00 frames. ], tot_loss[loss=0.1725, simple_loss=0.2605, pruned_loss=0.0422, over 2374379.12 frames. ], batch size: 99, lr: 4.26e-03, grad_scale: 16.0 2023-05-16 10:36:58,333 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.852e+02 2.631e+02 3.233e+02 4.193e+02 6.169e+02, threshold=6.466e+02, percent-clipped=0.0 2023-05-16 10:37:23,948 INFO [finetune.py:992] (0/2) Epoch 8, batch 8850, loss[loss=0.1674, simple_loss=0.2609, pruned_loss=0.03691, over 12128.00 frames. ], tot_loss[loss=0.1736, simple_loss=0.2617, pruned_loss=0.04271, over 2369557.91 frames. ], batch size: 39, lr: 4.26e-03, grad_scale: 16.0 2023-05-16 10:37:59,263 INFO [finetune.py:992] (0/2) Epoch 8, batch 8900, loss[loss=0.1945, simple_loss=0.2834, pruned_loss=0.05281, over 12074.00 frames. ], tot_loss[loss=0.1728, simple_loss=0.2608, pruned_loss=0.04237, over 2366543.77 frames. ], batch size: 42, lr: 4.26e-03, grad_scale: 16.0 2023-05-16 10:38:02,185 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=194540.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 10:38:10,023 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.855e+02 2.719e+02 3.310e+02 4.071e+02 9.706e+02, threshold=6.620e+02, percent-clipped=3.0 2023-05-16 10:38:36,139 INFO [finetune.py:992] (0/2) Epoch 8, batch 8950, loss[loss=0.1723, simple_loss=0.2652, pruned_loss=0.0397, over 12052.00 frames. ], tot_loss[loss=0.1721, simple_loss=0.2602, pruned_loss=0.04203, over 2372793.64 frames. ], batch size: 40, lr: 4.26e-03, grad_scale: 16.0 2023-05-16 10:38:36,936 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=194588.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 10:39:09,761 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=194634.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 10:39:11,522 INFO [finetune.py:992] (0/2) Epoch 8, batch 9000, loss[loss=0.1783, simple_loss=0.2626, pruned_loss=0.04697, over 12021.00 frames. ], tot_loss[loss=0.1739, simple_loss=0.2621, pruned_loss=0.04286, over 2365301.23 frames. ], batch size: 40, lr: 4.26e-03, grad_scale: 16.0 2023-05-16 10:39:11,523 INFO [finetune.py:1017] (0/2) Computing validation loss 2023-05-16 10:39:29,678 INFO [finetune.py:1026] (0/2) Epoch 8, validation: loss=0.3234, simple_loss=0.3994, pruned_loss=0.1237, over 1020973.00 frames. 2023-05-16 10:39:29,679 INFO [finetune.py:1027] (0/2) Maximum memory allocated so far is 12761MB 2023-05-16 10:39:40,954 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.927e+02 2.865e+02 3.334e+02 3.958e+02 1.099e+03, threshold=6.669e+02, percent-clipped=1.0 2023-05-16 10:39:57,598 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=194674.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 10:40:06,666 INFO [finetune.py:992] (0/2) Epoch 8, batch 9050, loss[loss=0.1563, simple_loss=0.2499, pruned_loss=0.03139, over 12367.00 frames. ], tot_loss[loss=0.1747, simple_loss=0.2628, pruned_loss=0.04327, over 2354581.29 frames. ], batch size: 35, lr: 4.26e-03, grad_scale: 16.0 2023-05-16 10:40:12,599 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=194695.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 10:40:17,478 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.3105, 2.4138, 3.1717, 4.1969, 2.0643, 4.2763, 4.3137, 4.3872], device='cuda:0'), covar=tensor([0.0173, 0.1293, 0.0537, 0.0199, 0.1457, 0.0251, 0.0199, 0.0122], device='cuda:0'), in_proj_covar=tensor([0.0119, 0.0203, 0.0180, 0.0115, 0.0187, 0.0174, 0.0172, 0.0118], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 10:40:20,442 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.7203, 2.8445, 4.5858, 4.8456, 2.8305, 2.7283, 3.0395, 2.1254], device='cuda:0'), covar=tensor([0.1460, 0.2865, 0.0439, 0.0359, 0.1277, 0.2185, 0.2599, 0.4001], device='cuda:0'), in_proj_covar=tensor([0.0282, 0.0375, 0.0265, 0.0293, 0.0259, 0.0291, 0.0362, 0.0361], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-16 10:40:21,980 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-05-16 10:40:22,407 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=194709.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 10:40:28,791 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=194718.0, num_to_drop=1, layers_to_drop={3} 2023-05-16 10:40:35,956 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=194728.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 10:40:42,196 INFO [finetune.py:992] (0/2) Epoch 8, batch 9100, loss[loss=0.1869, simple_loss=0.2822, pruned_loss=0.04573, over 12363.00 frames. ], tot_loss[loss=0.1751, simple_loss=0.2631, pruned_loss=0.04358, over 2351542.67 frames. ], batch size: 36, lr: 4.25e-03, grad_scale: 16.0 2023-05-16 10:40:44,607 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.5536, 3.5644, 3.2441, 3.1894, 2.9548, 2.8262, 3.6788, 2.4657], device='cuda:0'), covar=tensor([0.0370, 0.0192, 0.0178, 0.0211, 0.0363, 0.0329, 0.0130, 0.0426], device='cuda:0'), in_proj_covar=tensor([0.0191, 0.0161, 0.0156, 0.0183, 0.0202, 0.0200, 0.0167, 0.0199], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 10:40:52,206 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.833e+02 2.759e+02 3.384e+02 3.859e+02 6.643e+02, threshold=6.768e+02, percent-clipped=0.0 2023-05-16 10:41:10,293 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=194776.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 10:41:18,665 INFO [finetune.py:992] (0/2) Epoch 8, batch 9150, loss[loss=0.2894, simple_loss=0.342, pruned_loss=0.1184, over 7868.00 frames. ], tot_loss[loss=0.1754, simple_loss=0.2633, pruned_loss=0.04377, over 2349411.85 frames. ], batch size: 97, lr: 4.25e-03, grad_scale: 16.0 2023-05-16 10:41:40,858 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([6.0103, 6.0134, 5.7288, 5.2740, 5.1464, 5.8655, 5.4857, 5.2948], device='cuda:0'), covar=tensor([0.0732, 0.0770, 0.0691, 0.1639, 0.0748, 0.0753, 0.1553, 0.0943], device='cuda:0'), in_proj_covar=tensor([0.0587, 0.0527, 0.0494, 0.0612, 0.0398, 0.0688, 0.0749, 0.0550], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0004, 0.0004, 0.0003], device='cuda:0') 2023-05-16 10:41:54,878 INFO [finetune.py:992] (0/2) Epoch 8, batch 9200, loss[loss=0.1834, simple_loss=0.278, pruned_loss=0.04441, over 12154.00 frames. ], tot_loss[loss=0.1752, simple_loss=0.2634, pruned_loss=0.04348, over 2356668.49 frames. ], batch size: 36, lr: 4.25e-03, grad_scale: 16.0 2023-05-16 10:42:00,263 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.60 vs. limit=2.0 2023-05-16 10:42:02,207 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=194847.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 10:42:04,899 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.977e+02 2.634e+02 3.159e+02 3.699e+02 5.868e+02, threshold=6.317e+02, percent-clipped=0.0 2023-05-16 10:42:30,435 INFO [finetune.py:992] (0/2) Epoch 8, batch 9250, loss[loss=0.1757, simple_loss=0.2589, pruned_loss=0.04629, over 12252.00 frames. ], tot_loss[loss=0.175, simple_loss=0.2633, pruned_loss=0.0434, over 2359852.73 frames. ], batch size: 32, lr: 4.25e-03, grad_scale: 16.0 2023-05-16 10:42:45,482 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=194908.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 10:42:47,094 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=2.71 vs. limit=5.0 2023-05-16 10:42:51,804 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.9231, 3.9340, 3.9160, 4.0109, 3.7680, 3.7938, 3.7214, 3.9031], device='cuda:0'), covar=tensor([0.1161, 0.0662, 0.1374, 0.0648, 0.1514, 0.1236, 0.0537, 0.1005], device='cuda:0'), in_proj_covar=tensor([0.0518, 0.0674, 0.0585, 0.0608, 0.0828, 0.0717, 0.0533, 0.0474], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0005, 0.0004, 0.0004, 0.0005, 0.0005, 0.0004, 0.0003], device='cuda:0') 2023-05-16 10:43:02,991 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.6519, 4.2032, 4.3868, 4.5586, 4.3510, 4.5220, 4.3843, 2.7069], device='cuda:0'), covar=tensor([0.0082, 0.0088, 0.0103, 0.0064, 0.0053, 0.0084, 0.0088, 0.0731], device='cuda:0'), in_proj_covar=tensor([0.0066, 0.0077, 0.0080, 0.0072, 0.0058, 0.0089, 0.0078, 0.0095], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 10:43:06,233 INFO [finetune.py:992] (0/2) Epoch 8, batch 9300, loss[loss=0.1644, simple_loss=0.2567, pruned_loss=0.03606, over 12366.00 frames. ], tot_loss[loss=0.1747, simple_loss=0.2631, pruned_loss=0.04315, over 2359016.19 frames. ], batch size: 35, lr: 4.25e-03, grad_scale: 8.0 2023-05-16 10:43:17,504 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.907e+02 2.938e+02 3.434e+02 4.053e+02 6.793e+02, threshold=6.868e+02, percent-clipped=2.0 2023-05-16 10:43:33,149 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=194974.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 10:43:42,315 INFO [finetune.py:992] (0/2) Epoch 8, batch 9350, loss[loss=0.1561, simple_loss=0.243, pruned_loss=0.03463, over 12106.00 frames. ], tot_loss[loss=0.1736, simple_loss=0.2622, pruned_loss=0.04247, over 2370455.17 frames. ], batch size: 33, lr: 4.25e-03, grad_scale: 8.0 2023-05-16 10:43:44,680 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=194990.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 10:43:45,014 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 2023-05-16 10:43:56,171 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.3736, 4.8515, 5.3277, 4.6543, 4.9083, 4.7257, 5.3505, 5.0252], device='cuda:0'), covar=tensor([0.0215, 0.0403, 0.0243, 0.0288, 0.0372, 0.0309, 0.0217, 0.0305], device='cuda:0'), in_proj_covar=tensor([0.0247, 0.0259, 0.0279, 0.0255, 0.0254, 0.0251, 0.0233, 0.0210], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 10:43:58,311 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=195009.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 10:43:59,824 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=195011.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 10:44:03,305 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.4555, 2.4341, 3.1520, 4.4251, 1.9539, 4.3333, 4.4391, 4.4767], device='cuda:0'), covar=tensor([0.0110, 0.1240, 0.0481, 0.0125, 0.1429, 0.0209, 0.0124, 0.0091], device='cuda:0'), in_proj_covar=tensor([0.0119, 0.0201, 0.0179, 0.0115, 0.0186, 0.0175, 0.0172, 0.0117], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 10:44:04,717 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=195018.0, num_to_drop=1, layers_to_drop={0} 2023-05-16 10:44:07,436 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=195022.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 10:44:12,080 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.59 vs. limit=2.0 2023-05-16 10:44:18,087 INFO [finetune.py:992] (0/2) Epoch 8, batch 9400, loss[loss=0.1736, simple_loss=0.2609, pruned_loss=0.0431, over 12140.00 frames. ], tot_loss[loss=0.1732, simple_loss=0.262, pruned_loss=0.04221, over 2376189.99 frames. ], batch size: 34, lr: 4.25e-03, grad_scale: 8.0 2023-05-16 10:44:23,293 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=195044.0, num_to_drop=1, layers_to_drop={0} 2023-05-16 10:44:28,844 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.810e+02 2.895e+02 3.272e+02 4.035e+02 1.379e+03, threshold=6.543e+02, percent-clipped=4.0 2023-05-16 10:44:32,496 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=195057.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 10:44:33,943 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.0287, 4.6094, 4.1804, 4.1498, 4.6924, 4.0937, 4.2497, 4.1184], device='cuda:0'), covar=tensor([0.1511, 0.1069, 0.1187, 0.1998, 0.1085, 0.2100, 0.1698, 0.1421], device='cuda:0'), in_proj_covar=tensor([0.0357, 0.0477, 0.0381, 0.0425, 0.0452, 0.0431, 0.0385, 0.0378], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 10:44:38,706 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=195066.0, num_to_drop=1, layers_to_drop={0} 2023-05-16 10:44:42,334 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=195071.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 10:44:43,110 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=195072.0, num_to_drop=1, layers_to_drop={2} 2023-05-16 10:44:53,761 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.4198, 2.3115, 3.1111, 4.3455, 1.9887, 4.3476, 4.3831, 4.4576], device='cuda:0'), covar=tensor([0.0087, 0.1227, 0.0473, 0.0113, 0.1413, 0.0202, 0.0128, 0.0082], device='cuda:0'), in_proj_covar=tensor([0.0118, 0.0201, 0.0179, 0.0115, 0.0186, 0.0174, 0.0171, 0.0117], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 10:44:54,960 INFO [finetune.py:992] (0/2) Epoch 8, batch 9450, loss[loss=0.1858, simple_loss=0.2811, pruned_loss=0.04528, over 12147.00 frames. ], tot_loss[loss=0.1736, simple_loss=0.2623, pruned_loss=0.04244, over 2368805.77 frames. ], batch size: 36, lr: 4.25e-03, grad_scale: 8.0 2023-05-16 10:45:08,030 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=195105.0, num_to_drop=1, layers_to_drop={3} 2023-05-16 10:45:27,065 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=195132.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 10:45:30,271 INFO [finetune.py:992] (0/2) Epoch 8, batch 9500, loss[loss=0.1571, simple_loss=0.2523, pruned_loss=0.031, over 12251.00 frames. ], tot_loss[loss=0.1724, simple_loss=0.2611, pruned_loss=0.04185, over 2376327.38 frames. ], batch size: 32, lr: 4.25e-03, grad_scale: 8.0 2023-05-16 10:45:40,850 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.759e+02 2.744e+02 3.314e+02 3.809e+02 6.264e+02, threshold=6.628e+02, percent-clipped=0.0 2023-05-16 10:46:05,307 INFO [finetune.py:992] (0/2) Epoch 8, batch 9550, loss[loss=0.1684, simple_loss=0.2582, pruned_loss=0.03928, over 12362.00 frames. ], tot_loss[loss=0.1717, simple_loss=0.2604, pruned_loss=0.04149, over 2380623.52 frames. ], batch size: 35, lr: 4.25e-03, grad_scale: 8.0 2023-05-16 10:46:16,984 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=195203.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 10:46:26,480 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.3095, 4.5322, 4.0904, 4.9561, 4.3879, 2.7102, 4.1743, 3.0275], device='cuda:0'), covar=tensor([0.0764, 0.0826, 0.1457, 0.0392, 0.1179, 0.1813, 0.0963, 0.3255], device='cuda:0'), in_proj_covar=tensor([0.0294, 0.0372, 0.0352, 0.0282, 0.0361, 0.0264, 0.0335, 0.0357], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 10:46:31,945 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.7723, 2.8726, 4.7125, 4.7604, 2.9239, 2.7030, 3.0175, 2.2160], device='cuda:0'), covar=tensor([0.1399, 0.2952, 0.0408, 0.0438, 0.1235, 0.2161, 0.2581, 0.3894], device='cuda:0'), in_proj_covar=tensor([0.0283, 0.0376, 0.0268, 0.0295, 0.0261, 0.0293, 0.0365, 0.0363], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-16 10:46:42,145 INFO [finetune.py:992] (0/2) Epoch 8, batch 9600, loss[loss=0.1777, simple_loss=0.2712, pruned_loss=0.04213, over 12312.00 frames. ], tot_loss[loss=0.1725, simple_loss=0.2612, pruned_loss=0.04193, over 2378649.18 frames. ], batch size: 34, lr: 4.25e-03, grad_scale: 8.0 2023-05-16 10:46:52,603 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.986e+02 2.821e+02 3.505e+02 4.259e+02 9.159e+02, threshold=7.010e+02, percent-clipped=6.0 2023-05-16 10:47:17,319 INFO [finetune.py:992] (0/2) Epoch 8, batch 9650, loss[loss=0.1569, simple_loss=0.2513, pruned_loss=0.03127, over 12184.00 frames. ], tot_loss[loss=0.1724, simple_loss=0.2607, pruned_loss=0.04205, over 2376735.23 frames. ], batch size: 35, lr: 4.25e-03, grad_scale: 8.0 2023-05-16 10:47:19,537 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=195290.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 10:47:53,021 INFO [finetune.py:992] (0/2) Epoch 8, batch 9700, loss[loss=0.1851, simple_loss=0.2741, pruned_loss=0.04804, over 12006.00 frames. ], tot_loss[loss=0.1728, simple_loss=0.2611, pruned_loss=0.04226, over 2376261.94 frames. ], batch size: 40, lr: 4.25e-03, grad_scale: 8.0 2023-05-16 10:47:53,817 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=195338.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 10:48:03,541 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.974e+02 2.767e+02 3.197e+02 3.864e+02 7.277e+02, threshold=6.395e+02, percent-clipped=2.0 2023-05-16 10:48:08,651 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.0201, 2.3958, 3.5692, 2.9738, 3.3817, 3.1760, 2.5292, 3.4855], device='cuda:0'), covar=tensor([0.0101, 0.0358, 0.0135, 0.0215, 0.0132, 0.0145, 0.0350, 0.0111], device='cuda:0'), in_proj_covar=tensor([0.0169, 0.0200, 0.0183, 0.0179, 0.0207, 0.0156, 0.0192, 0.0177], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 10:48:10,402 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.93 vs. limit=2.0 2023-05-16 10:48:14,993 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=195367.0, num_to_drop=1, layers_to_drop={0} 2023-05-16 10:48:24,892 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.4522, 2.4449, 3.1105, 4.4127, 2.2033, 4.3915, 4.4318, 4.4728], device='cuda:0'), covar=tensor([0.0109, 0.1193, 0.0478, 0.0134, 0.1344, 0.0209, 0.0121, 0.0092], device='cuda:0'), in_proj_covar=tensor([0.0119, 0.0200, 0.0178, 0.0115, 0.0186, 0.0173, 0.0171, 0.0117], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 10:48:29,477 INFO [finetune.py:992] (0/2) Epoch 8, batch 9750, loss[loss=0.1776, simple_loss=0.2746, pruned_loss=0.04029, over 12079.00 frames. ], tot_loss[loss=0.1728, simple_loss=0.2611, pruned_loss=0.04227, over 2377863.05 frames. ], batch size: 40, lr: 4.25e-03, grad_scale: 8.0 2023-05-16 10:48:29,863 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.31 vs. limit=2.0 2023-05-16 10:48:33,160 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=195392.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 10:48:39,064 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=195400.0, num_to_drop=1, layers_to_drop={1} 2023-05-16 10:48:58,021 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=195427.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 10:49:04,955 INFO [finetune.py:992] (0/2) Epoch 8, batch 9800, loss[loss=0.1858, simple_loss=0.2818, pruned_loss=0.0449, over 12158.00 frames. ], tot_loss[loss=0.1727, simple_loss=0.2606, pruned_loss=0.04242, over 2374838.34 frames. ], batch size: 34, lr: 4.25e-03, grad_scale: 8.0 2023-05-16 10:49:06,823 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.41 vs. limit=2.0 2023-05-16 10:49:13,767 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.82 vs. limit=2.0 2023-05-16 10:49:15,533 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.074e+02 2.933e+02 3.542e+02 4.192e+02 1.012e+03, threshold=7.085e+02, percent-clipped=3.0 2023-05-16 10:49:16,460 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=195453.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 10:49:18,993 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.88 vs. limit=2.0 2023-05-16 10:49:29,385 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.3206, 3.1069, 4.7510, 2.3295, 2.6103, 3.5010, 3.3021, 3.6690], device='cuda:0'), covar=tensor([0.0551, 0.1274, 0.0337, 0.1445, 0.2107, 0.1548, 0.1363, 0.1173], device='cuda:0'), in_proj_covar=tensor([0.0236, 0.0235, 0.0243, 0.0180, 0.0236, 0.0290, 0.0224, 0.0259], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-16 10:49:40,738 INFO [finetune.py:992] (0/2) Epoch 8, batch 9850, loss[loss=0.1439, simple_loss=0.2255, pruned_loss=0.03115, over 12023.00 frames. ], tot_loss[loss=0.1718, simple_loss=0.2599, pruned_loss=0.0418, over 2374299.82 frames. ], batch size: 28, lr: 4.25e-03, grad_scale: 8.0 2023-05-16 10:49:47,329 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.1429, 4.7837, 5.1210, 4.9945, 4.8249, 5.0260, 4.9255, 2.9688], device='cuda:0'), covar=tensor([0.0099, 0.0064, 0.0064, 0.0062, 0.0045, 0.0093, 0.0072, 0.0616], device='cuda:0'), in_proj_covar=tensor([0.0066, 0.0077, 0.0079, 0.0072, 0.0058, 0.0089, 0.0079, 0.0095], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 10:49:48,072 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.9768, 2.0832, 3.2633, 3.9410, 3.4485, 3.9209, 3.5311, 2.9527], device='cuda:0'), covar=tensor([0.0047, 0.0429, 0.0140, 0.0053, 0.0126, 0.0073, 0.0119, 0.0335], device='cuda:0'), in_proj_covar=tensor([0.0086, 0.0120, 0.0103, 0.0074, 0.0099, 0.0115, 0.0091, 0.0137], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 10:49:52,344 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=195503.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 10:50:17,744 INFO [finetune.py:992] (0/2) Epoch 8, batch 9900, loss[loss=0.1462, simple_loss=0.2258, pruned_loss=0.03331, over 12191.00 frames. ], tot_loss[loss=0.1716, simple_loss=0.2601, pruned_loss=0.04154, over 2381456.94 frames. ], batch size: 29, lr: 4.25e-03, grad_scale: 8.0 2023-05-16 10:50:27,672 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=195551.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 10:50:28,283 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.980e+02 2.769e+02 3.203e+02 3.616e+02 6.223e+02, threshold=6.405e+02, percent-clipped=0.0 2023-05-16 10:50:52,647 INFO [finetune.py:992] (0/2) Epoch 8, batch 9950, loss[loss=0.1524, simple_loss=0.236, pruned_loss=0.03444, over 11754.00 frames. ], tot_loss[loss=0.1718, simple_loss=0.2601, pruned_loss=0.04178, over 2381085.55 frames. ], batch size: 26, lr: 4.25e-03, grad_scale: 8.0 2023-05-16 10:51:28,539 INFO [finetune.py:992] (0/2) Epoch 8, batch 10000, loss[loss=0.1608, simple_loss=0.2542, pruned_loss=0.03365, over 12349.00 frames. ], tot_loss[loss=0.1715, simple_loss=0.2597, pruned_loss=0.04162, over 2384646.94 frames. ], batch size: 36, lr: 4.25e-03, grad_scale: 8.0 2023-05-16 10:51:39,700 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.964e+02 2.768e+02 3.381e+02 3.907e+02 5.631e+02, threshold=6.763e+02, percent-clipped=0.0 2023-05-16 10:51:50,215 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=195667.0, num_to_drop=1, layers_to_drop={0} 2023-05-16 10:52:04,748 INFO [finetune.py:992] (0/2) Epoch 8, batch 10050, loss[loss=0.1856, simple_loss=0.2724, pruned_loss=0.04944, over 11133.00 frames. ], tot_loss[loss=0.1726, simple_loss=0.2606, pruned_loss=0.04228, over 2378983.40 frames. ], batch size: 55, lr: 4.24e-03, grad_scale: 8.0 2023-05-16 10:52:14,175 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=195700.0, num_to_drop=1, layers_to_drop={2} 2023-05-16 10:52:24,709 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=195715.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 10:52:33,304 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=195727.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 10:52:40,245 INFO [finetune.py:992] (0/2) Epoch 8, batch 10100, loss[loss=0.1655, simple_loss=0.2499, pruned_loss=0.04061, over 12153.00 frames. ], tot_loss[loss=0.1723, simple_loss=0.2604, pruned_loss=0.04207, over 2381945.47 frames. ], batch size: 36, lr: 4.24e-03, grad_scale: 8.0 2023-05-16 10:52:47,960 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=195748.0, num_to_drop=1, layers_to_drop={0} 2023-05-16 10:52:47,973 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=195748.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 10:52:50,506 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.087e+02 2.741e+02 3.375e+02 4.030e+02 6.794e+02, threshold=6.750e+02, percent-clipped=1.0 2023-05-16 10:53:06,471 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=195775.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 10:53:15,071 INFO [finetune.py:992] (0/2) Epoch 8, batch 10150, loss[loss=0.144, simple_loss=0.2265, pruned_loss=0.03074, over 12279.00 frames. ], tot_loss[loss=0.1725, simple_loss=0.2605, pruned_loss=0.0423, over 2381729.88 frames. ], batch size: 28, lr: 4.24e-03, grad_scale: 8.0 2023-05-16 10:53:36,339 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-05-16 10:53:52,377 INFO [finetune.py:992] (0/2) Epoch 8, batch 10200, loss[loss=0.1805, simple_loss=0.2725, pruned_loss=0.04427, over 12041.00 frames. ], tot_loss[loss=0.1726, simple_loss=0.2604, pruned_loss=0.04235, over 2370545.32 frames. ], batch size: 42, lr: 4.24e-03, grad_scale: 8.0 2023-05-16 10:53:57,955 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.96 vs. limit=2.0 2023-05-16 10:54:03,059 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.763e+02 2.806e+02 3.238e+02 4.023e+02 8.056e+02, threshold=6.475e+02, percent-clipped=2.0 2023-05-16 10:54:18,230 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 2023-05-16 10:54:27,946 INFO [finetune.py:992] (0/2) Epoch 8, batch 10250, loss[loss=0.1418, simple_loss=0.2233, pruned_loss=0.03011, over 11993.00 frames. ], tot_loss[loss=0.1723, simple_loss=0.2606, pruned_loss=0.04201, over 2373965.88 frames. ], batch size: 28, lr: 4.24e-03, grad_scale: 8.0 2023-05-16 10:55:03,420 INFO [finetune.py:992] (0/2) Epoch 8, batch 10300, loss[loss=0.1519, simple_loss=0.2382, pruned_loss=0.03279, over 12344.00 frames. ], tot_loss[loss=0.1708, simple_loss=0.2591, pruned_loss=0.04125, over 2377314.06 frames. ], batch size: 31, lr: 4.24e-03, grad_scale: 8.0 2023-05-16 10:55:14,639 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.639e+02 2.764e+02 3.180e+02 3.731e+02 8.237e+02, threshold=6.360e+02, percent-clipped=1.0 2023-05-16 10:55:36,560 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.5815, 5.0504, 5.4856, 4.8103, 5.0829, 4.9134, 5.5544, 5.1061], device='cuda:0'), covar=tensor([0.0186, 0.0333, 0.0257, 0.0230, 0.0372, 0.0257, 0.0192, 0.0220], device='cuda:0'), in_proj_covar=tensor([0.0247, 0.0258, 0.0280, 0.0253, 0.0254, 0.0252, 0.0230, 0.0209], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 10:55:40,012 INFO [finetune.py:992] (0/2) Epoch 8, batch 10350, loss[loss=0.1561, simple_loss=0.2488, pruned_loss=0.03173, over 12116.00 frames. ], tot_loss[loss=0.1722, simple_loss=0.2604, pruned_loss=0.04196, over 2368806.33 frames. ], batch size: 38, lr: 4.24e-03, grad_scale: 8.0 2023-05-16 10:55:40,226 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.6237, 2.5239, 3.6012, 4.6366, 3.9572, 4.5987, 4.0130, 3.3859], device='cuda:0'), covar=tensor([0.0028, 0.0354, 0.0131, 0.0028, 0.0101, 0.0059, 0.0094, 0.0287], device='cuda:0'), in_proj_covar=tensor([0.0086, 0.0120, 0.0104, 0.0074, 0.0099, 0.0116, 0.0092, 0.0138], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 10:55:47,446 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.4156, 4.7969, 3.0337, 2.7511, 4.0861, 2.4068, 4.0372, 3.2943], device='cuda:0'), covar=tensor([0.0634, 0.0394, 0.0986, 0.1379, 0.0289, 0.1547, 0.0448, 0.0787], device='cuda:0'), in_proj_covar=tensor([0.0186, 0.0249, 0.0172, 0.0195, 0.0138, 0.0177, 0.0193, 0.0172], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 10:55:49,769 INFO [checkpoint.py:75] (0/2) Saving checkpoint to pruned_transducer_stateless7/exp_giga_finetune/checkpoint-96000.pt 2023-05-16 10:56:03,099 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.4450, 2.3338, 3.5440, 4.4634, 3.8578, 4.4530, 3.8484, 3.0165], device='cuda:0'), covar=tensor([0.0034, 0.0410, 0.0123, 0.0028, 0.0092, 0.0066, 0.0104, 0.0385], device='cuda:0'), in_proj_covar=tensor([0.0086, 0.0121, 0.0104, 0.0074, 0.0099, 0.0116, 0.0092, 0.0138], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 10:56:12,835 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.6055, 4.4729, 4.5614, 4.6100, 4.3033, 4.3256, 4.1675, 4.5049], device='cuda:0'), covar=tensor([0.0740, 0.0672, 0.0843, 0.0596, 0.1884, 0.1342, 0.0659, 0.1107], device='cuda:0'), in_proj_covar=tensor([0.0524, 0.0680, 0.0591, 0.0609, 0.0831, 0.0718, 0.0536, 0.0478], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0005, 0.0004, 0.0004, 0.0005, 0.0005, 0.0004, 0.0003], device='cuda:0') 2023-05-16 10:56:18,472 INFO [finetune.py:992] (0/2) Epoch 8, batch 10400, loss[loss=0.1859, simple_loss=0.2815, pruned_loss=0.04514, over 12276.00 frames. ], tot_loss[loss=0.1727, simple_loss=0.2609, pruned_loss=0.04224, over 2363978.41 frames. ], batch size: 37, lr: 4.24e-03, grad_scale: 8.0 2023-05-16 10:56:26,273 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=196048.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 10:56:27,056 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.0133, 3.9058, 3.9171, 4.3243, 2.8749, 3.8820, 2.5226, 3.9247], device='cuda:0'), covar=tensor([0.1809, 0.0833, 0.1026, 0.0573, 0.1258, 0.0659, 0.1955, 0.1135], device='cuda:0'), in_proj_covar=tensor([0.0226, 0.0266, 0.0296, 0.0356, 0.0236, 0.0236, 0.0257, 0.0379], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 10:56:28,900 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.869e+02 2.918e+02 3.403e+02 4.012e+02 6.635e+02, threshold=6.805e+02, percent-clipped=1.0 2023-05-16 10:56:54,489 INFO [finetune.py:992] (0/2) Epoch 8, batch 10450, loss[loss=0.1821, simple_loss=0.2718, pruned_loss=0.04615, over 12152.00 frames. ], tot_loss[loss=0.1718, simple_loss=0.2605, pruned_loss=0.04156, over 2372352.35 frames. ], batch size: 34, lr: 4.24e-03, grad_scale: 8.0 2023-05-16 10:57:00,851 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=196096.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 10:57:18,520 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.1484, 4.9858, 5.1268, 5.1450, 4.8139, 4.8170, 4.5915, 5.0827], device='cuda:0'), covar=tensor([0.0769, 0.0563, 0.0715, 0.0585, 0.1693, 0.1305, 0.0590, 0.1063], device='cuda:0'), in_proj_covar=tensor([0.0527, 0.0684, 0.0592, 0.0611, 0.0833, 0.0722, 0.0538, 0.0481], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0005, 0.0004, 0.0004, 0.0005, 0.0005, 0.0004, 0.0004], device='cuda:0') 2023-05-16 10:57:30,577 INFO [finetune.py:992] (0/2) Epoch 8, batch 10500, loss[loss=0.1476, simple_loss=0.2335, pruned_loss=0.03082, over 12265.00 frames. ], tot_loss[loss=0.1723, simple_loss=0.2606, pruned_loss=0.04198, over 2372137.69 frames. ], batch size: 28, lr: 4.24e-03, grad_scale: 8.0 2023-05-16 10:57:41,283 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.887e+02 2.777e+02 3.207e+02 3.747e+02 7.521e+02, threshold=6.415e+02, percent-clipped=1.0 2023-05-16 10:57:43,316 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.38 vs. limit=2.0 2023-05-16 10:58:06,363 INFO [finetune.py:992] (0/2) Epoch 8, batch 10550, loss[loss=0.1847, simple_loss=0.2752, pruned_loss=0.04712, over 11124.00 frames. ], tot_loss[loss=0.1714, simple_loss=0.2598, pruned_loss=0.04149, over 2379325.05 frames. ], batch size: 55, lr: 4.24e-03, grad_scale: 8.0 2023-05-16 10:58:42,561 INFO [finetune.py:992] (0/2) Epoch 8, batch 10600, loss[loss=0.1884, simple_loss=0.2805, pruned_loss=0.04812, over 12367.00 frames. ], tot_loss[loss=0.1711, simple_loss=0.2593, pruned_loss=0.0415, over 2379005.51 frames. ], batch size: 35, lr: 4.24e-03, grad_scale: 8.0 2023-05-16 10:58:43,517 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.8011, 3.2351, 5.1782, 2.6698, 2.9168, 3.9075, 3.2729, 3.8019], device='cuda:0'), covar=tensor([0.0451, 0.1215, 0.0277, 0.1103, 0.1747, 0.1455, 0.1318, 0.1113], device='cuda:0'), in_proj_covar=tensor([0.0237, 0.0235, 0.0244, 0.0180, 0.0237, 0.0291, 0.0224, 0.0260], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-16 10:58:53,130 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.944e+02 2.800e+02 3.385e+02 4.224e+02 7.673e+02, threshold=6.769e+02, percent-clipped=3.0 2023-05-16 10:59:11,881 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.3259, 5.1481, 5.2621, 5.3103, 4.9099, 4.9024, 4.7492, 5.2245], device='cuda:0'), covar=tensor([0.0789, 0.0613, 0.0852, 0.0595, 0.1960, 0.1579, 0.0593, 0.1063], device='cuda:0'), in_proj_covar=tensor([0.0521, 0.0680, 0.0589, 0.0609, 0.0832, 0.0721, 0.0535, 0.0480], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0005, 0.0004, 0.0004, 0.0005, 0.0005, 0.0004, 0.0003], device='cuda:0') 2023-05-16 10:59:16,333 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.3795, 4.6865, 4.2707, 4.8988, 4.5628, 2.9502, 4.4396, 3.1001], device='cuda:0'), covar=tensor([0.0696, 0.0765, 0.1227, 0.0528, 0.1069, 0.1535, 0.1008, 0.3113], device='cuda:0'), in_proj_covar=tensor([0.0298, 0.0375, 0.0354, 0.0286, 0.0362, 0.0266, 0.0338, 0.0360], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 10:59:18,305 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.8363, 3.4669, 5.1298, 2.5007, 2.7301, 3.7581, 3.3530, 3.7305], device='cuda:0'), covar=tensor([0.0404, 0.1089, 0.0324, 0.1335, 0.2049, 0.1580, 0.1376, 0.1253], device='cuda:0'), in_proj_covar=tensor([0.0237, 0.0234, 0.0244, 0.0179, 0.0236, 0.0290, 0.0224, 0.0259], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-16 10:59:18,770 INFO [finetune.py:992] (0/2) Epoch 8, batch 10650, loss[loss=0.1724, simple_loss=0.2613, pruned_loss=0.04173, over 12293.00 frames. ], tot_loss[loss=0.1723, simple_loss=0.2604, pruned_loss=0.04213, over 2371842.16 frames. ], batch size: 34, lr: 4.24e-03, grad_scale: 8.0 2023-05-16 10:59:27,409 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 2023-05-16 10:59:42,630 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=196321.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 10:59:53,855 INFO [finetune.py:992] (0/2) Epoch 8, batch 10700, loss[loss=0.1448, simple_loss=0.2322, pruned_loss=0.02871, over 12127.00 frames. ], tot_loss[loss=0.1715, simple_loss=0.2593, pruned_loss=0.04179, over 2374491.83 frames. ], batch size: 30, lr: 4.24e-03, grad_scale: 8.0 2023-05-16 11:00:04,520 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.938e+02 2.733e+02 3.369e+02 4.182e+02 1.101e+03, threshold=6.737e+02, percent-clipped=2.0 2023-05-16 11:00:10,445 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=196360.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 11:00:26,797 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=196382.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 11:00:30,221 INFO [finetune.py:992] (0/2) Epoch 8, batch 10750, loss[loss=0.1664, simple_loss=0.2568, pruned_loss=0.03798, over 12297.00 frames. ], tot_loss[loss=0.1707, simple_loss=0.2585, pruned_loss=0.04145, over 2375901.48 frames. ], batch size: 37, lr: 4.24e-03, grad_scale: 8.0 2023-05-16 11:00:54,689 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=196421.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 11:01:01,432 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.35 vs. limit=2.0 2023-05-16 11:01:06,077 INFO [finetune.py:992] (0/2) Epoch 8, batch 10800, loss[loss=0.2151, simple_loss=0.2948, pruned_loss=0.0677, over 7734.00 frames. ], tot_loss[loss=0.1717, simple_loss=0.2598, pruned_loss=0.04181, over 2373841.66 frames. ], batch size: 99, lr: 4.24e-03, grad_scale: 8.0 2023-05-16 11:01:16,775 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.965e+02 2.678e+02 3.444e+02 4.139e+02 7.546e+02, threshold=6.888e+02, percent-clipped=3.0 2023-05-16 11:01:41,717 INFO [finetune.py:992] (0/2) Epoch 8, batch 10850, loss[loss=0.1536, simple_loss=0.2352, pruned_loss=0.03599, over 12290.00 frames. ], tot_loss[loss=0.1716, simple_loss=0.2599, pruned_loss=0.04167, over 2374470.83 frames. ], batch size: 28, lr: 4.24e-03, grad_scale: 8.0 2023-05-16 11:01:46,070 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.9619, 4.6376, 4.7327, 4.8318, 4.5479, 4.8011, 4.6911, 2.7045], device='cuda:0'), covar=tensor([0.0078, 0.0061, 0.0074, 0.0058, 0.0051, 0.0088, 0.0076, 0.0690], device='cuda:0'), in_proj_covar=tensor([0.0066, 0.0077, 0.0079, 0.0072, 0.0058, 0.0089, 0.0079, 0.0095], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 11:02:09,013 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.38 vs. limit=2.0 2023-05-16 11:02:18,940 INFO [finetune.py:992] (0/2) Epoch 8, batch 10900, loss[loss=0.1551, simple_loss=0.2424, pruned_loss=0.03395, over 12182.00 frames. ], tot_loss[loss=0.1723, simple_loss=0.2607, pruned_loss=0.04194, over 2376356.90 frames. ], batch size: 29, lr: 4.24e-03, grad_scale: 8.0 2023-05-16 11:02:29,325 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.316e+02 2.881e+02 3.421e+02 4.372e+02 1.882e+03, threshold=6.842e+02, percent-clipped=5.0 2023-05-16 11:02:48,869 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=196578.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 11:02:55,035 INFO [finetune.py:992] (0/2) Epoch 8, batch 10950, loss[loss=0.1797, simple_loss=0.2714, pruned_loss=0.04395, over 12319.00 frames. ], tot_loss[loss=0.1732, simple_loss=0.2617, pruned_loss=0.04237, over 2373462.72 frames. ], batch size: 34, lr: 4.24e-03, grad_scale: 8.0 2023-05-16 11:03:01,374 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.8793, 5.8825, 5.6420, 5.3260, 5.0551, 5.8328, 5.3952, 5.2216], device='cuda:0'), covar=tensor([0.0718, 0.0863, 0.0614, 0.1388, 0.0738, 0.0617, 0.1549, 0.0993], device='cuda:0'), in_proj_covar=tensor([0.0588, 0.0528, 0.0487, 0.0608, 0.0402, 0.0686, 0.0740, 0.0549], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0004, 0.0004, 0.0003], device='cuda:0') 2023-05-16 11:03:13,225 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.7475, 2.0207, 2.8404, 3.7571, 2.1614, 3.8322, 3.5939, 3.8450], device='cuda:0'), covar=tensor([0.0142, 0.1379, 0.0521, 0.0140, 0.1328, 0.0274, 0.0259, 0.0117], device='cuda:0'), in_proj_covar=tensor([0.0122, 0.0205, 0.0183, 0.0115, 0.0191, 0.0178, 0.0174, 0.0120], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 11:03:30,770 INFO [finetune.py:992] (0/2) Epoch 8, batch 11000, loss[loss=0.2174, simple_loss=0.3051, pruned_loss=0.06487, over 11656.00 frames. ], tot_loss[loss=0.1763, simple_loss=0.2647, pruned_loss=0.04395, over 2343566.80 frames. ], batch size: 48, lr: 4.23e-03, grad_scale: 8.0 2023-05-16 11:03:32,433 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=196639.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 11:03:42,017 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.811e+02 2.963e+02 3.550e+02 4.429e+02 1.214e+03, threshold=7.100e+02, percent-clipped=1.0 2023-05-16 11:03:54,290 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.8261, 2.5889, 3.5681, 3.6249, 2.8455, 2.6876, 2.6715, 2.4164], device='cuda:0'), covar=tensor([0.1156, 0.2268, 0.0575, 0.0499, 0.0949, 0.1888, 0.2372, 0.2971], device='cuda:0'), in_proj_covar=tensor([0.0285, 0.0379, 0.0271, 0.0296, 0.0262, 0.0295, 0.0367, 0.0364], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-16 11:03:58,125 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=4.89 vs. limit=5.0 2023-05-16 11:04:00,584 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=196677.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 11:04:06,031 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.83 vs. limit=2.0 2023-05-16 11:04:07,485 INFO [finetune.py:992] (0/2) Epoch 8, batch 11050, loss[loss=0.1795, simple_loss=0.2722, pruned_loss=0.04344, over 12142.00 frames. ], tot_loss[loss=0.1794, simple_loss=0.2674, pruned_loss=0.04568, over 2302315.37 frames. ], batch size: 34, lr: 4.23e-03, grad_scale: 8.0 2023-05-16 11:04:16,867 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.4683, 2.9098, 3.8325, 2.3077, 2.5869, 3.1072, 2.8864, 3.2150], device='cuda:0'), covar=tensor([0.0621, 0.1144, 0.0373, 0.1279, 0.1745, 0.1348, 0.1270, 0.1076], device='cuda:0'), in_proj_covar=tensor([0.0235, 0.0233, 0.0242, 0.0179, 0.0234, 0.0288, 0.0221, 0.0257], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-16 11:04:21,342 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.60 vs. limit=2.0 2023-05-16 11:04:28,002 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=196716.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 11:04:42,434 INFO [finetune.py:992] (0/2) Epoch 8, batch 11100, loss[loss=0.1725, simple_loss=0.2662, pruned_loss=0.03936, over 12105.00 frames. ], tot_loss[loss=0.1829, simple_loss=0.2707, pruned_loss=0.04754, over 2275295.85 frames. ], batch size: 33, lr: 4.23e-03, grad_scale: 8.0 2023-05-16 11:04:45,622 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.8787, 3.1183, 4.4557, 2.4939, 2.7086, 3.6135, 3.1414, 3.7007], device='cuda:0'), covar=tensor([0.0577, 0.1138, 0.0198, 0.1211, 0.1749, 0.1146, 0.1310, 0.0753], device='cuda:0'), in_proj_covar=tensor([0.0234, 0.0232, 0.0241, 0.0178, 0.0233, 0.0287, 0.0220, 0.0256], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-16 11:04:53,517 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.108e+02 3.403e+02 4.029e+02 4.819e+02 7.876e+02, threshold=8.057e+02, percent-clipped=4.0 2023-05-16 11:05:05,700 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.8480, 4.4586, 4.0035, 4.1433, 4.5403, 3.9636, 4.1383, 3.9729], device='cuda:0'), covar=tensor([0.1694, 0.1089, 0.1396, 0.1905, 0.1105, 0.2558, 0.1771, 0.1384], device='cuda:0'), in_proj_covar=tensor([0.0354, 0.0477, 0.0382, 0.0423, 0.0450, 0.0432, 0.0386, 0.0373], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 11:05:13,364 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=196779.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 11:05:18,881 INFO [finetune.py:992] (0/2) Epoch 8, batch 11150, loss[loss=0.1587, simple_loss=0.2457, pruned_loss=0.03584, over 12107.00 frames. ], tot_loss[loss=0.1894, simple_loss=0.2762, pruned_loss=0.05134, over 2214417.59 frames. ], batch size: 33, lr: 4.23e-03, grad_scale: 8.0 2023-05-16 11:05:54,737 INFO [finetune.py:992] (0/2) Epoch 8, batch 11200, loss[loss=0.1782, simple_loss=0.2728, pruned_loss=0.04185, over 12344.00 frames. ], tot_loss[loss=0.196, simple_loss=0.2818, pruned_loss=0.05506, over 2168315.33 frames. ], batch size: 36, lr: 4.23e-03, grad_scale: 8.0 2023-05-16 11:05:57,019 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=196840.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 11:06:05,861 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.285e+02 3.531e+02 4.319e+02 5.158e+02 1.234e+03, threshold=8.638e+02, percent-clipped=4.0 2023-05-16 11:06:07,448 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=196854.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 11:06:07,690 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.41 vs. limit=2.0 2023-05-16 11:06:26,285 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=196881.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 11:06:30,312 INFO [finetune.py:992] (0/2) Epoch 8, batch 11250, loss[loss=0.2619, simple_loss=0.341, pruned_loss=0.09143, over 6561.00 frames. ], tot_loss[loss=0.2042, simple_loss=0.289, pruned_loss=0.05966, over 2104890.25 frames. ], batch size: 98, lr: 4.23e-03, grad_scale: 8.0 2023-05-16 11:06:50,256 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=196915.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 11:06:59,833 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.7113, 3.2390, 5.2201, 2.6259, 2.8496, 4.0546, 3.4290, 4.0103], device='cuda:0'), covar=tensor([0.0448, 0.1161, 0.0191, 0.1172, 0.1772, 0.1146, 0.1169, 0.0923], device='cuda:0'), in_proj_covar=tensor([0.0229, 0.0227, 0.0235, 0.0175, 0.0228, 0.0281, 0.0215, 0.0250], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:0') 2023-05-16 11:07:03,868 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=196934.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 11:07:05,769 INFO [finetune.py:992] (0/2) Epoch 8, batch 11300, loss[loss=0.3278, simple_loss=0.3838, pruned_loss=0.1359, over 6709.00 frames. ], tot_loss[loss=0.2124, simple_loss=0.2959, pruned_loss=0.06439, over 2041162.93 frames. ], batch size: 98, lr: 4.23e-03, grad_scale: 16.0 2023-05-16 11:07:09,289 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=196942.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 11:07:15,921 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.372e+02 3.452e+02 4.076e+02 5.044e+02 8.286e+02, threshold=8.152e+02, percent-clipped=0.0 2023-05-16 11:07:34,046 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=196977.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 11:07:41,214 INFO [finetune.py:992] (0/2) Epoch 8, batch 11350, loss[loss=0.2263, simple_loss=0.2939, pruned_loss=0.0793, over 12400.00 frames. ], tot_loss[loss=0.2183, simple_loss=0.3007, pruned_loss=0.06793, over 1979303.38 frames. ], batch size: 32, lr: 4.23e-03, grad_scale: 16.0 2023-05-16 11:08:01,081 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=197016.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 11:08:02,940 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.87 vs. limit=2.0 2023-05-16 11:08:07,142 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=197025.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 11:08:13,103 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.3155, 2.7050, 3.7574, 2.3966, 2.6042, 3.2317, 2.8179, 3.1606], device='cuda:0'), covar=tensor([0.0671, 0.1220, 0.0289, 0.1503, 0.1678, 0.1203, 0.1360, 0.0948], device='cuda:0'), in_proj_covar=tensor([0.0226, 0.0225, 0.0231, 0.0174, 0.0226, 0.0278, 0.0213, 0.0247], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:0') 2023-05-16 11:08:15,533 INFO [finetune.py:992] (0/2) Epoch 8, batch 11400, loss[loss=0.1658, simple_loss=0.2496, pruned_loss=0.04097, over 12248.00 frames. ], tot_loss[loss=0.2227, simple_loss=0.3041, pruned_loss=0.07062, over 1940605.76 frames. ], batch size: 28, lr: 4.23e-03, grad_scale: 16.0 2023-05-16 11:08:25,270 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.064e+02 3.567e+02 4.120e+02 4.999e+02 1.164e+03, threshold=8.240e+02, percent-clipped=1.0 2023-05-16 11:08:34,357 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=197064.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 11:08:35,017 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.7926, 3.6744, 3.7550, 3.5809, 3.6991, 3.5503, 3.7517, 3.4454], device='cuda:0'), covar=tensor([0.0320, 0.0289, 0.0320, 0.0218, 0.0317, 0.0269, 0.0324, 0.1313], device='cuda:0'), in_proj_covar=tensor([0.0233, 0.0244, 0.0264, 0.0240, 0.0239, 0.0239, 0.0219, 0.0198], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 11:08:49,872 INFO [finetune.py:992] (0/2) Epoch 8, batch 11450, loss[loss=0.2226, simple_loss=0.3102, pruned_loss=0.06749, over 10477.00 frames. ], tot_loss[loss=0.2284, simple_loss=0.3084, pruned_loss=0.07425, over 1888409.60 frames. ], batch size: 68, lr: 4.23e-03, grad_scale: 16.0 2023-05-16 11:09:23,443 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=197135.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 11:09:24,675 INFO [finetune.py:992] (0/2) Epoch 8, batch 11500, loss[loss=0.2806, simple_loss=0.345, pruned_loss=0.1081, over 7015.00 frames. ], tot_loss[loss=0.2322, simple_loss=0.3112, pruned_loss=0.07659, over 1866591.94 frames. ], batch size: 98, lr: 4.23e-03, grad_scale: 16.0 2023-05-16 11:09:30,417 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 2023-05-16 11:09:34,835 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.491e+02 3.515e+02 4.207e+02 4.859e+02 7.657e+02, threshold=8.415e+02, percent-clipped=0.0 2023-05-16 11:10:00,158 INFO [finetune.py:992] (0/2) Epoch 8, batch 11550, loss[loss=0.2236, simple_loss=0.3161, pruned_loss=0.06556, over 10273.00 frames. ], tot_loss[loss=0.2348, simple_loss=0.313, pruned_loss=0.07836, over 1834833.14 frames. ], batch size: 68, lr: 4.23e-03, grad_scale: 16.0 2023-05-16 11:10:02,529 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=4.50 vs. limit=5.0 2023-05-16 11:10:16,005 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=197210.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 11:10:32,961 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=197234.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 11:10:34,861 INFO [finetune.py:992] (0/2) Epoch 8, batch 11600, loss[loss=0.2517, simple_loss=0.317, pruned_loss=0.09323, over 7041.00 frames. ], tot_loss[loss=0.2373, simple_loss=0.3146, pruned_loss=0.08005, over 1795744.55 frames. ], batch size: 100, lr: 4.23e-03, grad_scale: 16.0 2023-05-16 11:10:34,971 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=197237.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 11:10:35,247 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 2023-05-16 11:10:45,610 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.588e+02 3.360e+02 3.974e+02 4.737e+02 1.581e+03, threshold=7.948e+02, percent-clipped=1.0 2023-05-16 11:11:05,997 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.4746, 4.4526, 4.3419, 4.0230, 4.0700, 4.4578, 4.2152, 4.0128], device='cuda:0'), covar=tensor([0.0891, 0.0870, 0.0673, 0.1316, 0.2479, 0.0841, 0.1392, 0.1163], device='cuda:0'), in_proj_covar=tensor([0.0562, 0.0505, 0.0464, 0.0576, 0.0381, 0.0650, 0.0696, 0.0521], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002], device='cuda:0') 2023-05-16 11:11:07,460 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=197282.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 11:11:11,472 INFO [finetune.py:992] (0/2) Epoch 8, batch 11650, loss[loss=0.3304, simple_loss=0.3864, pruned_loss=0.1372, over 7268.00 frames. ], tot_loss[loss=0.2377, simple_loss=0.3146, pruned_loss=0.08042, over 1788585.81 frames. ], batch size: 97, lr: 4.23e-03, grad_scale: 16.0 2023-05-16 11:11:15,949 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.4339, 5.2365, 5.3232, 5.3811, 5.0768, 5.0499, 4.8771, 5.2601], device='cuda:0'), covar=tensor([0.0629, 0.0518, 0.0788, 0.0476, 0.1423, 0.1292, 0.0546, 0.1000], device='cuda:0'), in_proj_covar=tensor([0.0489, 0.0636, 0.0553, 0.0565, 0.0766, 0.0670, 0.0501, 0.0451], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0005, 0.0004, 0.0003, 0.0003], device='cuda:0') 2023-05-16 11:11:23,554 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.97 vs. limit=2.0 2023-05-16 11:11:36,281 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.8031, 3.7949, 3.7162, 3.8712, 3.6800, 3.6766, 3.6078, 3.7675], device='cuda:0'), covar=tensor([0.1148, 0.0695, 0.1535, 0.0708, 0.1403, 0.1268, 0.0575, 0.0984], device='cuda:0'), in_proj_covar=tensor([0.0489, 0.0637, 0.0554, 0.0564, 0.0766, 0.0670, 0.0501, 0.0452], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0005, 0.0004, 0.0003, 0.0003], device='cuda:0') 2023-05-16 11:11:45,980 INFO [finetune.py:992] (0/2) Epoch 8, batch 11700, loss[loss=0.2925, simple_loss=0.3519, pruned_loss=0.1165, over 6739.00 frames. ], tot_loss[loss=0.2381, simple_loss=0.3143, pruned_loss=0.08092, over 1762643.70 frames. ], batch size: 98, lr: 4.23e-03, grad_scale: 16.0 2023-05-16 11:11:56,330 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.471e+02 3.569e+02 3.981e+02 5.157e+02 1.224e+03, threshold=7.962e+02, percent-clipped=1.0 2023-05-16 11:12:20,624 INFO [finetune.py:992] (0/2) Epoch 8, batch 11750, loss[loss=0.1841, simple_loss=0.2761, pruned_loss=0.04599, over 11198.00 frames. ], tot_loss[loss=0.2394, simple_loss=0.315, pruned_loss=0.08196, over 1732540.52 frames. ], batch size: 55, lr: 4.23e-03, grad_scale: 16.0 2023-05-16 11:12:40,276 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=197415.0, num_to_drop=1, layers_to_drop={0} 2023-05-16 11:12:48,951 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.9070, 2.2749, 2.6491, 2.9785, 2.3088, 3.0293, 2.9409, 3.0266], device='cuda:0'), covar=tensor([0.0152, 0.0974, 0.0483, 0.0165, 0.1040, 0.0247, 0.0257, 0.0159], device='cuda:0'), in_proj_covar=tensor([0.0117, 0.0199, 0.0178, 0.0110, 0.0186, 0.0170, 0.0164, 0.0115], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 11:12:53,855 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.8577, 2.3650, 3.4491, 3.5298, 2.9526, 2.6507, 2.5638, 2.4398], device='cuda:0'), covar=tensor([0.1142, 0.3168, 0.0700, 0.0512, 0.0882, 0.2258, 0.2717, 0.4201], device='cuda:0'), in_proj_covar=tensor([0.0277, 0.0365, 0.0261, 0.0282, 0.0252, 0.0285, 0.0355, 0.0353], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-16 11:12:54,420 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=197435.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 11:12:55,550 INFO [finetune.py:992] (0/2) Epoch 8, batch 11800, loss[loss=0.2675, simple_loss=0.3404, pruned_loss=0.09727, over 6566.00 frames. ], tot_loss[loss=0.2432, simple_loss=0.3179, pruned_loss=0.08425, over 1710472.74 frames. ], batch size: 98, lr: 4.23e-03, grad_scale: 16.0 2023-05-16 11:13:05,853 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.262e+02 3.453e+02 4.425e+02 5.306e+02 1.147e+03, threshold=8.850e+02, percent-clipped=3.0 2023-05-16 11:13:23,031 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=197476.0, num_to_drop=1, layers_to_drop={1} 2023-05-16 11:13:27,572 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=197483.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 11:13:28,711 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.72 vs. limit=2.0 2023-05-16 11:13:30,158 INFO [finetune.py:992] (0/2) Epoch 8, batch 11850, loss[loss=0.2677, simple_loss=0.3314, pruned_loss=0.102, over 6520.00 frames. ], tot_loss[loss=0.2423, simple_loss=0.3181, pruned_loss=0.08326, over 1726022.48 frames. ], batch size: 98, lr: 4.23e-03, grad_scale: 16.0 2023-05-16 11:13:46,329 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=197510.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 11:13:50,340 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=197516.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 11:14:05,194 INFO [finetune.py:992] (0/2) Epoch 8, batch 11900, loss[loss=0.2201, simple_loss=0.3045, pruned_loss=0.06782, over 6609.00 frames. ], tot_loss[loss=0.2401, simple_loss=0.317, pruned_loss=0.08162, over 1722973.86 frames. ], batch size: 97, lr: 4.23e-03, grad_scale: 16.0 2023-05-16 11:14:05,389 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=197537.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 11:14:10,780 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=197545.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 11:14:16,136 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.200e+02 3.439e+02 3.960e+02 4.770e+02 1.222e+03, threshold=7.919e+02, percent-clipped=4.0 2023-05-16 11:14:20,176 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=197558.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 11:14:33,108 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=197577.0, num_to_drop=1, layers_to_drop={1} 2023-05-16 11:14:36,354 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.8154, 3.0946, 2.4483, 2.2325, 2.8538, 2.3053, 3.1108, 2.6333], device='cuda:0'), covar=tensor([0.0613, 0.0561, 0.0889, 0.1405, 0.0290, 0.1268, 0.0497, 0.0771], device='cuda:0'), in_proj_covar=tensor([0.0176, 0.0230, 0.0164, 0.0190, 0.0128, 0.0172, 0.0180, 0.0163], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 11:14:38,959 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=197585.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 11:14:40,176 INFO [finetune.py:992] (0/2) Epoch 8, batch 11950, loss[loss=0.2162, simple_loss=0.2955, pruned_loss=0.06846, over 6717.00 frames. ], tot_loss[loss=0.2351, simple_loss=0.3134, pruned_loss=0.07839, over 1716420.17 frames. ], batch size: 98, lr: 4.22e-03, grad_scale: 16.0 2023-05-16 11:14:53,530 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=197606.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 11:15:15,139 INFO [finetune.py:992] (0/2) Epoch 8, batch 12000, loss[loss=0.1933, simple_loss=0.2857, pruned_loss=0.05041, over 10019.00 frames. ], tot_loss[loss=0.2293, simple_loss=0.309, pruned_loss=0.07481, over 1710581.80 frames. ], batch size: 68, lr: 4.22e-03, grad_scale: 16.0 2023-05-16 11:15:15,139 INFO [finetune.py:1017] (0/2) Computing validation loss 2023-05-16 11:15:26,899 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([1.9709, 3.1245, 3.4076, 3.7674, 2.6392, 3.2028, 2.2574, 3.2294], device='cuda:0'), covar=tensor([0.2184, 0.1290, 0.1335, 0.0686, 0.1512, 0.1082, 0.2674, 0.1342], device='cuda:0'), in_proj_covar=tensor([0.0222, 0.0258, 0.0285, 0.0340, 0.0231, 0.0229, 0.0255, 0.0364], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 11:15:34,456 INFO [finetune.py:1026] (0/2) Epoch 8, validation: loss=0.2946, simple_loss=0.3674, pruned_loss=0.1109, over 1020973.00 frames. 2023-05-16 11:15:34,457 INFO [finetune.py:1027] (0/2) Maximum memory allocated so far is 12761MB 2023-05-16 11:15:44,914 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.171e+02 3.035e+02 3.587e+02 4.219e+02 7.957e+02, threshold=7.174e+02, percent-clipped=2.0 2023-05-16 11:16:09,199 INFO [finetune.py:992] (0/2) Epoch 8, batch 12050, loss[loss=0.2092, simple_loss=0.2872, pruned_loss=0.06561, over 6884.00 frames. ], tot_loss[loss=0.2237, simple_loss=0.3045, pruned_loss=0.07149, over 1713267.36 frames. ], batch size: 98, lr: 4.22e-03, grad_scale: 16.0 2023-05-16 11:16:42,037 INFO [finetune.py:992] (0/2) Epoch 8, batch 12100, loss[loss=0.2457, simple_loss=0.3089, pruned_loss=0.09127, over 7017.00 frames. ], tot_loss[loss=0.2235, simple_loss=0.3042, pruned_loss=0.07143, over 1708322.17 frames. ], batch size: 100, lr: 4.22e-03, grad_scale: 16.0 2023-05-16 11:16:44,085 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.7495, 3.6408, 3.6418, 3.7537, 3.4584, 3.7808, 3.7775, 3.8265], device='cuda:0'), covar=tensor([0.0210, 0.0164, 0.0166, 0.0258, 0.0534, 0.0273, 0.0166, 0.0231], device='cuda:0'), in_proj_covar=tensor([0.0165, 0.0164, 0.0159, 0.0205, 0.0204, 0.0182, 0.0147, 0.0194], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:0') 2023-05-16 11:16:51,590 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.093e+02 3.331e+02 3.770e+02 4.406e+02 9.373e+02, threshold=7.541e+02, percent-clipped=1.0 2023-05-16 11:16:56,866 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.0196, 2.2278, 2.6567, 3.0813, 2.2744, 3.1227, 3.0126, 3.1344], device='cuda:0'), covar=tensor([0.0140, 0.1029, 0.0457, 0.0156, 0.1073, 0.0250, 0.0267, 0.0144], device='cuda:0'), in_proj_covar=tensor([0.0114, 0.0195, 0.0174, 0.0108, 0.0183, 0.0166, 0.0160, 0.0112], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 11:17:03,877 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=197771.0, num_to_drop=1, layers_to_drop={2} 2023-05-16 11:17:07,722 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.7047, 3.5403, 3.5472, 3.6857, 3.5736, 3.7449, 3.6370, 2.5871], device='cuda:0'), covar=tensor([0.0094, 0.0091, 0.0141, 0.0083, 0.0073, 0.0108, 0.0084, 0.0776], device='cuda:0'), in_proj_covar=tensor([0.0060, 0.0070, 0.0073, 0.0066, 0.0053, 0.0081, 0.0071, 0.0089], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-16 11:17:13,970 INFO [finetune.py:992] (0/2) Epoch 8, batch 12150, loss[loss=0.2298, simple_loss=0.3032, pruned_loss=0.07825, over 6630.00 frames. ], tot_loss[loss=0.224, simple_loss=0.3044, pruned_loss=0.07183, over 1686314.56 frames. ], batch size: 97, lr: 4.22e-03, grad_scale: 16.0 2023-05-16 11:17:45,951 INFO [finetune.py:992] (0/2) Epoch 8, batch 12200, loss[loss=0.2375, simple_loss=0.3023, pruned_loss=0.08634, over 6260.00 frames. ], tot_loss[loss=0.2252, simple_loss=0.3053, pruned_loss=0.07255, over 1669649.53 frames. ], batch size: 97, lr: 4.22e-03, grad_scale: 16.0 2023-05-16 11:17:55,228 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.389e+02 3.087e+02 3.794e+02 4.591e+02 1.477e+03, threshold=7.588e+02, percent-clipped=2.0 2023-05-16 11:18:07,835 INFO [checkpoint.py:75] (0/2) Saving checkpoint to pruned_transducer_stateless7/exp_giga_finetune/epoch-8.pt 2023-05-16 11:18:32,504 INFO [finetune.py:992] (0/2) Epoch 9, batch 0, loss[loss=0.2028, simple_loss=0.289, pruned_loss=0.05833, over 12149.00 frames. ], tot_loss[loss=0.2028, simple_loss=0.289, pruned_loss=0.05833, over 12149.00 frames. ], batch size: 39, lr: 4.22e-03, grad_scale: 16.0 2023-05-16 11:18:32,505 INFO [finetune.py:1017] (0/2) Computing validation loss 2023-05-16 11:18:44,358 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.4083, 4.3322, 4.4755, 4.4529, 4.2034, 4.2176, 4.1531, 4.3120], device='cuda:0'), covar=tensor([0.0816, 0.0631, 0.0692, 0.0661, 0.1469, 0.1346, 0.0561, 0.1085], device='cuda:0'), in_proj_covar=tensor([0.0466, 0.0610, 0.0529, 0.0541, 0.0723, 0.0641, 0.0476, 0.0434], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0003], device='cuda:0') 2023-05-16 11:18:49,800 INFO [finetune.py:1026] (0/2) Epoch 9, validation: loss=0.2943, simple_loss=0.3667, pruned_loss=0.111, over 1020973.00 frames. 2023-05-16 11:18:49,801 INFO [finetune.py:1027] (0/2) Maximum memory allocated so far is 12761MB 2023-05-16 11:18:50,591 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=197872.0, num_to_drop=1, layers_to_drop={2} 2023-05-16 11:19:05,330 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=197892.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 11:19:11,796 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=197901.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 11:19:23,870 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.9971, 5.0185, 4.8466, 4.9135, 4.5391, 4.9534, 5.0449, 5.1399], device='cuda:0'), covar=tensor([0.0203, 0.0132, 0.0199, 0.0271, 0.0674, 0.0254, 0.0148, 0.0180], device='cuda:0'), in_proj_covar=tensor([0.0164, 0.0162, 0.0159, 0.0204, 0.0202, 0.0181, 0.0147, 0.0192], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:0') 2023-05-16 11:19:25,828 INFO [finetune.py:992] (0/2) Epoch 9, batch 50, loss[loss=0.1743, simple_loss=0.2648, pruned_loss=0.0419, over 12308.00 frames. ], tot_loss[loss=0.184, simple_loss=0.2738, pruned_loss=0.04712, over 540741.16 frames. ], batch size: 34, lr: 4.22e-03, grad_scale: 16.0 2023-05-16 11:19:47,773 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.811e+02 2.989e+02 3.499e+02 4.265e+02 8.208e+02, threshold=6.998e+02, percent-clipped=1.0 2023-05-16 11:19:48,656 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=197953.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 11:20:01,259 INFO [finetune.py:992] (0/2) Epoch 9, batch 100, loss[loss=0.1808, simple_loss=0.2799, pruned_loss=0.04083, over 12363.00 frames. ], tot_loss[loss=0.1815, simple_loss=0.2709, pruned_loss=0.04606, over 947742.20 frames. ], batch size: 38, lr: 4.22e-03, grad_scale: 16.0 2023-05-16 11:20:19,002 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.5696, 2.6425, 3.7096, 4.6676, 3.9651, 4.7022, 3.9158, 3.1777], device='cuda:0'), covar=tensor([0.0028, 0.0361, 0.0109, 0.0027, 0.0097, 0.0050, 0.0087, 0.0337], device='cuda:0'), in_proj_covar=tensor([0.0081, 0.0117, 0.0099, 0.0071, 0.0095, 0.0111, 0.0088, 0.0132], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 11:20:20,096 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.39 vs. limit=2.0 2023-05-16 11:20:22,632 INFO [checkpoint.py:75] (0/2) Saving checkpoint to pruned_transducer_stateless7/exp_giga_finetune/checkpoint-98000.pt 2023-05-16 11:20:36,883 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.1765, 4.7232, 5.1550, 4.5539, 4.8580, 4.6111, 5.1947, 4.8558], device='cuda:0'), covar=tensor([0.0275, 0.0424, 0.0306, 0.0270, 0.0385, 0.0326, 0.0253, 0.0296], device='cuda:0'), in_proj_covar=tensor([0.0225, 0.0234, 0.0254, 0.0232, 0.0230, 0.0230, 0.0211, 0.0189], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 11:20:40,194 INFO [finetune.py:992] (0/2) Epoch 9, batch 150, loss[loss=0.1751, simple_loss=0.2723, pruned_loss=0.039, over 12268.00 frames. ], tot_loss[loss=0.1794, simple_loss=0.2681, pruned_loss=0.04533, over 1264390.03 frames. ], batch size: 37, lr: 4.22e-03, grad_scale: 16.0 2023-05-16 11:20:51,376 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.93 vs. limit=2.0 2023-05-16 11:21:02,900 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.039e+02 2.862e+02 3.446e+02 4.094e+02 7.701e+02, threshold=6.892e+02, percent-clipped=0.0 2023-05-16 11:21:16,474 INFO [finetune.py:992] (0/2) Epoch 9, batch 200, loss[loss=0.1457, simple_loss=0.2271, pruned_loss=0.0321, over 11984.00 frames. ], tot_loss[loss=0.1755, simple_loss=0.264, pruned_loss=0.04347, over 1522065.12 frames. ], batch size: 28, lr: 4.22e-03, grad_scale: 16.0 2023-05-16 11:21:16,604 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=198071.0, num_to_drop=1, layers_to_drop={2} 2023-05-16 11:21:23,776 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=198081.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 11:21:24,602 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.2069, 4.5689, 3.9666, 4.8383, 4.4206, 2.6809, 4.3255, 2.8876], device='cuda:0'), covar=tensor([0.0837, 0.0828, 0.1470, 0.0525, 0.1220, 0.1939, 0.0934, 0.3567], device='cuda:0'), in_proj_covar=tensor([0.0290, 0.0363, 0.0342, 0.0270, 0.0353, 0.0260, 0.0329, 0.0354], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 11:21:50,764 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=198119.0, num_to_drop=1, layers_to_drop={1} 2023-05-16 11:21:52,009 INFO [finetune.py:992] (0/2) Epoch 9, batch 250, loss[loss=0.21, simple_loss=0.2935, pruned_loss=0.0633, over 12055.00 frames. ], tot_loss[loss=0.1757, simple_loss=0.2645, pruned_loss=0.04348, over 1710826.70 frames. ], batch size: 42, lr: 4.22e-03, grad_scale: 16.0 2023-05-16 11:22:06,618 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.32 vs. limit=2.0 2023-05-16 11:22:07,793 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=198142.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 11:22:14,506 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.981e+02 2.780e+02 3.190e+02 3.913e+02 1.023e+03, threshold=6.381e+02, percent-clipped=3.0 2023-05-16 11:22:18,994 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.2404, 2.3209, 3.0268, 4.1632, 2.1273, 4.1348, 4.1236, 4.2657], device='cuda:0'), covar=tensor([0.0127, 0.1225, 0.0489, 0.0133, 0.1467, 0.0186, 0.0174, 0.0096], device='cuda:0'), in_proj_covar=tensor([0.0116, 0.0198, 0.0178, 0.0110, 0.0188, 0.0169, 0.0165, 0.0114], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 11:22:27,859 INFO [finetune.py:992] (0/2) Epoch 9, batch 300, loss[loss=0.1525, simple_loss=0.2372, pruned_loss=0.03388, over 11997.00 frames. ], tot_loss[loss=0.175, simple_loss=0.2635, pruned_loss=0.04324, over 1864602.28 frames. ], batch size: 28, lr: 4.22e-03, grad_scale: 16.0 2023-05-16 11:22:28,688 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=198172.0, num_to_drop=1, layers_to_drop={2} 2023-05-16 11:22:37,451 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.8210, 3.3665, 5.1950, 2.8310, 2.8988, 3.8136, 3.3742, 3.8306], device='cuda:0'), covar=tensor([0.0373, 0.1307, 0.0325, 0.1189, 0.2080, 0.1489, 0.1370, 0.1316], device='cuda:0'), in_proj_covar=tensor([0.0226, 0.0229, 0.0232, 0.0178, 0.0231, 0.0279, 0.0217, 0.0248], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:0') 2023-05-16 11:22:49,843 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=198201.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 11:23:01,026 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=5.01 vs. limit=5.0 2023-05-16 11:23:02,263 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-05-16 11:23:04,136 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=198220.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 11:23:04,726 INFO [finetune.py:992] (0/2) Epoch 9, batch 350, loss[loss=0.1765, simple_loss=0.2674, pruned_loss=0.04282, over 12190.00 frames. ], tot_loss[loss=0.1745, simple_loss=0.2634, pruned_loss=0.04277, over 1982807.99 frames. ], batch size: 35, lr: 4.22e-03, grad_scale: 16.0 2023-05-16 11:23:24,063 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=198248.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 11:23:24,779 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=198249.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 11:23:26,510 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=4.62 vs. limit=5.0 2023-05-16 11:23:26,766 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.923e+02 2.757e+02 3.319e+02 3.988e+02 6.623e+02, threshold=6.637e+02, percent-clipped=1.0 2023-05-16 11:23:41,078 INFO [finetune.py:992] (0/2) Epoch 9, batch 400, loss[loss=0.1697, simple_loss=0.267, pruned_loss=0.03621, over 12369.00 frames. ], tot_loss[loss=0.1752, simple_loss=0.2639, pruned_loss=0.04325, over 2059666.23 frames. ], batch size: 35, lr: 4.22e-03, grad_scale: 16.0 2023-05-16 11:24:16,782 INFO [finetune.py:992] (0/2) Epoch 9, batch 450, loss[loss=0.2162, simple_loss=0.3033, pruned_loss=0.06455, over 8258.00 frames. ], tot_loss[loss=0.1754, simple_loss=0.264, pruned_loss=0.04343, over 2134345.24 frames. ], batch size: 99, lr: 4.22e-03, grad_scale: 16.0 2023-05-16 11:24:39,781 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.276e+02 2.819e+02 3.326e+02 4.197e+02 7.125e+02, threshold=6.653e+02, percent-clipped=2.0 2023-05-16 11:24:49,840 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.6922, 2.6610, 3.7708, 4.6980, 4.1914, 4.6541, 4.0930, 3.5598], device='cuda:0'), covar=tensor([0.0026, 0.0381, 0.0121, 0.0028, 0.0077, 0.0069, 0.0079, 0.0276], device='cuda:0'), in_proj_covar=tensor([0.0083, 0.0121, 0.0102, 0.0073, 0.0097, 0.0114, 0.0091, 0.0136], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 11:24:53,152 INFO [finetune.py:992] (0/2) Epoch 9, batch 500, loss[loss=0.1585, simple_loss=0.2457, pruned_loss=0.03567, over 12126.00 frames. ], tot_loss[loss=0.1745, simple_loss=0.2628, pruned_loss=0.04306, over 2191461.91 frames. ], batch size: 30, lr: 4.21e-03, grad_scale: 16.0 2023-05-16 11:25:11,270 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.0437, 4.5828, 3.8254, 4.7390, 4.3907, 2.3511, 3.9098, 2.8803], device='cuda:0'), covar=tensor([0.0754, 0.0602, 0.1513, 0.0442, 0.1036, 0.1879, 0.1105, 0.3093], device='cuda:0'), in_proj_covar=tensor([0.0292, 0.0364, 0.0345, 0.0273, 0.0354, 0.0261, 0.0331, 0.0355], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 11:25:23,965 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=2.63 vs. limit=5.0 2023-05-16 11:25:29,264 INFO [finetune.py:992] (0/2) Epoch 9, batch 550, loss[loss=0.1563, simple_loss=0.2545, pruned_loss=0.02906, over 12352.00 frames. ], tot_loss[loss=0.1741, simple_loss=0.2627, pruned_loss=0.04272, over 2235467.97 frames. ], batch size: 35, lr: 4.21e-03, grad_scale: 16.0 2023-05-16 11:25:29,656 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.88 vs. limit=2.0 2023-05-16 11:25:41,411 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=198437.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 11:25:48,722 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.1447, 5.1463, 4.9909, 5.0707, 4.6450, 5.1693, 5.0728, 5.3441], device='cuda:0'), covar=tensor([0.0207, 0.0121, 0.0156, 0.0248, 0.0673, 0.0265, 0.0142, 0.0140], device='cuda:0'), in_proj_covar=tensor([0.0178, 0.0177, 0.0173, 0.0223, 0.0220, 0.0198, 0.0160, 0.0206], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:0') 2023-05-16 11:25:52,023 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.765e+02 2.772e+02 3.197e+02 3.691e+02 5.962e+02, threshold=6.394e+02, percent-clipped=0.0 2023-05-16 11:25:57,380 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.5088, 4.8753, 3.1486, 2.8398, 4.3863, 2.5866, 4.1616, 3.3247], device='cuda:0'), covar=tensor([0.0697, 0.0556, 0.1096, 0.1511, 0.0231, 0.1569, 0.0498, 0.0888], device='cuda:0'), in_proj_covar=tensor([0.0183, 0.0243, 0.0172, 0.0196, 0.0134, 0.0179, 0.0189, 0.0171], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 11:26:05,765 INFO [finetune.py:992] (0/2) Epoch 9, batch 600, loss[loss=0.1949, simple_loss=0.2797, pruned_loss=0.05506, over 12061.00 frames. ], tot_loss[loss=0.1738, simple_loss=0.2623, pruned_loss=0.04265, over 2261085.49 frames. ], batch size: 37, lr: 4.21e-03, grad_scale: 16.0 2023-05-16 11:26:09,616 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.3878, 2.4215, 3.2361, 4.3594, 2.2045, 4.2764, 4.3699, 4.4549], device='cuda:0'), covar=tensor([0.0137, 0.1224, 0.0456, 0.0157, 0.1321, 0.0209, 0.0152, 0.0108], device='cuda:0'), in_proj_covar=tensor([0.0116, 0.0196, 0.0176, 0.0109, 0.0186, 0.0169, 0.0164, 0.0114], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 11:26:26,866 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=198499.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 11:26:42,403 INFO [finetune.py:992] (0/2) Epoch 9, batch 650, loss[loss=0.1663, simple_loss=0.2547, pruned_loss=0.03896, over 12117.00 frames. ], tot_loss[loss=0.1729, simple_loss=0.2611, pruned_loss=0.04236, over 2279893.79 frames. ], batch size: 33, lr: 4.21e-03, grad_scale: 16.0 2023-05-16 11:26:56,343 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.61 vs. limit=2.0 2023-05-16 11:27:01,639 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=198548.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 11:27:04,375 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.816e+02 2.661e+02 3.289e+02 3.726e+02 5.491e+02, threshold=6.578e+02, percent-clipped=0.0 2023-05-16 11:27:10,247 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=198560.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 11:27:17,794 INFO [finetune.py:992] (0/2) Epoch 9, batch 700, loss[loss=0.1899, simple_loss=0.2788, pruned_loss=0.05045, over 11243.00 frames. ], tot_loss[loss=0.1724, simple_loss=0.261, pruned_loss=0.0419, over 2306153.53 frames. ], batch size: 55, lr: 4.21e-03, grad_scale: 16.0 2023-05-16 11:27:35,944 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=198596.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 11:27:49,902 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=198615.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 11:27:53,995 INFO [finetune.py:992] (0/2) Epoch 9, batch 750, loss[loss=0.1862, simple_loss=0.274, pruned_loss=0.04919, over 12350.00 frames. ], tot_loss[loss=0.1733, simple_loss=0.2618, pruned_loss=0.04241, over 2319715.23 frames. ], batch size: 31, lr: 4.21e-03, grad_scale: 16.0 2023-05-16 11:28:09,743 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.7821, 3.5988, 3.3318, 3.2491, 3.0235, 2.8810, 3.6637, 2.2782], device='cuda:0'), covar=tensor([0.0303, 0.0132, 0.0148, 0.0176, 0.0346, 0.0308, 0.0108, 0.0478], device='cuda:0'), in_proj_covar=tensor([0.0184, 0.0156, 0.0151, 0.0179, 0.0196, 0.0192, 0.0161, 0.0194], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 11:28:16,594 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.757e+02 2.769e+02 3.237e+02 3.971e+02 5.764e+02, threshold=6.473e+02, percent-clipped=0.0 2023-05-16 11:28:30,060 INFO [finetune.py:992] (0/2) Epoch 9, batch 800, loss[loss=0.1484, simple_loss=0.2289, pruned_loss=0.03394, over 12288.00 frames. ], tot_loss[loss=0.1733, simple_loss=0.2622, pruned_loss=0.04223, over 2328660.26 frames. ], batch size: 28, lr: 4.21e-03, grad_scale: 16.0 2023-05-16 11:28:33,727 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=198676.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 11:28:40,551 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.41 vs. limit=2.0 2023-05-16 11:29:05,766 INFO [finetune.py:992] (0/2) Epoch 9, batch 850, loss[loss=0.1781, simple_loss=0.2642, pruned_loss=0.04599, over 12263.00 frames. ], tot_loss[loss=0.1742, simple_loss=0.263, pruned_loss=0.04274, over 2333180.45 frames. ], batch size: 37, lr: 4.21e-03, grad_scale: 16.0 2023-05-16 11:29:17,323 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=198737.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 11:29:27,743 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.967e+02 2.779e+02 3.276e+02 3.903e+02 7.256e+02, threshold=6.551e+02, percent-clipped=1.0 2023-05-16 11:29:41,452 INFO [finetune.py:992] (0/2) Epoch 9, batch 900, loss[loss=0.1695, simple_loss=0.2487, pruned_loss=0.04514, over 12250.00 frames. ], tot_loss[loss=0.1731, simple_loss=0.2619, pruned_loss=0.04218, over 2343853.21 frames. ], batch size: 32, lr: 4.21e-03, grad_scale: 16.0 2023-05-16 11:29:51,555 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=198785.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 11:30:18,283 INFO [finetune.py:992] (0/2) Epoch 9, batch 950, loss[loss=0.1654, simple_loss=0.2619, pruned_loss=0.03442, over 12160.00 frames. ], tot_loss[loss=0.1719, simple_loss=0.2607, pruned_loss=0.04158, over 2356633.47 frames. ], batch size: 36, lr: 4.21e-03, grad_scale: 16.0 2023-05-16 11:30:40,144 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.623e+02 2.863e+02 3.440e+02 4.069e+02 1.108e+03, threshold=6.879e+02, percent-clipped=2.0 2023-05-16 11:30:40,386 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=198852.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 11:30:42,383 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=198855.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 11:30:54,260 INFO [finetune.py:992] (0/2) Epoch 9, batch 1000, loss[loss=0.1411, simple_loss=0.2256, pruned_loss=0.02832, over 12107.00 frames. ], tot_loss[loss=0.1717, simple_loss=0.2603, pruned_loss=0.04156, over 2368574.11 frames. ], batch size: 30, lr: 4.21e-03, grad_scale: 16.0 2023-05-16 11:31:02,904 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=198883.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 11:31:24,257 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=198913.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 11:31:27,102 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.9148, 5.9445, 5.6084, 5.1819, 5.0735, 5.8015, 5.4300, 5.2272], device='cuda:0'), covar=tensor([0.0857, 0.1013, 0.0763, 0.1646, 0.0691, 0.0935, 0.1870, 0.1098], device='cuda:0'), in_proj_covar=tensor([0.0582, 0.0523, 0.0480, 0.0600, 0.0393, 0.0675, 0.0732, 0.0540], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-05-16 11:31:29,888 INFO [finetune.py:992] (0/2) Epoch 9, batch 1050, loss[loss=0.1693, simple_loss=0.2601, pruned_loss=0.03925, over 12344.00 frames. ], tot_loss[loss=0.1709, simple_loss=0.2591, pruned_loss=0.04138, over 2374134.93 frames. ], batch size: 31, lr: 4.21e-03, grad_scale: 16.0 2023-05-16 11:31:34,289 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=198927.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 11:31:39,591 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=4.68 vs. limit=5.0 2023-05-16 11:31:46,149 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=198944.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 11:31:52,314 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.829e+02 2.715e+02 3.357e+02 3.940e+02 9.082e+02, threshold=6.713e+02, percent-clipped=1.0 2023-05-16 11:32:03,452 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.39 vs. limit=2.0 2023-05-16 11:32:05,601 INFO [finetune.py:992] (0/2) Epoch 9, batch 1100, loss[loss=0.1796, simple_loss=0.2712, pruned_loss=0.04405, over 12358.00 frames. ], tot_loss[loss=0.1712, simple_loss=0.2595, pruned_loss=0.04146, over 2374438.91 frames. ], batch size: 35, lr: 4.21e-03, grad_scale: 32.0 2023-05-16 11:32:05,684 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=198971.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 11:32:17,944 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=198988.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 11:32:30,962 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.0065, 4.9375, 4.8342, 5.0358, 3.7095, 5.2943, 5.1376, 5.1880], device='cuda:0'), covar=tensor([0.0308, 0.0227, 0.0239, 0.0313, 0.1413, 0.0286, 0.0192, 0.0213], device='cuda:0'), in_proj_covar=tensor([0.0186, 0.0185, 0.0181, 0.0234, 0.0232, 0.0206, 0.0167, 0.0214], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0005, 0.0004, 0.0004, 0.0003, 0.0005], device='cuda:0') 2023-05-16 11:32:41,474 INFO [finetune.py:992] (0/2) Epoch 9, batch 1150, loss[loss=0.1838, simple_loss=0.2694, pruned_loss=0.04909, over 12049.00 frames. ], tot_loss[loss=0.1711, simple_loss=0.2592, pruned_loss=0.04151, over 2372857.27 frames. ], batch size: 37, lr: 4.21e-03, grad_scale: 32.0 2023-05-16 11:32:51,068 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.6567, 3.6375, 3.2978, 3.2821, 2.9461, 2.8148, 3.7048, 2.2120], device='cuda:0'), covar=tensor([0.0353, 0.0128, 0.0182, 0.0191, 0.0373, 0.0353, 0.0127, 0.0507], device='cuda:0'), in_proj_covar=tensor([0.0184, 0.0156, 0.0151, 0.0179, 0.0197, 0.0192, 0.0160, 0.0194], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 11:33:05,170 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.969e+02 2.724e+02 3.258e+02 3.920e+02 1.307e+03, threshold=6.516e+02, percent-clipped=3.0 2023-05-16 11:33:17,844 INFO [finetune.py:992] (0/2) Epoch 9, batch 1200, loss[loss=0.1551, simple_loss=0.2516, pruned_loss=0.02931, over 12146.00 frames. ], tot_loss[loss=0.1711, simple_loss=0.2591, pruned_loss=0.04151, over 2371008.14 frames. ], batch size: 34, lr: 4.21e-03, grad_scale: 16.0 2023-05-16 11:33:20,481 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.45 vs. limit=2.0 2023-05-16 11:33:23,100 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.5134, 4.0272, 4.0660, 4.4557, 3.1709, 4.1419, 2.7061, 4.2076], device='cuda:0'), covar=tensor([0.1476, 0.0767, 0.0954, 0.0617, 0.1103, 0.0540, 0.1844, 0.1265], device='cuda:0'), in_proj_covar=tensor([0.0234, 0.0270, 0.0298, 0.0358, 0.0242, 0.0240, 0.0267, 0.0383], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 11:33:53,889 INFO [finetune.py:992] (0/2) Epoch 9, batch 1250, loss[loss=0.1627, simple_loss=0.2547, pruned_loss=0.03536, over 12270.00 frames. ], tot_loss[loss=0.1708, simple_loss=0.2589, pruned_loss=0.04135, over 2370061.31 frames. ], batch size: 37, lr: 4.21e-03, grad_scale: 16.0 2023-05-16 11:34:12,585 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.9341, 5.9233, 5.6600, 5.1679, 4.9786, 5.7561, 5.4152, 5.1442], device='cuda:0'), covar=tensor([0.0741, 0.0845, 0.0667, 0.1648, 0.0780, 0.0845, 0.1582, 0.1228], device='cuda:0'), in_proj_covar=tensor([0.0580, 0.0526, 0.0481, 0.0601, 0.0395, 0.0677, 0.0732, 0.0544], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-05-16 11:34:16,573 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.991e+02 2.725e+02 3.192e+02 3.846e+02 8.183e+02, threshold=6.384e+02, percent-clipped=4.0 2023-05-16 11:34:18,196 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=199155.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 11:34:22,562 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.2546, 4.5494, 3.9785, 4.8204, 4.6700, 2.9538, 4.2736, 2.9298], device='cuda:0'), covar=tensor([0.0796, 0.0849, 0.1515, 0.0517, 0.0880, 0.1680, 0.0998, 0.3544], device='cuda:0'), in_proj_covar=tensor([0.0295, 0.0369, 0.0349, 0.0276, 0.0358, 0.0263, 0.0333, 0.0358], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 11:34:30,056 INFO [finetune.py:992] (0/2) Epoch 9, batch 1300, loss[loss=0.1512, simple_loss=0.2437, pruned_loss=0.02938, over 12169.00 frames. ], tot_loss[loss=0.1712, simple_loss=0.2597, pruned_loss=0.04138, over 2379814.27 frames. ], batch size: 31, lr: 4.21e-03, grad_scale: 16.0 2023-05-16 11:34:33,776 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=199176.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 11:34:52,896 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=199203.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 11:34:56,516 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=199208.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 11:35:05,594 INFO [finetune.py:992] (0/2) Epoch 9, batch 1350, loss[loss=0.2045, simple_loss=0.2835, pruned_loss=0.06273, over 7907.00 frames. ], tot_loss[loss=0.1717, simple_loss=0.2603, pruned_loss=0.04158, over 2378309.60 frames. ], batch size: 98, lr: 4.21e-03, grad_scale: 16.0 2023-05-16 11:35:16,943 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=199237.0, num_to_drop=1, layers_to_drop={2} 2023-05-16 11:35:18,805 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=199239.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 11:35:28,444 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.792e+02 2.776e+02 3.275e+02 3.775e+02 9.733e+02, threshold=6.551e+02, percent-clipped=4.0 2023-05-16 11:35:41,239 INFO [finetune.py:992] (0/2) Epoch 9, batch 1400, loss[loss=0.1714, simple_loss=0.2713, pruned_loss=0.03577, over 10338.00 frames. ], tot_loss[loss=0.1722, simple_loss=0.2605, pruned_loss=0.04197, over 2366281.49 frames. ], batch size: 68, lr: 4.21e-03, grad_scale: 16.0 2023-05-16 11:35:41,399 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=199271.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 11:35:49,894 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=199283.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 11:36:16,313 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=199319.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 11:36:17,648 INFO [finetune.py:992] (0/2) Epoch 9, batch 1450, loss[loss=0.1859, simple_loss=0.2756, pruned_loss=0.04812, over 12380.00 frames. ], tot_loss[loss=0.1714, simple_loss=0.2599, pruned_loss=0.04142, over 2370681.08 frames. ], batch size: 38, lr: 4.20e-03, grad_scale: 16.0 2023-05-16 11:36:32,233 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2023-05-16 11:36:40,396 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.815e+02 2.705e+02 3.180e+02 3.708e+02 6.435e+02, threshold=6.359e+02, percent-clipped=0.0 2023-05-16 11:36:53,056 INFO [finetune.py:992] (0/2) Epoch 9, batch 1500, loss[loss=0.1495, simple_loss=0.2397, pruned_loss=0.02966, over 12170.00 frames. ], tot_loss[loss=0.1711, simple_loss=0.2598, pruned_loss=0.04117, over 2372334.15 frames. ], batch size: 31, lr: 4.20e-03, grad_scale: 16.0 2023-05-16 11:37:14,632 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=199400.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 11:37:27,973 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=3.31 vs. limit=5.0 2023-05-16 11:37:28,033 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.38 vs. limit=2.0 2023-05-16 11:37:29,779 INFO [finetune.py:992] (0/2) Epoch 9, batch 1550, loss[loss=0.1584, simple_loss=0.2531, pruned_loss=0.03184, over 12294.00 frames. ], tot_loss[loss=0.1707, simple_loss=0.2596, pruned_loss=0.04093, over 2377921.93 frames. ], batch size: 33, lr: 4.20e-03, grad_scale: 16.0 2023-05-16 11:37:52,557 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.003e+02 2.743e+02 3.183e+02 3.898e+02 7.323e+02, threshold=6.366e+02, percent-clipped=2.0 2023-05-16 11:37:54,581 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.86 vs. limit=2.0 2023-05-16 11:37:56,331 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=199458.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 11:37:58,452 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=199461.0, num_to_drop=1, layers_to_drop={1} 2023-05-16 11:38:05,999 INFO [finetune.py:992] (0/2) Epoch 9, batch 1600, loss[loss=0.1766, simple_loss=0.2683, pruned_loss=0.04248, over 12189.00 frames. ], tot_loss[loss=0.1701, simple_loss=0.259, pruned_loss=0.04059, over 2383662.91 frames. ], batch size: 35, lr: 4.20e-03, grad_scale: 16.0 2023-05-16 11:38:22,615 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=199494.0, num_to_drop=1, layers_to_drop={1} 2023-05-16 11:38:32,474 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=199508.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 11:38:40,461 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=199519.0, num_to_drop=1, layers_to_drop={0} 2023-05-16 11:38:41,660 INFO [finetune.py:992] (0/2) Epoch 9, batch 1650, loss[loss=0.1898, simple_loss=0.2783, pruned_loss=0.05061, over 11788.00 frames. ], tot_loss[loss=0.1705, simple_loss=0.2595, pruned_loss=0.04078, over 2386689.46 frames. ], batch size: 44, lr: 4.20e-03, grad_scale: 16.0 2023-05-16 11:38:49,649 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=199532.0, num_to_drop=1, layers_to_drop={3} 2023-05-16 11:38:52,435 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=199535.0, num_to_drop=1, layers_to_drop={0} 2023-05-16 11:38:55,047 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=199539.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 11:38:58,861 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.61 vs. limit=2.0 2023-05-16 11:39:04,802 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.685e+02 2.622e+02 3.079e+02 3.563e+02 8.648e+02, threshold=6.157e+02, percent-clipped=1.0 2023-05-16 11:39:06,426 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=199555.0, num_to_drop=1, layers_to_drop={3} 2023-05-16 11:39:06,977 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=199556.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 11:39:17,506 INFO [finetune.py:992] (0/2) Epoch 9, batch 1700, loss[loss=0.1545, simple_loss=0.2409, pruned_loss=0.03401, over 12022.00 frames. ], tot_loss[loss=0.171, simple_loss=0.2599, pruned_loss=0.04106, over 2386984.34 frames. ], batch size: 31, lr: 4.20e-03, grad_scale: 16.0 2023-05-16 11:39:24,796 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.8238, 3.2784, 5.2249, 2.6889, 2.8305, 3.8042, 3.4497, 3.9952], device='cuda:0'), covar=tensor([0.0443, 0.1142, 0.0283, 0.1075, 0.1860, 0.1410, 0.1152, 0.0951], device='cuda:0'), in_proj_covar=tensor([0.0228, 0.0230, 0.0236, 0.0179, 0.0232, 0.0281, 0.0219, 0.0251], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:0') 2023-05-16 11:39:26,183 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=199583.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 11:39:28,901 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=199587.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 11:39:35,536 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=199596.0, num_to_drop=1, layers_to_drop={0} 2023-05-16 11:39:37,547 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.4581, 5.0253, 5.4285, 4.7769, 5.0793, 4.8145, 5.4814, 5.0842], device='cuda:0'), covar=tensor([0.0231, 0.0347, 0.0259, 0.0247, 0.0291, 0.0320, 0.0196, 0.0350], device='cuda:0'), in_proj_covar=tensor([0.0242, 0.0253, 0.0274, 0.0248, 0.0248, 0.0249, 0.0227, 0.0207], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 11:39:53,787 INFO [finetune.py:992] (0/2) Epoch 9, batch 1750, loss[loss=0.1819, simple_loss=0.2668, pruned_loss=0.04852, over 11830.00 frames. ], tot_loss[loss=0.1713, simple_loss=0.2604, pruned_loss=0.04108, over 2392742.71 frames. ], batch size: 44, lr: 4.20e-03, grad_scale: 16.0 2023-05-16 11:40:01,097 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=199631.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 11:40:16,865 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.948e+02 2.823e+02 3.191e+02 3.852e+02 7.415e+02, threshold=6.382e+02, percent-clipped=2.0 2023-05-16 11:40:29,899 INFO [finetune.py:992] (0/2) Epoch 9, batch 1800, loss[loss=0.146, simple_loss=0.2332, pruned_loss=0.02941, over 12173.00 frames. ], tot_loss[loss=0.1717, simple_loss=0.2605, pruned_loss=0.04149, over 2382927.37 frames. ], batch size: 31, lr: 4.20e-03, grad_scale: 16.0 2023-05-16 11:41:06,301 INFO [finetune.py:992] (0/2) Epoch 9, batch 1850, loss[loss=0.1533, simple_loss=0.2497, pruned_loss=0.02851, over 12187.00 frames. ], tot_loss[loss=0.1709, simple_loss=0.2596, pruned_loss=0.04112, over 2378745.25 frames. ], batch size: 31, lr: 4.20e-03, grad_scale: 16.0 2023-05-16 11:41:25,538 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.4037, 4.9713, 5.3475, 4.7056, 4.9575, 4.7428, 5.3999, 5.0141], device='cuda:0'), covar=tensor([0.0257, 0.0364, 0.0308, 0.0260, 0.0385, 0.0322, 0.0236, 0.0316], device='cuda:0'), in_proj_covar=tensor([0.0245, 0.0255, 0.0278, 0.0250, 0.0250, 0.0252, 0.0230, 0.0209], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 11:41:28,996 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.560e+02 2.717e+02 3.262e+02 3.822e+02 8.454e+02, threshold=6.525e+02, percent-clipped=2.0 2023-05-16 11:41:31,236 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=199756.0, num_to_drop=1, layers_to_drop={0} 2023-05-16 11:41:42,430 INFO [finetune.py:992] (0/2) Epoch 9, batch 1900, loss[loss=0.1565, simple_loss=0.2392, pruned_loss=0.03692, over 12162.00 frames. ], tot_loss[loss=0.1709, simple_loss=0.2596, pruned_loss=0.04109, over 2374399.96 frames. ], batch size: 29, lr: 4.20e-03, grad_scale: 16.0 2023-05-16 11:41:53,482 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.98 vs. limit=2.0 2023-05-16 11:41:53,622 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.37 vs. limit=2.0 2023-05-16 11:41:55,106 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.40 vs. limit=2.0 2023-05-16 11:42:13,427 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=199814.0, num_to_drop=1, layers_to_drop={0} 2023-05-16 11:42:18,249 INFO [finetune.py:992] (0/2) Epoch 9, batch 1950, loss[loss=0.1797, simple_loss=0.2484, pruned_loss=0.05551, over 12292.00 frames. ], tot_loss[loss=0.1714, simple_loss=0.2599, pruned_loss=0.04143, over 2368820.21 frames. ], batch size: 28, lr: 4.20e-03, grad_scale: 16.0 2023-05-16 11:42:26,109 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=199832.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 11:42:39,367 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=199850.0, num_to_drop=1, layers_to_drop={3} 2023-05-16 11:42:41,364 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.935e+02 2.711e+02 3.179e+02 3.779e+02 6.587e+02, threshold=6.359e+02, percent-clipped=1.0 2023-05-16 11:42:54,194 INFO [finetune.py:992] (0/2) Epoch 9, batch 2000, loss[loss=0.1541, simple_loss=0.239, pruned_loss=0.03463, over 12263.00 frames. ], tot_loss[loss=0.1714, simple_loss=0.2602, pruned_loss=0.04129, over 2373140.69 frames. ], batch size: 32, lr: 4.20e-03, grad_scale: 16.0 2023-05-16 11:43:01,017 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=199880.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 11:43:08,758 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=199891.0, num_to_drop=1, layers_to_drop={1} 2023-05-16 11:43:11,017 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.1300, 3.8251, 3.8232, 4.1883, 2.7784, 3.7637, 2.5241, 3.8619], device='cuda:0'), covar=tensor([0.1682, 0.0707, 0.0921, 0.0740, 0.1203, 0.0627, 0.1830, 0.1115], device='cuda:0'), in_proj_covar=tensor([0.0231, 0.0268, 0.0297, 0.0356, 0.0241, 0.0239, 0.0264, 0.0381], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 11:43:22,302 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=199909.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 11:43:30,666 INFO [finetune.py:992] (0/2) Epoch 9, batch 2050, loss[loss=0.1737, simple_loss=0.26, pruned_loss=0.04369, over 12164.00 frames. ], tot_loss[loss=0.1711, simple_loss=0.2598, pruned_loss=0.04116, over 2372478.99 frames. ], batch size: 34, lr: 4.20e-03, grad_scale: 16.0 2023-05-16 11:43:47,802 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=3.20 vs. limit=5.0 2023-05-16 11:43:49,986 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.85 vs. limit=2.0 2023-05-16 11:43:53,839 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.235e+02 2.846e+02 3.318e+02 3.919e+02 6.845e+02, threshold=6.636e+02, percent-clipped=2.0 2023-05-16 11:43:54,851 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.3404, 4.6140, 2.7007, 2.5111, 4.0299, 2.7190, 3.9945, 3.1295], device='cuda:0'), covar=tensor([0.0649, 0.0490, 0.1281, 0.1610, 0.0277, 0.1306, 0.0482, 0.0846], device='cuda:0'), in_proj_covar=tensor([0.0185, 0.0249, 0.0175, 0.0198, 0.0138, 0.0181, 0.0194, 0.0172], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 11:44:06,097 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.2319, 2.0539, 2.9703, 3.1935, 3.1448, 3.1228, 3.0016, 2.5146], device='cuda:0'), covar=tensor([0.0067, 0.0417, 0.0136, 0.0061, 0.0112, 0.0112, 0.0118, 0.0345], device='cuda:0'), in_proj_covar=tensor([0.0083, 0.0119, 0.0102, 0.0073, 0.0098, 0.0114, 0.0091, 0.0134], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 11:44:06,129 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=199970.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 11:44:06,560 INFO [finetune.py:992] (0/2) Epoch 9, batch 2100, loss[loss=0.2112, simple_loss=0.283, pruned_loss=0.06976, over 8395.00 frames. ], tot_loss[loss=0.1713, simple_loss=0.2598, pruned_loss=0.04141, over 2359694.96 frames. ], batch size: 97, lr: 4.20e-03, grad_scale: 16.0 2023-05-16 11:44:28,303 INFO [checkpoint.py:75] (0/2) Saving checkpoint to pruned_transducer_stateless7/exp_giga_finetune/checkpoint-100000.pt 2023-05-16 11:44:46,120 INFO [finetune.py:992] (0/2) Epoch 9, batch 2150, loss[loss=0.1721, simple_loss=0.2618, pruned_loss=0.04122, over 12253.00 frames. ], tot_loss[loss=0.1706, simple_loss=0.259, pruned_loss=0.04117, over 2367068.04 frames. ], batch size: 32, lr: 4.20e-03, grad_scale: 16.0 2023-05-16 11:45:09,897 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.946e+02 2.662e+02 3.100e+02 3.609e+02 6.112e+02, threshold=6.199e+02, percent-clipped=0.0 2023-05-16 11:45:12,309 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=200056.0, num_to_drop=1, layers_to_drop={2} 2023-05-16 11:45:22,849 INFO [finetune.py:992] (0/2) Epoch 9, batch 2200, loss[loss=0.1579, simple_loss=0.2409, pruned_loss=0.03744, over 11743.00 frames. ], tot_loss[loss=0.1712, simple_loss=0.2593, pruned_loss=0.04157, over 2358583.78 frames. ], batch size: 26, lr: 4.20e-03, grad_scale: 16.0 2023-05-16 11:45:46,356 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=200104.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 11:45:53,085 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.4293, 2.5688, 3.1132, 4.3754, 2.2555, 4.4311, 4.3898, 4.5613], device='cuda:0'), covar=tensor([0.0153, 0.1211, 0.0520, 0.0166, 0.1503, 0.0229, 0.0216, 0.0095], device='cuda:0'), in_proj_covar=tensor([0.0119, 0.0201, 0.0180, 0.0115, 0.0190, 0.0174, 0.0171, 0.0117], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 11:45:53,730 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=200114.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 11:45:56,521 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.7552, 5.7470, 5.5715, 5.0718, 5.0584, 5.6553, 5.3120, 5.1914], device='cuda:0'), covar=tensor([0.0711, 0.0788, 0.0553, 0.1625, 0.0758, 0.0760, 0.1448, 0.0997], device='cuda:0'), in_proj_covar=tensor([0.0591, 0.0538, 0.0490, 0.0617, 0.0400, 0.0690, 0.0747, 0.0556], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0004, 0.0004, 0.0003], device='cuda:0') 2023-05-16 11:45:58,491 INFO [finetune.py:992] (0/2) Epoch 9, batch 2250, loss[loss=0.1905, simple_loss=0.2895, pruned_loss=0.04572, over 10490.00 frames. ], tot_loss[loss=0.1704, simple_loss=0.2588, pruned_loss=0.04106, over 2360326.02 frames. ], batch size: 68, lr: 4.20e-03, grad_scale: 16.0 2023-05-16 11:46:19,581 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=200150.0, num_to_drop=1, layers_to_drop={2} 2023-05-16 11:46:21,557 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.949e+02 2.841e+02 3.279e+02 4.097e+02 1.905e+03, threshold=6.557e+02, percent-clipped=6.0 2023-05-16 11:46:28,025 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=200162.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 11:46:32,422 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.9679, 4.6223, 4.8362, 4.7944, 4.5863, 4.8816, 4.7244, 2.6686], device='cuda:0'), covar=tensor([0.0102, 0.0067, 0.0085, 0.0070, 0.0051, 0.0084, 0.0086, 0.0774], device='cuda:0'), in_proj_covar=tensor([0.0065, 0.0076, 0.0079, 0.0071, 0.0057, 0.0087, 0.0078, 0.0095], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-16 11:46:34,352 INFO [finetune.py:992] (0/2) Epoch 9, batch 2300, loss[loss=0.1863, simple_loss=0.2881, pruned_loss=0.04225, over 12366.00 frames. ], tot_loss[loss=0.1712, simple_loss=0.2596, pruned_loss=0.0414, over 2372206.46 frames. ], batch size: 38, lr: 4.20e-03, grad_scale: 16.0 2023-05-16 11:46:38,688 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=200177.0, num_to_drop=1, layers_to_drop={0} 2023-05-16 11:46:49,336 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=200191.0, num_to_drop=1, layers_to_drop={1} 2023-05-16 11:46:53,979 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=200198.0, num_to_drop=1, layers_to_drop={1} 2023-05-16 11:47:10,423 INFO [finetune.py:992] (0/2) Epoch 9, batch 2350, loss[loss=0.1835, simple_loss=0.2759, pruned_loss=0.0455, over 10528.00 frames. ], tot_loss[loss=0.1714, simple_loss=0.2599, pruned_loss=0.04142, over 2369702.98 frames. ], batch size: 68, lr: 4.20e-03, grad_scale: 16.0 2023-05-16 11:47:22,598 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=200238.0, num_to_drop=1, layers_to_drop={3} 2023-05-16 11:47:23,181 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=200239.0, num_to_drop=1, layers_to_drop={0} 2023-05-16 11:47:33,030 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.922e+02 2.910e+02 3.318e+02 3.930e+02 2.008e+03, threshold=6.636e+02, percent-clipped=7.0 2023-05-16 11:47:41,241 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=200265.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 11:47:45,425 INFO [finetune.py:992] (0/2) Epoch 9, batch 2400, loss[loss=0.1779, simple_loss=0.2704, pruned_loss=0.04276, over 12343.00 frames. ], tot_loss[loss=0.1715, simple_loss=0.2598, pruned_loss=0.04153, over 2369679.96 frames. ], batch size: 35, lr: 4.19e-03, grad_scale: 16.0 2023-05-16 11:47:51,866 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=200279.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 11:48:21,835 INFO [finetune.py:992] (0/2) Epoch 9, batch 2450, loss[loss=0.186, simple_loss=0.2788, pruned_loss=0.04663, over 10443.00 frames. ], tot_loss[loss=0.1712, simple_loss=0.2594, pruned_loss=0.04146, over 2369864.54 frames. ], batch size: 68, lr: 4.19e-03, grad_scale: 16.0 2023-05-16 11:48:36,351 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=200340.0, num_to_drop=1, layers_to_drop={1} 2023-05-16 11:48:45,374 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.956e+02 2.761e+02 3.380e+02 3.986e+02 6.954e+02, threshold=6.761e+02, percent-clipped=1.0 2023-05-16 11:48:45,611 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.3669, 3.2602, 3.0289, 3.0211, 2.7174, 2.5177, 3.2384, 2.2467], device='cuda:0'), covar=tensor([0.0350, 0.0153, 0.0198, 0.0207, 0.0405, 0.0374, 0.0136, 0.0415], device='cuda:0'), in_proj_covar=tensor([0.0189, 0.0161, 0.0156, 0.0187, 0.0202, 0.0197, 0.0166, 0.0198], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 11:48:51,284 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.7756, 3.4147, 5.1456, 2.5867, 2.9001, 3.8140, 3.2888, 3.8467], device='cuda:0'), covar=tensor([0.0391, 0.1063, 0.0298, 0.1144, 0.1874, 0.1420, 0.1290, 0.1107], device='cuda:0'), in_proj_covar=tensor([0.0232, 0.0232, 0.0238, 0.0180, 0.0235, 0.0285, 0.0221, 0.0255], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:0') 2023-05-16 11:48:52,682 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.7469, 3.7732, 3.3459, 3.4696, 3.0807, 2.9692, 3.8098, 2.5480], device='cuda:0'), covar=tensor([0.0323, 0.0135, 0.0181, 0.0163, 0.0350, 0.0302, 0.0107, 0.0396], device='cuda:0'), in_proj_covar=tensor([0.0189, 0.0161, 0.0157, 0.0187, 0.0203, 0.0197, 0.0167, 0.0199], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 11:48:58,195 INFO [finetune.py:992] (0/2) Epoch 9, batch 2500, loss[loss=0.165, simple_loss=0.2617, pruned_loss=0.03415, over 12179.00 frames. ], tot_loss[loss=0.171, simple_loss=0.2595, pruned_loss=0.04127, over 2370912.34 frames. ], batch size: 35, lr: 4.19e-03, grad_scale: 16.0 2023-05-16 11:49:34,579 INFO [finetune.py:992] (0/2) Epoch 9, batch 2550, loss[loss=0.1692, simple_loss=0.2544, pruned_loss=0.04203, over 12341.00 frames. ], tot_loss[loss=0.1706, simple_loss=0.259, pruned_loss=0.04108, over 2371262.06 frames. ], batch size: 31, lr: 4.19e-03, grad_scale: 16.0 2023-05-16 11:49:44,866 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.4463, 4.0804, 4.2641, 4.5096, 3.2427, 4.0296, 2.6641, 4.0936], device='cuda:0'), covar=tensor([0.1629, 0.0743, 0.0859, 0.0650, 0.1116, 0.0557, 0.1748, 0.1494], device='cuda:0'), in_proj_covar=tensor([0.0227, 0.0264, 0.0294, 0.0352, 0.0236, 0.0236, 0.0260, 0.0375], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 11:49:48,440 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.9494, 2.3332, 3.3734, 2.9227, 3.2110, 3.0852, 2.3803, 3.2983], device='cuda:0'), covar=tensor([0.0117, 0.0331, 0.0156, 0.0220, 0.0122, 0.0158, 0.0337, 0.0110], device='cuda:0'), in_proj_covar=tensor([0.0172, 0.0201, 0.0182, 0.0181, 0.0208, 0.0157, 0.0194, 0.0177], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 11:49:49,213 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.4161, 4.6152, 4.1123, 5.0563, 4.6072, 3.0682, 4.1665, 3.0634], device='cuda:0'), covar=tensor([0.0768, 0.0901, 0.1469, 0.0442, 0.1045, 0.1507, 0.1174, 0.3351], device='cuda:0'), in_proj_covar=tensor([0.0297, 0.0372, 0.0350, 0.0279, 0.0359, 0.0263, 0.0334, 0.0358], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 11:49:55,860 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=2.01 vs. limit=2.0 2023-05-16 11:49:57,450 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.601e+02 2.735e+02 3.226e+02 3.817e+02 9.182e+02, threshold=6.453e+02, percent-clipped=2.0 2023-05-16 11:50:06,669 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.46 vs. limit=2.0 2023-05-16 11:50:10,326 INFO [finetune.py:992] (0/2) Epoch 9, batch 2600, loss[loss=0.1814, simple_loss=0.2669, pruned_loss=0.04793, over 12069.00 frames. ], tot_loss[loss=0.1697, simple_loss=0.2577, pruned_loss=0.04081, over 2365790.07 frames. ], batch size: 42, lr: 4.19e-03, grad_scale: 16.0 2023-05-16 11:50:23,976 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.1125, 3.8628, 4.0106, 4.4598, 3.0196, 3.8633, 2.5549, 4.0208], device='cuda:0'), covar=tensor([0.1876, 0.0841, 0.0947, 0.0557, 0.1261, 0.0674, 0.1991, 0.1245], device='cuda:0'), in_proj_covar=tensor([0.0227, 0.0264, 0.0294, 0.0353, 0.0237, 0.0236, 0.0261, 0.0375], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 11:50:45,189 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2023-05-16 11:50:46,264 INFO [finetune.py:992] (0/2) Epoch 9, batch 2650, loss[loss=0.1503, simple_loss=0.2289, pruned_loss=0.03581, over 12275.00 frames. ], tot_loss[loss=0.1699, simple_loss=0.2578, pruned_loss=0.04105, over 2360993.22 frames. ], batch size: 28, lr: 4.19e-03, grad_scale: 16.0 2023-05-16 11:50:54,991 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=200533.0, num_to_drop=1, layers_to_drop={1} 2023-05-16 11:51:08,744 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.966e+02 2.749e+02 3.270e+02 3.848e+02 7.144e+02, threshold=6.540e+02, percent-clipped=2.0 2023-05-16 11:51:18,127 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=200565.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 11:51:22,211 INFO [finetune.py:992] (0/2) Epoch 9, batch 2700, loss[loss=0.1593, simple_loss=0.2379, pruned_loss=0.04035, over 12184.00 frames. ], tot_loss[loss=0.1697, simple_loss=0.2577, pruned_loss=0.04086, over 2360922.02 frames. ], batch size: 29, lr: 4.19e-03, grad_scale: 16.0 2023-05-16 11:51:52,355 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=200613.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 11:51:58,032 INFO [finetune.py:992] (0/2) Epoch 9, batch 2750, loss[loss=0.1618, simple_loss=0.2521, pruned_loss=0.0358, over 12163.00 frames. ], tot_loss[loss=0.1695, simple_loss=0.2574, pruned_loss=0.04077, over 2363544.70 frames. ], batch size: 34, lr: 4.19e-03, grad_scale: 16.0 2023-05-16 11:52:08,739 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=200635.0, num_to_drop=1, layers_to_drop={3} 2023-05-16 11:52:15,976 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=200645.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 11:52:16,683 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([6.1896, 6.1257, 5.9482, 5.4489, 5.2110, 6.0464, 5.6909, 5.4747], device='cuda:0'), covar=tensor([0.0523, 0.0774, 0.0546, 0.1542, 0.0636, 0.0649, 0.1395, 0.1025], device='cuda:0'), in_proj_covar=tensor([0.0592, 0.0538, 0.0496, 0.0620, 0.0402, 0.0691, 0.0748, 0.0555], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0004, 0.0004, 0.0003], device='cuda:0') 2023-05-16 11:52:17,493 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.5240, 4.8561, 3.1010, 2.7136, 4.2796, 2.5744, 4.0794, 3.3979], device='cuda:0'), covar=tensor([0.0636, 0.0431, 0.1085, 0.1466, 0.0211, 0.1348, 0.0453, 0.0736], device='cuda:0'), in_proj_covar=tensor([0.0187, 0.0252, 0.0176, 0.0200, 0.0139, 0.0181, 0.0195, 0.0174], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 11:52:21,443 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.960e+02 2.726e+02 3.164e+02 4.050e+02 6.073e+02, threshold=6.328e+02, percent-clipped=0.0 2023-05-16 11:52:34,315 INFO [finetune.py:992] (0/2) Epoch 9, batch 2800, loss[loss=0.1563, simple_loss=0.2348, pruned_loss=0.03894, over 12018.00 frames. ], tot_loss[loss=0.1694, simple_loss=0.2573, pruned_loss=0.04076, over 2363491.05 frames. ], batch size: 31, lr: 4.19e-03, grad_scale: 16.0 2023-05-16 11:52:59,296 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=200706.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 11:53:10,423 INFO [finetune.py:992] (0/2) Epoch 9, batch 2850, loss[loss=0.2251, simple_loss=0.2957, pruned_loss=0.07726, over 8039.00 frames. ], tot_loss[loss=0.17, simple_loss=0.258, pruned_loss=0.04098, over 2368956.71 frames. ], batch size: 98, lr: 4.19e-03, grad_scale: 16.0 2023-05-16 11:53:27,397 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.44 vs. limit=2.0 2023-05-16 11:53:32,651 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.9096, 5.8764, 5.6867, 5.1659, 5.1203, 5.7740, 5.3786, 5.2026], device='cuda:0'), covar=tensor([0.0789, 0.1048, 0.0687, 0.1681, 0.0779, 0.0756, 0.1692, 0.1163], device='cuda:0'), in_proj_covar=tensor([0.0593, 0.0538, 0.0496, 0.0619, 0.0400, 0.0692, 0.0747, 0.0556], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0004, 0.0004, 0.0003], device='cuda:0') 2023-05-16 11:53:33,240 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.932e+02 2.724e+02 3.281e+02 3.924e+02 1.531e+03, threshold=6.562e+02, percent-clipped=4.0 2023-05-16 11:53:46,048 INFO [finetune.py:992] (0/2) Epoch 9, batch 2900, loss[loss=0.1752, simple_loss=0.2691, pruned_loss=0.04069, over 12151.00 frames. ], tot_loss[loss=0.1696, simple_loss=0.258, pruned_loss=0.04059, over 2374159.40 frames. ], batch size: 34, lr: 4.19e-03, grad_scale: 16.0 2023-05-16 11:54:23,015 INFO [finetune.py:992] (0/2) Epoch 9, batch 2950, loss[loss=0.1696, simple_loss=0.2593, pruned_loss=0.03998, over 12300.00 frames. ], tot_loss[loss=0.1694, simple_loss=0.2584, pruned_loss=0.04023, over 2383247.89 frames. ], batch size: 34, lr: 4.19e-03, grad_scale: 16.0 2023-05-16 11:54:31,738 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=200833.0, num_to_drop=1, layers_to_drop={1} 2023-05-16 11:54:43,063 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.3058, 4.0294, 4.2453, 4.4396, 3.1245, 4.0198, 2.8137, 4.0743], device='cuda:0'), covar=tensor([0.1629, 0.0754, 0.0811, 0.0647, 0.1087, 0.0575, 0.1688, 0.1173], device='cuda:0'), in_proj_covar=tensor([0.0225, 0.0263, 0.0290, 0.0350, 0.0234, 0.0234, 0.0257, 0.0372], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 11:54:45,608 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.993e+02 2.805e+02 3.316e+02 3.739e+02 7.927e+02, threshold=6.633e+02, percent-clipped=2.0 2023-05-16 11:54:58,964 INFO [finetune.py:992] (0/2) Epoch 9, batch 3000, loss[loss=0.1694, simple_loss=0.2651, pruned_loss=0.03679, over 12136.00 frames. ], tot_loss[loss=0.1698, simple_loss=0.2588, pruned_loss=0.04034, over 2368193.63 frames. ], batch size: 36, lr: 4.19e-03, grad_scale: 16.0 2023-05-16 11:54:58,965 INFO [finetune.py:1017] (0/2) Computing validation loss 2023-05-16 11:55:16,375 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.0456, 1.7062, 2.0711, 1.8216, 2.0312, 2.1087, 1.6071, 2.0574], device='cuda:0'), covar=tensor([0.0098, 0.0247, 0.0105, 0.0130, 0.0098, 0.0097, 0.0222, 0.0104], device='cuda:0'), in_proj_covar=tensor([0.0171, 0.0201, 0.0182, 0.0181, 0.0209, 0.0159, 0.0194, 0.0178], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 11:55:16,844 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.4659, 2.8460, 4.1624, 4.3688, 2.8925, 2.6598, 2.8037, 2.0268], device='cuda:0'), covar=tensor([0.1586, 0.2473, 0.0571, 0.0454, 0.1208, 0.2212, 0.2812, 0.4744], device='cuda:0'), in_proj_covar=tensor([0.0283, 0.0376, 0.0268, 0.0290, 0.0260, 0.0292, 0.0365, 0.0362], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-16 11:55:17,474 INFO [finetune.py:1026] (0/2) Epoch 9, validation: loss=0.3195, simple_loss=0.3965, pruned_loss=0.1213, over 1020973.00 frames. 2023-05-16 11:55:17,475 INFO [finetune.py:1027] (0/2) Maximum memory allocated so far is 12761MB 2023-05-16 11:55:25,238 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=200881.0, num_to_drop=1, layers_to_drop={0} 2023-05-16 11:55:36,438 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.36 vs. limit=2.0 2023-05-16 11:55:53,793 INFO [finetune.py:992] (0/2) Epoch 9, batch 3050, loss[loss=0.1999, simple_loss=0.2831, pruned_loss=0.05837, over 12113.00 frames. ], tot_loss[loss=0.1692, simple_loss=0.2581, pruned_loss=0.04021, over 2379659.16 frames. ], batch size: 38, lr: 4.19e-03, grad_scale: 16.0 2023-05-16 11:56:03,870 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=200935.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 11:56:16,430 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.029e+02 2.822e+02 3.318e+02 4.013e+02 5.698e+02, threshold=6.637e+02, percent-clipped=0.0 2023-05-16 11:56:30,092 INFO [finetune.py:992] (0/2) Epoch 9, batch 3100, loss[loss=0.1571, simple_loss=0.2461, pruned_loss=0.03406, over 12178.00 frames. ], tot_loss[loss=0.1692, simple_loss=0.258, pruned_loss=0.04021, over 2377745.37 frames. ], batch size: 31, lr: 4.19e-03, grad_scale: 16.0 2023-05-16 11:56:38,833 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=200983.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 11:56:51,827 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=201001.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 11:57:06,171 INFO [finetune.py:992] (0/2) Epoch 9, batch 3150, loss[loss=0.1688, simple_loss=0.2569, pruned_loss=0.04034, over 12250.00 frames. ], tot_loss[loss=0.1686, simple_loss=0.2573, pruned_loss=0.03995, over 2383292.61 frames. ], batch size: 32, lr: 4.19e-03, grad_scale: 16.0 2023-05-16 11:57:29,726 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.942e+02 2.683e+02 3.143e+02 3.728e+02 6.490e+02, threshold=6.287e+02, percent-clipped=0.0 2023-05-16 11:57:42,700 INFO [finetune.py:992] (0/2) Epoch 9, batch 3200, loss[loss=0.1604, simple_loss=0.235, pruned_loss=0.04294, over 12272.00 frames. ], tot_loss[loss=0.1695, simple_loss=0.2582, pruned_loss=0.04044, over 2379236.79 frames. ], batch size: 28, lr: 4.19e-03, grad_scale: 32.0 2023-05-16 11:57:42,904 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=201071.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 11:57:55,914 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.5564, 3.5401, 3.2424, 3.2030, 2.9263, 2.7189, 3.5391, 2.3230], device='cuda:0'), covar=tensor([0.0343, 0.0124, 0.0154, 0.0176, 0.0378, 0.0320, 0.0119, 0.0448], device='cuda:0'), in_proj_covar=tensor([0.0191, 0.0162, 0.0158, 0.0187, 0.0203, 0.0198, 0.0168, 0.0200], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 11:58:19,339 INFO [finetune.py:992] (0/2) Epoch 9, batch 3250, loss[loss=0.189, simple_loss=0.2809, pruned_loss=0.04849, over 12122.00 frames. ], tot_loss[loss=0.1695, simple_loss=0.2581, pruned_loss=0.04049, over 2373795.64 frames. ], batch size: 39, lr: 4.19e-03, grad_scale: 32.0 2023-05-16 11:58:27,425 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=201132.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 11:58:41,965 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.718e+02 2.910e+02 3.291e+02 3.848e+02 5.664e+02, threshold=6.582e+02, percent-clipped=0.0 2023-05-16 11:58:46,560 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.4357, 3.1154, 3.0144, 2.9330, 2.6706, 2.5567, 3.0975, 2.1823], device='cuda:0'), covar=tensor([0.0311, 0.0143, 0.0153, 0.0175, 0.0319, 0.0276, 0.0140, 0.0384], device='cuda:0'), in_proj_covar=tensor([0.0190, 0.0161, 0.0157, 0.0187, 0.0202, 0.0197, 0.0167, 0.0200], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 11:58:54,696 INFO [finetune.py:992] (0/2) Epoch 9, batch 3300, loss[loss=0.177, simple_loss=0.2669, pruned_loss=0.04349, over 12150.00 frames. ], tot_loss[loss=0.1693, simple_loss=0.2577, pruned_loss=0.04045, over 2374818.42 frames. ], batch size: 34, lr: 4.19e-03, grad_scale: 32.0 2023-05-16 11:59:03,057 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.3519, 4.5607, 4.0335, 4.9288, 4.5278, 3.1656, 4.1772, 2.9820], device='cuda:0'), covar=tensor([0.0825, 0.0826, 0.1406, 0.0478, 0.1190, 0.1528, 0.1143, 0.3444], device='cuda:0'), in_proj_covar=tensor([0.0300, 0.0375, 0.0353, 0.0284, 0.0363, 0.0265, 0.0337, 0.0361], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 11:59:10,778 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.4327, 5.1983, 5.3719, 5.4013, 5.0032, 5.0298, 4.8283, 5.3352], device='cuda:0'), covar=tensor([0.0633, 0.0625, 0.0726, 0.0547, 0.1807, 0.1237, 0.0562, 0.1049], device='cuda:0'), in_proj_covar=tensor([0.0507, 0.0668, 0.0572, 0.0588, 0.0806, 0.0708, 0.0520, 0.0464], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0005, 0.0004, 0.0004, 0.0005, 0.0005, 0.0003, 0.0003], device='cuda:0') 2023-05-16 11:59:31,517 INFO [finetune.py:992] (0/2) Epoch 9, batch 3350, loss[loss=0.1816, simple_loss=0.2697, pruned_loss=0.04673, over 11882.00 frames. ], tot_loss[loss=0.1693, simple_loss=0.2575, pruned_loss=0.04054, over 2370648.01 frames. ], batch size: 44, lr: 4.18e-03, grad_scale: 32.0 2023-05-16 11:59:46,715 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.4722, 4.8015, 3.0876, 2.3932, 4.1940, 2.3862, 4.0698, 3.2814], device='cuda:0'), covar=tensor([0.0602, 0.0433, 0.1007, 0.1686, 0.0314, 0.1445, 0.0407, 0.0826], device='cuda:0'), in_proj_covar=tensor([0.0186, 0.0252, 0.0176, 0.0199, 0.0139, 0.0180, 0.0195, 0.0173], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 11:59:54,540 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.755e+02 2.735e+02 3.170e+02 3.586e+02 6.394e+02, threshold=6.340e+02, percent-clipped=0.0 2023-05-16 11:59:54,733 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=201253.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 12:00:08,019 INFO [finetune.py:992] (0/2) Epoch 9, batch 3400, loss[loss=0.1976, simple_loss=0.2898, pruned_loss=0.05265, over 12359.00 frames. ], tot_loss[loss=0.1694, simple_loss=0.2578, pruned_loss=0.04046, over 2380270.76 frames. ], batch size: 36, lr: 4.18e-03, grad_scale: 32.0 2023-05-16 12:00:10,401 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=201274.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 12:00:11,181 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.1259, 3.9064, 2.6791, 2.2682, 3.4605, 2.2846, 3.5629, 2.8870], device='cuda:0'), covar=tensor([0.0693, 0.0618, 0.1063, 0.1506, 0.0323, 0.1397, 0.0499, 0.0786], device='cuda:0'), in_proj_covar=tensor([0.0186, 0.0251, 0.0176, 0.0198, 0.0139, 0.0180, 0.0195, 0.0173], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 12:00:29,562 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=201301.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 12:00:38,772 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=201314.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 12:00:43,490 INFO [finetune.py:992] (0/2) Epoch 9, batch 3450, loss[loss=0.1993, simple_loss=0.2889, pruned_loss=0.05486, over 12007.00 frames. ], tot_loss[loss=0.1695, simple_loss=0.2581, pruned_loss=0.04049, over 2380947.72 frames. ], batch size: 42, lr: 4.18e-03, grad_scale: 32.0 2023-05-16 12:00:54,390 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=201335.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 12:01:04,232 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=201349.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 12:01:07,016 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.976e+02 2.752e+02 3.275e+02 4.140e+02 7.327e+02, threshold=6.550e+02, percent-clipped=2.0 2023-05-16 12:01:19,613 INFO [finetune.py:992] (0/2) Epoch 9, batch 3500, loss[loss=0.1558, simple_loss=0.2496, pruned_loss=0.03097, over 11857.00 frames. ], tot_loss[loss=0.1688, simple_loss=0.2571, pruned_loss=0.04028, over 2370803.31 frames. ], batch size: 26, lr: 4.18e-03, grad_scale: 32.0 2023-05-16 12:01:56,023 INFO [finetune.py:992] (0/2) Epoch 9, batch 3550, loss[loss=0.1732, simple_loss=0.2598, pruned_loss=0.04331, over 12103.00 frames. ], tot_loss[loss=0.17, simple_loss=0.258, pruned_loss=0.04099, over 2352889.27 frames. ], batch size: 32, lr: 4.18e-03, grad_scale: 32.0 2023-05-16 12:02:00,336 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=201427.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 12:02:19,017 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.012e+02 2.890e+02 3.315e+02 3.847e+02 6.638e+02, threshold=6.631e+02, percent-clipped=2.0 2023-05-16 12:02:27,306 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.39 vs. limit=2.0 2023-05-16 12:02:31,739 INFO [finetune.py:992] (0/2) Epoch 9, batch 3600, loss[loss=0.1634, simple_loss=0.2467, pruned_loss=0.04002, over 12038.00 frames. ], tot_loss[loss=0.1697, simple_loss=0.2577, pruned_loss=0.04086, over 2362462.15 frames. ], batch size: 31, lr: 4.18e-03, grad_scale: 32.0 2023-05-16 12:02:31,871 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.9416, 5.9167, 5.6672, 5.1787, 5.0916, 5.7918, 5.4004, 5.2411], device='cuda:0'), covar=tensor([0.0661, 0.0856, 0.0691, 0.1427, 0.0708, 0.0705, 0.1528, 0.1051], device='cuda:0'), in_proj_covar=tensor([0.0595, 0.0540, 0.0501, 0.0620, 0.0404, 0.0694, 0.0752, 0.0558], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0004, 0.0004, 0.0003], device='cuda:0') 2023-05-16 12:03:06,157 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.6837, 2.6311, 4.0879, 4.3279, 2.8156, 2.5238, 2.8137, 2.0868], device='cuda:0'), covar=tensor([0.1469, 0.2897, 0.0557, 0.0432, 0.1181, 0.2148, 0.2556, 0.4035], device='cuda:0'), in_proj_covar=tensor([0.0284, 0.0377, 0.0268, 0.0291, 0.0262, 0.0293, 0.0365, 0.0364], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-16 12:03:08,095 INFO [finetune.py:992] (0/2) Epoch 9, batch 3650, loss[loss=0.1578, simple_loss=0.2545, pruned_loss=0.03051, over 12112.00 frames. ], tot_loss[loss=0.1695, simple_loss=0.2577, pruned_loss=0.04072, over 2364033.72 frames. ], batch size: 33, lr: 4.18e-03, grad_scale: 32.0 2023-05-16 12:03:15,116 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.7499, 5.7169, 5.4948, 4.9699, 5.0083, 5.6264, 5.2417, 5.0527], device='cuda:0'), covar=tensor([0.0692, 0.0868, 0.0689, 0.1530, 0.0789, 0.0759, 0.1489, 0.1141], device='cuda:0'), in_proj_covar=tensor([0.0600, 0.0543, 0.0505, 0.0626, 0.0406, 0.0698, 0.0757, 0.0563], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0004, 0.0004, 0.0003], device='cuda:0') 2023-05-16 12:03:24,079 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.43 vs. limit=2.0 2023-05-16 12:03:29,227 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.38 vs. limit=2.0 2023-05-16 12:03:31,411 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.735e+02 2.865e+02 3.266e+02 3.969e+02 6.770e+02, threshold=6.532e+02, percent-clipped=0.0 2023-05-16 12:03:44,098 INFO [finetune.py:992] (0/2) Epoch 9, batch 3700, loss[loss=0.1684, simple_loss=0.2495, pruned_loss=0.0436, over 12020.00 frames. ], tot_loss[loss=0.1699, simple_loss=0.258, pruned_loss=0.04083, over 2369731.38 frames. ], batch size: 31, lr: 4.18e-03, grad_scale: 32.0 2023-05-16 12:04:11,421 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=201609.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 12:04:19,576 INFO [finetune.py:992] (0/2) Epoch 9, batch 3750, loss[loss=0.1589, simple_loss=0.2421, pruned_loss=0.0379, over 12133.00 frames. ], tot_loss[loss=0.1689, simple_loss=0.2567, pruned_loss=0.04054, over 2364500.88 frames. ], batch size: 30, lr: 4.18e-03, grad_scale: 32.0 2023-05-16 12:04:26,775 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=201630.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 12:04:36,854 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=201644.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 12:04:41,362 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=2.75 vs. limit=5.0 2023-05-16 12:04:43,132 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.902e+02 2.871e+02 3.276e+02 3.780e+02 6.507e+02, threshold=6.551e+02, percent-clipped=1.0 2023-05-16 12:04:48,970 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=201661.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 12:04:55,989 INFO [finetune.py:992] (0/2) Epoch 9, batch 3800, loss[loss=0.1592, simple_loss=0.2372, pruned_loss=0.0406, over 12358.00 frames. ], tot_loss[loss=0.1686, simple_loss=0.2568, pruned_loss=0.04014, over 2370663.63 frames. ], batch size: 30, lr: 4.18e-03, grad_scale: 32.0 2023-05-16 12:05:21,081 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=201705.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 12:05:21,810 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=201706.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 12:05:32,302 INFO [finetune.py:992] (0/2) Epoch 9, batch 3850, loss[loss=0.1776, simple_loss=0.2657, pruned_loss=0.04479, over 12359.00 frames. ], tot_loss[loss=0.1693, simple_loss=0.2576, pruned_loss=0.04047, over 2364312.33 frames. ], batch size: 36, lr: 4.18e-03, grad_scale: 32.0 2023-05-16 12:05:33,221 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=201722.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 12:05:36,688 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=201727.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 12:05:48,880 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=3.35 vs. limit=5.0 2023-05-16 12:05:54,881 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.815e+02 2.704e+02 3.228e+02 3.890e+02 7.833e+02, threshold=6.456e+02, percent-clipped=1.0 2023-05-16 12:06:05,716 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=201767.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 12:06:08,319 INFO [finetune.py:992] (0/2) Epoch 9, batch 3900, loss[loss=0.1644, simple_loss=0.2568, pruned_loss=0.03601, over 12104.00 frames. ], tot_loss[loss=0.17, simple_loss=0.2582, pruned_loss=0.04084, over 2367956.54 frames. ], batch size: 33, lr: 4.18e-03, grad_scale: 32.0 2023-05-16 12:06:11,163 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=201775.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 12:06:43,893 INFO [finetune.py:992] (0/2) Epoch 9, batch 3950, loss[loss=0.2222, simple_loss=0.318, pruned_loss=0.06318, over 11994.00 frames. ], tot_loss[loss=0.1706, simple_loss=0.2591, pruned_loss=0.04111, over 2376194.03 frames. ], batch size: 42, lr: 4.18e-03, grad_scale: 32.0 2023-05-16 12:06:46,517 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.86 vs. limit=2.0 2023-05-16 12:07:07,292 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.902e+02 2.742e+02 3.284e+02 3.856e+02 7.422e+02, threshold=6.567e+02, percent-clipped=1.0 2023-05-16 12:07:20,252 INFO [finetune.py:992] (0/2) Epoch 9, batch 4000, loss[loss=0.1706, simple_loss=0.2586, pruned_loss=0.04128, over 12156.00 frames. ], tot_loss[loss=0.1704, simple_loss=0.2589, pruned_loss=0.04095, over 2379160.04 frames. ], batch size: 34, lr: 4.18e-03, grad_scale: 32.0 2023-05-16 12:07:27,565 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.2876, 3.9033, 3.9895, 4.3970, 3.0354, 4.0014, 2.4276, 4.0372], device='cuda:0'), covar=tensor([0.1596, 0.0822, 0.1034, 0.0661, 0.1188, 0.0567, 0.1933, 0.1101], device='cuda:0'), in_proj_covar=tensor([0.0221, 0.0261, 0.0289, 0.0350, 0.0233, 0.0233, 0.0256, 0.0372], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 12:07:36,384 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=4.70 vs. limit=5.0 2023-05-16 12:07:44,408 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.3018, 4.8434, 5.0595, 5.0687, 4.8566, 5.1101, 4.9222, 2.7520], device='cuda:0'), covar=tensor([0.0085, 0.0070, 0.0080, 0.0062, 0.0045, 0.0090, 0.0083, 0.0758], device='cuda:0'), in_proj_covar=tensor([0.0065, 0.0076, 0.0078, 0.0072, 0.0058, 0.0088, 0.0077, 0.0095], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-16 12:07:47,318 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.5618, 5.1458, 5.5694, 4.9253, 5.2132, 4.9566, 5.5953, 5.1448], device='cuda:0'), covar=tensor([0.0223, 0.0295, 0.0232, 0.0210, 0.0295, 0.0279, 0.0158, 0.0241], device='cuda:0'), in_proj_covar=tensor([0.0249, 0.0255, 0.0278, 0.0251, 0.0250, 0.0251, 0.0230, 0.0210], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 12:07:47,328 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=201909.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 12:07:56,303 INFO [finetune.py:992] (0/2) Epoch 9, batch 4050, loss[loss=0.1594, simple_loss=0.2482, pruned_loss=0.03528, over 12122.00 frames. ], tot_loss[loss=0.1708, simple_loss=0.2597, pruned_loss=0.04094, over 2382266.12 frames. ], batch size: 30, lr: 4.18e-03, grad_scale: 16.0 2023-05-16 12:08:02,895 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=201930.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 12:08:20,087 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.994e+02 2.634e+02 3.114e+02 4.085e+02 9.505e+02, threshold=6.228e+02, percent-clipped=2.0 2023-05-16 12:08:22,275 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=201957.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 12:08:31,937 INFO [finetune.py:992] (0/2) Epoch 9, batch 4100, loss[loss=0.1785, simple_loss=0.2639, pruned_loss=0.0465, over 12048.00 frames. ], tot_loss[loss=0.1705, simple_loss=0.2591, pruned_loss=0.04094, over 2377347.09 frames. ], batch size: 42, lr: 4.18e-03, grad_scale: 16.0 2023-05-16 12:08:37,877 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=201978.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 12:08:46,417 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=201990.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 12:08:53,747 INFO [checkpoint.py:75] (0/2) Saving checkpoint to pruned_transducer_stateless7/exp_giga_finetune/checkpoint-102000.pt 2023-05-16 12:08:56,477 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=202000.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 12:08:57,239 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.3843, 4.9642, 5.3622, 4.6721, 5.0290, 4.7165, 5.3985, 5.1101], device='cuda:0'), covar=tensor([0.0236, 0.0307, 0.0259, 0.0256, 0.0332, 0.0307, 0.0201, 0.0211], device='cuda:0'), in_proj_covar=tensor([0.0247, 0.0254, 0.0276, 0.0250, 0.0248, 0.0250, 0.0229, 0.0209], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 12:09:08,605 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=202017.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 12:09:11,429 INFO [finetune.py:992] (0/2) Epoch 9, batch 4150, loss[loss=0.16, simple_loss=0.2484, pruned_loss=0.03579, over 12110.00 frames. ], tot_loss[loss=0.1703, simple_loss=0.2585, pruned_loss=0.04102, over 2366271.77 frames. ], batch size: 33, lr: 4.18e-03, grad_scale: 16.0 2023-05-16 12:09:21,804 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=4.92 vs. limit=5.0 2023-05-16 12:09:32,611 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=202051.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 12:09:34,575 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.872e+02 2.800e+02 3.305e+02 4.028e+02 6.958e+02, threshold=6.610e+02, percent-clipped=2.0 2023-05-16 12:09:40,323 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=202062.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 12:09:47,166 INFO [finetune.py:992] (0/2) Epoch 9, batch 4200, loss[loss=0.1736, simple_loss=0.2639, pruned_loss=0.0416, over 12367.00 frames. ], tot_loss[loss=0.1701, simple_loss=0.2582, pruned_loss=0.041, over 2375194.50 frames. ], batch size: 35, lr: 4.18e-03, grad_scale: 16.0 2023-05-16 12:10:05,413 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.1498, 2.3989, 3.7002, 3.1091, 3.4667, 3.1953, 2.5343, 3.4894], device='cuda:0'), covar=tensor([0.0109, 0.0335, 0.0117, 0.0207, 0.0140, 0.0150, 0.0276, 0.0116], device='cuda:0'), in_proj_covar=tensor([0.0170, 0.0199, 0.0181, 0.0181, 0.0210, 0.0157, 0.0193, 0.0177], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 12:10:09,611 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.1223, 3.9323, 2.6371, 2.3245, 3.4790, 2.3033, 3.5640, 2.8565], device='cuda:0'), covar=tensor([0.0661, 0.0579, 0.0959, 0.1442, 0.0345, 0.1333, 0.0518, 0.0797], device='cuda:0'), in_proj_covar=tensor([0.0186, 0.0248, 0.0174, 0.0196, 0.0138, 0.0178, 0.0193, 0.0172], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 12:10:22,475 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-05-16 12:10:23,547 INFO [finetune.py:992] (0/2) Epoch 9, batch 4250, loss[loss=0.1511, simple_loss=0.2332, pruned_loss=0.03453, over 12016.00 frames. ], tot_loss[loss=0.1708, simple_loss=0.2591, pruned_loss=0.04121, over 2369571.44 frames. ], batch size: 28, lr: 4.18e-03, grad_scale: 16.0 2023-05-16 12:10:26,759 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.71 vs. limit=2.0 2023-05-16 12:10:39,533 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.42 vs. limit=2.0 2023-05-16 12:10:41,561 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=2.24 vs. limit=5.0 2023-05-16 12:10:46,835 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.965e+02 2.761e+02 3.356e+02 3.983e+02 6.332e+02, threshold=6.712e+02, percent-clipped=0.0 2023-05-16 12:10:57,629 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.9967, 3.4533, 5.2611, 2.7687, 3.1157, 3.9736, 3.2798, 3.9035], device='cuda:0'), covar=tensor([0.0356, 0.1053, 0.0394, 0.1205, 0.1697, 0.1282, 0.1311, 0.1090], device='cuda:0'), in_proj_covar=tensor([0.0235, 0.0233, 0.0242, 0.0183, 0.0236, 0.0291, 0.0224, 0.0258], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-16 12:10:58,811 INFO [finetune.py:992] (0/2) Epoch 9, batch 4300, loss[loss=0.167, simple_loss=0.2494, pruned_loss=0.04232, over 11843.00 frames. ], tot_loss[loss=0.1707, simple_loss=0.2591, pruned_loss=0.04118, over 2377238.64 frames. ], batch size: 26, lr: 4.18e-03, grad_scale: 16.0 2023-05-16 12:11:06,845 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.6056, 2.6213, 4.2707, 4.4000, 2.8679, 2.5174, 2.7087, 2.0882], device='cuda:0'), covar=tensor([0.1627, 0.2864, 0.0491, 0.0458, 0.1235, 0.2414, 0.2762, 0.4085], device='cuda:0'), in_proj_covar=tensor([0.0286, 0.0379, 0.0268, 0.0292, 0.0263, 0.0293, 0.0369, 0.0366], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-16 12:11:07,395 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.0179, 5.9655, 5.4284, 5.5223, 5.9982, 5.2513, 5.5359, 5.5130], device='cuda:0'), covar=tensor([0.1273, 0.0951, 0.0983, 0.1657, 0.0878, 0.2107, 0.1744, 0.1049], device='cuda:0'), in_proj_covar=tensor([0.0353, 0.0475, 0.0378, 0.0428, 0.0452, 0.0429, 0.0383, 0.0371], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 12:11:31,446 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.0656, 4.5817, 4.7570, 4.8280, 4.6150, 4.8088, 4.7666, 2.6494], device='cuda:0'), covar=tensor([0.0080, 0.0077, 0.0086, 0.0077, 0.0059, 0.0115, 0.0085, 0.0775], device='cuda:0'), in_proj_covar=tensor([0.0065, 0.0075, 0.0078, 0.0071, 0.0058, 0.0088, 0.0077, 0.0095], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-16 12:11:34,813 INFO [finetune.py:992] (0/2) Epoch 9, batch 4350, loss[loss=0.1245, simple_loss=0.2196, pruned_loss=0.01473, over 12086.00 frames. ], tot_loss[loss=0.171, simple_loss=0.2596, pruned_loss=0.04115, over 2377162.35 frames. ], batch size: 32, lr: 4.17e-03, grad_scale: 16.0 2023-05-16 12:11:35,943 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.61 vs. limit=2.0 2023-05-16 12:11:58,306 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.986e+02 2.751e+02 3.212e+02 3.727e+02 1.205e+03, threshold=6.424e+02, percent-clipped=3.0 2023-05-16 12:12:10,924 INFO [finetune.py:992] (0/2) Epoch 9, batch 4400, loss[loss=0.1836, simple_loss=0.2778, pruned_loss=0.04468, over 10431.00 frames. ], tot_loss[loss=0.1707, simple_loss=0.2592, pruned_loss=0.04107, over 2364543.58 frames. ], batch size: 68, lr: 4.17e-03, grad_scale: 16.0 2023-05-16 12:12:24,597 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.0515, 3.4250, 5.3451, 2.7538, 2.8622, 3.9406, 3.3836, 3.9202], device='cuda:0'), covar=tensor([0.0388, 0.1066, 0.0215, 0.1197, 0.1947, 0.1343, 0.1252, 0.1103], device='cuda:0'), in_proj_covar=tensor([0.0235, 0.0233, 0.0241, 0.0183, 0.0236, 0.0290, 0.0223, 0.0259], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-16 12:12:31,111 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 2023-05-16 12:12:31,371 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=202300.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 12:12:43,453 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=202317.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 12:12:46,161 INFO [finetune.py:992] (0/2) Epoch 9, batch 4450, loss[loss=0.1679, simple_loss=0.263, pruned_loss=0.03638, over 11363.00 frames. ], tot_loss[loss=0.1702, simple_loss=0.2586, pruned_loss=0.0409, over 2368439.20 frames. ], batch size: 55, lr: 4.17e-03, grad_scale: 16.0 2023-05-16 12:13:03,772 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=202346.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 12:13:05,202 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=202348.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 12:13:06,931 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.3151, 4.5925, 4.0589, 4.8719, 4.5866, 2.9064, 4.3068, 3.1038], device='cuda:0'), covar=tensor([0.0800, 0.0709, 0.1378, 0.0536, 0.1068, 0.1596, 0.0983, 0.3203], device='cuda:0'), in_proj_covar=tensor([0.0301, 0.0377, 0.0356, 0.0289, 0.0366, 0.0265, 0.0338, 0.0362], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 12:13:10,073 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.794e+02 2.951e+02 3.496e+02 3.984e+02 8.134e+02, threshold=6.992e+02, percent-clipped=1.0 2023-05-16 12:13:10,192 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.9335, 5.8126, 5.3928, 5.3304, 5.9576, 5.1440, 5.4189, 5.4510], device='cuda:0'), covar=tensor([0.1481, 0.1002, 0.1122, 0.2082, 0.0873, 0.2167, 0.1714, 0.1169], device='cuda:0'), in_proj_covar=tensor([0.0352, 0.0475, 0.0376, 0.0426, 0.0450, 0.0427, 0.0380, 0.0369], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 12:13:10,602 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=4.44 vs. limit=5.0 2023-05-16 12:13:15,917 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=202362.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 12:13:17,968 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=202365.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 12:13:22,059 INFO [finetune.py:992] (0/2) Epoch 9, batch 4500, loss[loss=0.1752, simple_loss=0.2672, pruned_loss=0.04162, over 12247.00 frames. ], tot_loss[loss=0.1705, simple_loss=0.2589, pruned_loss=0.04103, over 2368081.81 frames. ], batch size: 32, lr: 4.17e-03, grad_scale: 16.0 2023-05-16 12:13:44,257 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.99 vs. limit=2.0 2023-05-16 12:13:50,923 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=202410.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 12:13:58,713 INFO [finetune.py:992] (0/2) Epoch 9, batch 4550, loss[loss=0.211, simple_loss=0.2872, pruned_loss=0.06739, over 8015.00 frames. ], tot_loss[loss=0.1706, simple_loss=0.2591, pruned_loss=0.041, over 2364782.00 frames. ], batch size: 99, lr: 4.17e-03, grad_scale: 16.0 2023-05-16 12:14:15,302 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.2681, 4.7435, 2.8474, 2.4317, 4.1729, 2.2471, 3.9868, 3.0358], device='cuda:0'), covar=tensor([0.0752, 0.0416, 0.1109, 0.1584, 0.0220, 0.1590, 0.0416, 0.0882], device='cuda:0'), in_proj_covar=tensor([0.0187, 0.0250, 0.0175, 0.0197, 0.0139, 0.0178, 0.0194, 0.0173], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 12:14:22,243 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.019e+02 2.840e+02 3.198e+02 3.742e+02 1.519e+03, threshold=6.396e+02, percent-clipped=2.0 2023-05-16 12:14:34,385 INFO [finetune.py:992] (0/2) Epoch 9, batch 4600, loss[loss=0.1388, simple_loss=0.2255, pruned_loss=0.02605, over 12338.00 frames. ], tot_loss[loss=0.1696, simple_loss=0.2583, pruned_loss=0.04038, over 2375782.12 frames. ], batch size: 30, lr: 4.17e-03, grad_scale: 16.0 2023-05-16 12:15:08,499 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.2256, 4.8003, 4.8719, 5.0414, 4.7806, 4.9381, 4.8881, 2.7670], device='cuda:0'), covar=tensor([0.0060, 0.0063, 0.0083, 0.0059, 0.0046, 0.0090, 0.0070, 0.0672], device='cuda:0'), in_proj_covar=tensor([0.0066, 0.0076, 0.0079, 0.0072, 0.0058, 0.0089, 0.0078, 0.0095], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 12:15:10,548 INFO [finetune.py:992] (0/2) Epoch 9, batch 4650, loss[loss=0.1612, simple_loss=0.2405, pruned_loss=0.04101, over 12344.00 frames. ], tot_loss[loss=0.1704, simple_loss=0.2589, pruned_loss=0.04096, over 2377352.54 frames. ], batch size: 30, lr: 4.17e-03, grad_scale: 16.0 2023-05-16 12:15:27,783 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.1414, 2.3740, 3.7005, 2.9818, 3.4557, 3.2224, 2.4594, 3.5357], device='cuda:0'), covar=tensor([0.0131, 0.0397, 0.0115, 0.0285, 0.0159, 0.0182, 0.0384, 0.0126], device='cuda:0'), in_proj_covar=tensor([0.0173, 0.0202, 0.0184, 0.0184, 0.0214, 0.0160, 0.0197, 0.0181], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 12:15:34,764 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.795e+02 2.899e+02 3.227e+02 3.837e+02 6.733e+02, threshold=6.454e+02, percent-clipped=1.0 2023-05-16 12:15:46,583 INFO [finetune.py:992] (0/2) Epoch 9, batch 4700, loss[loss=0.1881, simple_loss=0.2771, pruned_loss=0.04954, over 11737.00 frames. ], tot_loss[loss=0.17, simple_loss=0.2584, pruned_loss=0.04083, over 2382762.98 frames. ], batch size: 44, lr: 4.17e-03, grad_scale: 16.0 2023-05-16 12:16:03,902 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.0780, 3.8406, 3.8624, 4.2444, 2.7718, 3.7087, 2.5717, 3.8017], device='cuda:0'), covar=tensor([0.1917, 0.0865, 0.1057, 0.0735, 0.1475, 0.0745, 0.2053, 0.1349], device='cuda:0'), in_proj_covar=tensor([0.0223, 0.0263, 0.0292, 0.0354, 0.0235, 0.0234, 0.0257, 0.0376], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 12:16:22,521 INFO [finetune.py:992] (0/2) Epoch 9, batch 4750, loss[loss=0.1629, simple_loss=0.2556, pruned_loss=0.03506, over 12095.00 frames. ], tot_loss[loss=0.1708, simple_loss=0.2591, pruned_loss=0.04124, over 2381881.44 frames. ], batch size: 33, lr: 4.17e-03, grad_scale: 16.0 2023-05-16 12:16:32,000 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.6100, 2.6799, 3.3397, 4.5792, 2.5500, 4.5805, 4.5520, 4.6956], device='cuda:0'), covar=tensor([0.0114, 0.1207, 0.0471, 0.0136, 0.1287, 0.0208, 0.0125, 0.0073], device='cuda:0'), in_proj_covar=tensor([0.0119, 0.0203, 0.0184, 0.0116, 0.0189, 0.0176, 0.0174, 0.0118], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 12:16:40,971 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=202646.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 12:16:46,561 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.122e+02 2.882e+02 3.213e+02 4.113e+02 1.161e+03, threshold=6.425e+02, percent-clipped=2.0 2023-05-16 12:16:58,512 INFO [finetune.py:992] (0/2) Epoch 9, batch 4800, loss[loss=0.1484, simple_loss=0.2281, pruned_loss=0.03437, over 12266.00 frames. ], tot_loss[loss=0.1709, simple_loss=0.2593, pruned_loss=0.04128, over 2371323.93 frames. ], batch size: 28, lr: 4.17e-03, grad_scale: 16.0 2023-05-16 12:17:15,854 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=202694.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 12:17:28,772 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.7508, 4.4817, 4.5039, 4.5988, 4.5284, 4.5657, 4.4647, 2.5397], device='cuda:0'), covar=tensor([0.0188, 0.0106, 0.0148, 0.0131, 0.0078, 0.0162, 0.0186, 0.0950], device='cuda:0'), in_proj_covar=tensor([0.0065, 0.0076, 0.0079, 0.0071, 0.0058, 0.0088, 0.0078, 0.0095], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 12:17:34,873 INFO [finetune.py:992] (0/2) Epoch 9, batch 4850, loss[loss=0.1537, simple_loss=0.2517, pruned_loss=0.02786, over 12191.00 frames. ], tot_loss[loss=0.1714, simple_loss=0.2596, pruned_loss=0.04161, over 2371753.38 frames. ], batch size: 35, lr: 4.17e-03, grad_scale: 16.0 2023-05-16 12:17:36,017 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.13 vs. limit=2.0 2023-05-16 12:17:50,151 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.37 vs. limit=2.0 2023-05-16 12:17:58,003 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.973e+02 2.677e+02 3.036e+02 3.907e+02 8.573e+02, threshold=6.072e+02, percent-clipped=3.0 2023-05-16 12:18:10,313 INFO [finetune.py:992] (0/2) Epoch 9, batch 4900, loss[loss=0.174, simple_loss=0.2634, pruned_loss=0.04233, over 12149.00 frames. ], tot_loss[loss=0.1709, simple_loss=0.2593, pruned_loss=0.04123, over 2385770.12 frames. ], batch size: 36, lr: 4.17e-03, grad_scale: 16.0 2023-05-16 12:18:46,623 INFO [finetune.py:992] (0/2) Epoch 9, batch 4950, loss[loss=0.193, simple_loss=0.2723, pruned_loss=0.05679, over 11321.00 frames. ], tot_loss[loss=0.1714, simple_loss=0.2596, pruned_loss=0.04159, over 2377987.38 frames. ], batch size: 55, lr: 4.17e-03, grad_scale: 16.0 2023-05-16 12:18:53,766 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=202830.0, num_to_drop=1, layers_to_drop={1} 2023-05-16 12:19:10,492 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.853e+02 2.757e+02 3.415e+02 4.105e+02 1.382e+03, threshold=6.831e+02, percent-clipped=3.0 2023-05-16 12:19:15,712 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.5938, 2.4557, 3.2578, 4.5635, 2.3152, 4.4911, 4.5668, 4.7060], device='cuda:0'), covar=tensor([0.0125, 0.1327, 0.0527, 0.0121, 0.1418, 0.0206, 0.0152, 0.0084], device='cuda:0'), in_proj_covar=tensor([0.0119, 0.0204, 0.0184, 0.0117, 0.0190, 0.0177, 0.0175, 0.0119], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 12:19:22,525 INFO [finetune.py:992] (0/2) Epoch 9, batch 5000, loss[loss=0.1451, simple_loss=0.2326, pruned_loss=0.02881, over 12380.00 frames. ], tot_loss[loss=0.1716, simple_loss=0.2595, pruned_loss=0.04184, over 2373513.81 frames. ], batch size: 30, lr: 4.17e-03, grad_scale: 16.0 2023-05-16 12:19:36,890 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=202891.0, num_to_drop=1, layers_to_drop={0} 2023-05-16 12:19:41,241 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.2406, 4.0813, 4.1011, 4.4094, 3.0253, 4.0540, 2.7999, 4.1714], device='cuda:0'), covar=tensor([0.1477, 0.0692, 0.0964, 0.0656, 0.1166, 0.0541, 0.1654, 0.1184], device='cuda:0'), in_proj_covar=tensor([0.0221, 0.0260, 0.0289, 0.0350, 0.0233, 0.0232, 0.0254, 0.0371], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 12:19:57,691 INFO [finetune.py:992] (0/2) Epoch 9, batch 5050, loss[loss=0.1849, simple_loss=0.2655, pruned_loss=0.05216, over 12328.00 frames. ], tot_loss[loss=0.1711, simple_loss=0.259, pruned_loss=0.04161, over 2374649.25 frames. ], batch size: 36, lr: 4.17e-03, grad_scale: 16.0 2023-05-16 12:20:09,788 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2023-05-16 12:20:21,655 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.657e+02 2.730e+02 3.199e+02 3.913e+02 1.181e+03, threshold=6.397e+02, percent-clipped=3.0 2023-05-16 12:20:29,084 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.5894, 2.7615, 4.3849, 4.5762, 2.7893, 2.5360, 2.8208, 2.1535], device='cuda:0'), covar=tensor([0.1570, 0.3048, 0.0521, 0.0450, 0.1327, 0.2380, 0.2909, 0.3972], device='cuda:0'), in_proj_covar=tensor([0.0289, 0.0385, 0.0273, 0.0296, 0.0267, 0.0297, 0.0373, 0.0369], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-16 12:20:33,892 INFO [finetune.py:992] (0/2) Epoch 9, batch 5100, loss[loss=0.1619, simple_loss=0.2468, pruned_loss=0.03855, over 12342.00 frames. ], tot_loss[loss=0.171, simple_loss=0.2589, pruned_loss=0.04153, over 2371457.33 frames. ], batch size: 31, lr: 4.17e-03, grad_scale: 16.0 2023-05-16 12:20:51,995 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.4351, 3.5017, 3.1884, 3.1044, 2.8611, 2.5830, 3.6358, 2.2169], device='cuda:0'), covar=tensor([0.0375, 0.0151, 0.0218, 0.0198, 0.0444, 0.0387, 0.0131, 0.0493], device='cuda:0'), in_proj_covar=tensor([0.0190, 0.0161, 0.0156, 0.0185, 0.0199, 0.0197, 0.0166, 0.0199], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 12:21:10,325 INFO [finetune.py:992] (0/2) Epoch 9, batch 5150, loss[loss=0.2157, simple_loss=0.2931, pruned_loss=0.06919, over 7753.00 frames. ], tot_loss[loss=0.171, simple_loss=0.2592, pruned_loss=0.04145, over 2373456.00 frames. ], batch size: 101, lr: 4.17e-03, grad_scale: 16.0 2023-05-16 12:21:15,484 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.4942, 2.2918, 3.2048, 4.5358, 2.3433, 4.5246, 4.4987, 4.5969], device='cuda:0'), covar=tensor([0.0155, 0.1404, 0.0508, 0.0150, 0.1392, 0.0198, 0.0170, 0.0104], device='cuda:0'), in_proj_covar=tensor([0.0119, 0.0204, 0.0184, 0.0117, 0.0191, 0.0177, 0.0175, 0.0119], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 12:21:22,554 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=203038.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 12:21:33,501 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.075e+02 2.774e+02 3.219e+02 3.722e+02 9.351e+02, threshold=6.437e+02, percent-clipped=3.0 2023-05-16 12:21:43,601 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=2.62 vs. limit=5.0 2023-05-16 12:21:45,399 INFO [finetune.py:992] (0/2) Epoch 9, batch 5200, loss[loss=0.1782, simple_loss=0.2727, pruned_loss=0.04182, over 12363.00 frames. ], tot_loss[loss=0.1715, simple_loss=0.2598, pruned_loss=0.04162, over 2364371.09 frames. ], batch size: 35, lr: 4.17e-03, grad_scale: 16.0 2023-05-16 12:21:56,029 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=203085.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 12:22:05,963 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=203099.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 12:22:08,799 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.3617, 3.3292, 3.0709, 3.0218, 2.7351, 2.5477, 3.3826, 2.1511], device='cuda:0'), covar=tensor([0.0389, 0.0141, 0.0206, 0.0199, 0.0408, 0.0383, 0.0150, 0.0499], device='cuda:0'), in_proj_covar=tensor([0.0190, 0.0162, 0.0156, 0.0186, 0.0201, 0.0197, 0.0166, 0.0200], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 12:22:21,518 INFO [finetune.py:992] (0/2) Epoch 9, batch 5250, loss[loss=0.1618, simple_loss=0.2474, pruned_loss=0.03813, over 12088.00 frames. ], tot_loss[loss=0.1712, simple_loss=0.2597, pruned_loss=0.04137, over 2371329.79 frames. ], batch size: 32, lr: 4.17e-03, grad_scale: 16.0 2023-05-16 12:22:37,326 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.3338, 4.8432, 5.3170, 4.6632, 4.9528, 4.7238, 5.3747, 4.9583], device='cuda:0'), covar=tensor([0.0247, 0.0355, 0.0261, 0.0262, 0.0361, 0.0299, 0.0183, 0.0272], device='cuda:0'), in_proj_covar=tensor([0.0250, 0.0259, 0.0281, 0.0253, 0.0253, 0.0253, 0.0234, 0.0212], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 12:22:40,056 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=203146.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 12:22:45,579 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.827e+02 2.746e+02 3.239e+02 3.710e+02 7.459e+02, threshold=6.478e+02, percent-clipped=1.0 2023-05-16 12:22:57,865 INFO [finetune.py:992] (0/2) Epoch 9, batch 5300, loss[loss=0.187, simple_loss=0.2769, pruned_loss=0.04854, over 11817.00 frames. ], tot_loss[loss=0.1708, simple_loss=0.2594, pruned_loss=0.04109, over 2371614.54 frames. ], batch size: 44, lr: 4.16e-03, grad_scale: 16.0 2023-05-16 12:23:02,364 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([6.1168, 6.0400, 5.8490, 5.3009, 5.1826, 6.0014, 5.6158, 5.3876], device='cuda:0'), covar=tensor([0.0584, 0.0882, 0.0615, 0.1509, 0.0645, 0.0703, 0.1445, 0.0991], device='cuda:0'), in_proj_covar=tensor([0.0605, 0.0540, 0.0501, 0.0629, 0.0403, 0.0700, 0.0756, 0.0562], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0004, 0.0004, 0.0003], device='cuda:0') 2023-05-16 12:23:05,363 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.69 vs. limit=2.0 2023-05-16 12:23:07,472 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.6674, 3.1332, 5.0032, 2.6206, 2.5703, 3.7192, 3.2338, 3.8114], device='cuda:0'), covar=tensor([0.0454, 0.1182, 0.0359, 0.1184, 0.2010, 0.1402, 0.1393, 0.1092], device='cuda:0'), in_proj_covar=tensor([0.0235, 0.0231, 0.0242, 0.0182, 0.0234, 0.0288, 0.0222, 0.0258], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-16 12:23:08,759 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=203186.0, num_to_drop=1, layers_to_drop={0} 2023-05-16 12:23:21,060 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.2226, 3.1631, 2.9377, 2.8740, 2.6315, 2.4317, 3.2147, 2.0739], device='cuda:0'), covar=tensor([0.0386, 0.0151, 0.0207, 0.0188, 0.0400, 0.0362, 0.0142, 0.0459], device='cuda:0'), in_proj_covar=tensor([0.0187, 0.0160, 0.0155, 0.0184, 0.0198, 0.0195, 0.0165, 0.0198], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 12:23:33,610 INFO [finetune.py:992] (0/2) Epoch 9, batch 5350, loss[loss=0.1441, simple_loss=0.2278, pruned_loss=0.03015, over 12179.00 frames. ], tot_loss[loss=0.1706, simple_loss=0.2594, pruned_loss=0.04083, over 2375824.72 frames. ], batch size: 29, lr: 4.16e-03, grad_scale: 16.0 2023-05-16 12:23:57,471 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.805e+02 2.707e+02 3.219e+02 3.949e+02 9.779e+02, threshold=6.438e+02, percent-clipped=1.0 2023-05-16 12:24:09,256 INFO [finetune.py:992] (0/2) Epoch 9, batch 5400, loss[loss=0.1679, simple_loss=0.2596, pruned_loss=0.03812, over 12149.00 frames. ], tot_loss[loss=0.1698, simple_loss=0.2585, pruned_loss=0.04054, over 2384541.64 frames. ], batch size: 34, lr: 4.16e-03, grad_scale: 16.0 2023-05-16 12:24:40,081 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.7431, 3.1981, 5.0618, 2.5911, 2.6792, 3.7378, 3.2335, 3.8216], device='cuda:0'), covar=tensor([0.0387, 0.1061, 0.0358, 0.1164, 0.1960, 0.1382, 0.1302, 0.1079], device='cuda:0'), in_proj_covar=tensor([0.0234, 0.0231, 0.0242, 0.0182, 0.0234, 0.0287, 0.0221, 0.0258], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-16 12:24:44,760 INFO [finetune.py:992] (0/2) Epoch 9, batch 5450, loss[loss=0.1629, simple_loss=0.2537, pruned_loss=0.03602, over 12153.00 frames. ], tot_loss[loss=0.1702, simple_loss=0.2585, pruned_loss=0.04089, over 2374541.48 frames. ], batch size: 34, lr: 4.16e-03, grad_scale: 16.0 2023-05-16 12:25:08,197 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.016e+02 2.673e+02 3.050e+02 3.823e+02 9.002e+02, threshold=6.100e+02, percent-clipped=2.0 2023-05-16 12:25:18,745 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.2404, 5.0853, 5.1566, 5.2448, 4.8023, 4.8560, 4.6200, 5.1662], device='cuda:0'), covar=tensor([0.0669, 0.0568, 0.0907, 0.0541, 0.1934, 0.1418, 0.0561, 0.0995], device='cuda:0'), in_proj_covar=tensor([0.0524, 0.0689, 0.0601, 0.0609, 0.0845, 0.0733, 0.0545, 0.0483], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0005, 0.0004, 0.0004, 0.0005, 0.0005, 0.0004, 0.0003], device='cuda:0') 2023-05-16 12:25:19,990 INFO [finetune.py:992] (0/2) Epoch 9, batch 5500, loss[loss=0.1882, simple_loss=0.2903, pruned_loss=0.04305, over 10552.00 frames. ], tot_loss[loss=0.1706, simple_loss=0.2589, pruned_loss=0.04117, over 2373774.63 frames. ], batch size: 68, lr: 4.16e-03, grad_scale: 16.0 2023-05-16 12:25:36,865 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=203394.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 12:25:42,363 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-05-16 12:25:56,688 INFO [finetune.py:992] (0/2) Epoch 9, batch 5550, loss[loss=0.1512, simple_loss=0.238, pruned_loss=0.03219, over 12341.00 frames. ], tot_loss[loss=0.1704, simple_loss=0.2588, pruned_loss=0.04105, over 2371910.14 frames. ], batch size: 31, lr: 4.16e-03, grad_scale: 16.0 2023-05-16 12:26:11,067 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=203441.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 12:26:20,183 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.796e+02 2.800e+02 3.328e+02 3.811e+02 1.182e+03, threshold=6.657e+02, percent-clipped=6.0 2023-05-16 12:26:29,469 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=203467.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 12:26:32,193 INFO [finetune.py:992] (0/2) Epoch 9, batch 5600, loss[loss=0.144, simple_loss=0.2274, pruned_loss=0.03025, over 12179.00 frames. ], tot_loss[loss=0.1701, simple_loss=0.2585, pruned_loss=0.04089, over 2370571.38 frames. ], batch size: 29, lr: 4.16e-03, grad_scale: 16.0 2023-05-16 12:26:43,006 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=203486.0, num_to_drop=1, layers_to_drop={1} 2023-05-16 12:26:46,492 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.7248, 2.9657, 5.0320, 2.7148, 2.7244, 3.8245, 3.1511, 3.9377], device='cuda:0'), covar=tensor([0.0440, 0.1401, 0.0369, 0.1189, 0.2030, 0.1404, 0.1443, 0.1076], device='cuda:0'), in_proj_covar=tensor([0.0236, 0.0234, 0.0245, 0.0184, 0.0237, 0.0290, 0.0224, 0.0260], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-16 12:27:07,445 INFO [finetune.py:992] (0/2) Epoch 9, batch 5650, loss[loss=0.1734, simple_loss=0.2671, pruned_loss=0.03987, over 12165.00 frames. ], tot_loss[loss=0.1709, simple_loss=0.259, pruned_loss=0.04136, over 2374158.75 frames. ], batch size: 34, lr: 4.16e-03, grad_scale: 16.0 2023-05-16 12:27:12,638 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=203528.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 12:27:17,372 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=203534.0, num_to_drop=1, layers_to_drop={0} 2023-05-16 12:27:31,138 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.190e+02 2.781e+02 3.248e+02 4.061e+02 7.168e+02, threshold=6.495e+02, percent-clipped=2.0 2023-05-16 12:27:44,047 INFO [finetune.py:992] (0/2) Epoch 9, batch 5700, loss[loss=0.1904, simple_loss=0.2678, pruned_loss=0.05646, over 12103.00 frames. ], tot_loss[loss=0.1705, simple_loss=0.2584, pruned_loss=0.0413, over 2373069.27 frames. ], batch size: 42, lr: 4.16e-03, grad_scale: 16.0 2023-05-16 12:28:09,033 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.5395, 5.0525, 5.5265, 4.8275, 5.1474, 4.8731, 5.5578, 5.1348], device='cuda:0'), covar=tensor([0.0228, 0.0335, 0.0239, 0.0219, 0.0318, 0.0259, 0.0174, 0.0210], device='cuda:0'), in_proj_covar=tensor([0.0250, 0.0257, 0.0282, 0.0253, 0.0253, 0.0254, 0.0233, 0.0211], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 12:28:19,510 INFO [finetune.py:992] (0/2) Epoch 9, batch 5750, loss[loss=0.2429, simple_loss=0.3131, pruned_loss=0.08633, over 8240.00 frames. ], tot_loss[loss=0.1718, simple_loss=0.2595, pruned_loss=0.04204, over 2370670.41 frames. ], batch size: 98, lr: 4.16e-03, grad_scale: 16.0 2023-05-16 12:28:42,798 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.803e+02 2.683e+02 3.193e+02 4.148e+02 8.960e+02, threshold=6.386e+02, percent-clipped=3.0 2023-05-16 12:28:49,762 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([6.0057, 5.9875, 5.7874, 5.2340, 5.1631, 5.9228, 5.5380, 5.3391], device='cuda:0'), covar=tensor([0.0731, 0.0932, 0.0628, 0.1642, 0.0632, 0.0746, 0.1632, 0.0931], device='cuda:0'), in_proj_covar=tensor([0.0603, 0.0534, 0.0496, 0.0620, 0.0401, 0.0699, 0.0753, 0.0559], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0004, 0.0004, 0.0003], device='cuda:0') 2023-05-16 12:28:49,881 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.1781, 2.4712, 3.7427, 3.1064, 3.5144, 3.1621, 2.5659, 3.6660], device='cuda:0'), covar=tensor([0.0124, 0.0357, 0.0117, 0.0203, 0.0131, 0.0180, 0.0328, 0.0102], device='cuda:0'), in_proj_covar=tensor([0.0172, 0.0201, 0.0183, 0.0181, 0.0212, 0.0159, 0.0194, 0.0178], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 12:28:54,513 INFO [finetune.py:992] (0/2) Epoch 9, batch 5800, loss[loss=0.1915, simple_loss=0.2729, pruned_loss=0.055, over 12049.00 frames. ], tot_loss[loss=0.1727, simple_loss=0.2606, pruned_loss=0.04246, over 2365633.43 frames. ], batch size: 40, lr: 4.16e-03, grad_scale: 16.0 2023-05-16 12:29:11,675 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=203694.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 12:29:29,365 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.7339, 2.7464, 4.6103, 4.7944, 2.7952, 2.7206, 2.9980, 2.2589], device='cuda:0'), covar=tensor([0.1487, 0.3102, 0.0450, 0.0423, 0.1248, 0.2231, 0.2576, 0.3770], device='cuda:0'), in_proj_covar=tensor([0.0286, 0.0382, 0.0270, 0.0293, 0.0265, 0.0295, 0.0369, 0.0364], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-16 12:29:31,173 INFO [finetune.py:992] (0/2) Epoch 9, batch 5850, loss[loss=0.1737, simple_loss=0.2611, pruned_loss=0.04316, over 12069.00 frames. ], tot_loss[loss=0.1722, simple_loss=0.2602, pruned_loss=0.0421, over 2372701.60 frames. ], batch size: 42, lr: 4.16e-03, grad_scale: 16.0 2023-05-16 12:29:39,349 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=4.46 vs. limit=5.0 2023-05-16 12:29:45,448 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=203741.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 12:29:46,061 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=203742.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 12:29:49,787 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=203747.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 12:29:54,425 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.934e+02 2.770e+02 3.365e+02 3.985e+02 7.334e+02, threshold=6.729e+02, percent-clipped=1.0 2023-05-16 12:29:56,080 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.3231, 4.7354, 3.0223, 2.7007, 4.1031, 2.8519, 4.0643, 3.5057], device='cuda:0'), covar=tensor([0.0643, 0.0485, 0.1037, 0.1427, 0.0283, 0.1158, 0.0413, 0.0648], device='cuda:0'), in_proj_covar=tensor([0.0186, 0.0252, 0.0175, 0.0198, 0.0139, 0.0179, 0.0195, 0.0173], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 12:29:58,345 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.97 vs. limit=2.0 2023-05-16 12:30:03,037 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=203766.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 12:30:06,239 INFO [finetune.py:992] (0/2) Epoch 9, batch 5900, loss[loss=0.1731, simple_loss=0.2651, pruned_loss=0.04049, over 11699.00 frames. ], tot_loss[loss=0.1714, simple_loss=0.2596, pruned_loss=0.04165, over 2379496.09 frames. ], batch size: 48, lr: 4.16e-03, grad_scale: 16.0 2023-05-16 12:30:18,966 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=203789.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 12:30:25,093 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.43 vs. limit=2.0 2023-05-16 12:30:32,993 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=203808.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 12:30:42,940 INFO [finetune.py:992] (0/2) Epoch 9, batch 5950, loss[loss=0.1704, simple_loss=0.2623, pruned_loss=0.03922, over 12189.00 frames. ], tot_loss[loss=0.1718, simple_loss=0.2599, pruned_loss=0.04186, over 2373564.87 frames. ], batch size: 35, lr: 4.16e-03, grad_scale: 16.0 2023-05-16 12:30:44,071 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.93 vs. limit=2.0 2023-05-16 12:30:44,455 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=203823.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 12:30:47,345 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=203827.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 12:31:02,353 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=203847.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 12:31:07,172 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.843e+02 2.790e+02 3.214e+02 3.808e+02 7.515e+02, threshold=6.427e+02, percent-clipped=3.0 2023-05-16 12:31:16,719 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=2.31 vs. limit=5.0 2023-05-16 12:31:19,275 INFO [finetune.py:992] (0/2) Epoch 9, batch 6000, loss[loss=0.1497, simple_loss=0.2382, pruned_loss=0.03054, over 12122.00 frames. ], tot_loss[loss=0.1719, simple_loss=0.2602, pruned_loss=0.04183, over 2378812.00 frames. ], batch size: 30, lr: 4.16e-03, grad_scale: 16.0 2023-05-16 12:31:19,275 INFO [finetune.py:1017] (0/2) Computing validation loss 2023-05-16 12:31:32,295 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.5306, 2.1519, 2.8325, 3.5839, 2.2183, 3.6689, 3.2637, 3.6302], device='cuda:0'), covar=tensor([0.0124, 0.1256, 0.0513, 0.0149, 0.1271, 0.0162, 0.0305, 0.0125], device='cuda:0'), in_proj_covar=tensor([0.0120, 0.0203, 0.0182, 0.0115, 0.0189, 0.0176, 0.0173, 0.0119], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 12:31:37,762 INFO [finetune.py:1026] (0/2) Epoch 9, validation: loss=0.3191, simple_loss=0.3958, pruned_loss=0.1212, over 1020973.00 frames. 2023-05-16 12:31:37,763 INFO [finetune.py:1027] (0/2) Maximum memory allocated so far is 12761MB 2023-05-16 12:31:58,150 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=203900.0, num_to_drop=1, layers_to_drop={0} 2023-05-16 12:32:03,691 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=203908.0, num_to_drop=1, layers_to_drop={1} 2023-05-16 12:32:13,531 INFO [finetune.py:992] (0/2) Epoch 9, batch 6050, loss[loss=0.1504, simple_loss=0.2439, pruned_loss=0.02847, over 12111.00 frames. ], tot_loss[loss=0.1721, simple_loss=0.2605, pruned_loss=0.04188, over 2380608.47 frames. ], batch size: 33, lr: 4.16e-03, grad_scale: 32.0 2023-05-16 12:32:37,126 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.072e+02 2.790e+02 3.279e+02 3.879e+02 6.411e+02, threshold=6.558e+02, percent-clipped=0.0 2023-05-16 12:32:41,517 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.0975, 5.9017, 5.3579, 5.3883, 5.9875, 5.3883, 5.4250, 5.4901], device='cuda:0'), covar=tensor([0.1475, 0.0931, 0.1111, 0.2040, 0.0898, 0.1970, 0.1833, 0.0968], device='cuda:0'), in_proj_covar=tensor([0.0357, 0.0479, 0.0382, 0.0433, 0.0452, 0.0430, 0.0387, 0.0373], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 12:32:42,353 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=203961.0, num_to_drop=1, layers_to_drop={0} 2023-05-16 12:32:49,130 INFO [finetune.py:992] (0/2) Epoch 9, batch 6100, loss[loss=0.1468, simple_loss=0.2345, pruned_loss=0.02956, over 12336.00 frames. ], tot_loss[loss=0.1731, simple_loss=0.2614, pruned_loss=0.04238, over 2374146.63 frames. ], batch size: 31, lr: 4.16e-03, grad_scale: 32.0 2023-05-16 12:33:10,142 INFO [checkpoint.py:75] (0/2) Saving checkpoint to pruned_transducer_stateless7/exp_giga_finetune/checkpoint-104000.pt 2023-05-16 12:33:19,281 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.7512, 5.4331, 4.9149, 5.0183, 5.5354, 4.8529, 5.0229, 4.9912], device='cuda:0'), covar=tensor([0.1531, 0.0946, 0.0973, 0.1839, 0.0942, 0.2114, 0.1677, 0.1090], device='cuda:0'), in_proj_covar=tensor([0.0358, 0.0481, 0.0382, 0.0436, 0.0454, 0.0431, 0.0388, 0.0374], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 12:33:27,556 INFO [finetune.py:992] (0/2) Epoch 9, batch 6150, loss[loss=0.173, simple_loss=0.261, pruned_loss=0.04256, over 12129.00 frames. ], tot_loss[loss=0.1723, simple_loss=0.2606, pruned_loss=0.04206, over 2377437.33 frames. ], batch size: 38, lr: 4.16e-03, grad_scale: 16.0 2023-05-16 12:33:33,653 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=2.96 vs. limit=5.0 2023-05-16 12:33:51,600 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.027e+02 2.976e+02 3.458e+02 4.063e+02 7.171e+02, threshold=6.916e+02, percent-clipped=1.0 2023-05-16 12:34:03,550 INFO [finetune.py:992] (0/2) Epoch 9, batch 6200, loss[loss=0.2353, simple_loss=0.304, pruned_loss=0.08329, over 7788.00 frames. ], tot_loss[loss=0.1737, simple_loss=0.2615, pruned_loss=0.04288, over 2363679.06 frames. ], batch size: 99, lr: 4.16e-03, grad_scale: 16.0 2023-05-16 12:34:07,999 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=204077.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 12:34:08,895 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 2023-05-16 12:34:27,061 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=204103.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 12:34:39,446 INFO [finetune.py:992] (0/2) Epoch 9, batch 6250, loss[loss=0.1439, simple_loss=0.2256, pruned_loss=0.03104, over 12022.00 frames. ], tot_loss[loss=0.1725, simple_loss=0.2604, pruned_loss=0.04226, over 2368958.96 frames. ], batch size: 28, lr: 4.15e-03, grad_scale: 16.0 2023-05-16 12:34:40,218 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=204122.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 12:34:40,392 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.4729, 4.2187, 3.8613, 4.4262, 3.5888, 4.0665, 2.6838, 4.4316], device='cuda:0'), covar=tensor([0.1182, 0.0550, 0.1313, 0.0912, 0.0735, 0.0461, 0.1457, 0.1015], device='cuda:0'), in_proj_covar=tensor([0.0222, 0.0262, 0.0292, 0.0353, 0.0234, 0.0235, 0.0257, 0.0372], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 12:34:41,041 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=204123.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 12:34:47,557 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.5824, 2.4444, 3.6139, 4.4972, 3.7961, 4.3864, 3.8473, 3.0935], device='cuda:0'), covar=tensor([0.0032, 0.0406, 0.0151, 0.0037, 0.0113, 0.0095, 0.0102, 0.0362], device='cuda:0'), in_proj_covar=tensor([0.0084, 0.0124, 0.0105, 0.0075, 0.0102, 0.0116, 0.0093, 0.0138], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 12:34:51,826 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=204138.0, num_to_drop=1, layers_to_drop={1} 2023-05-16 12:34:59,892 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=4.43 vs. limit=5.0 2023-05-16 12:35:03,519 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.788e+02 2.845e+02 3.277e+02 3.871e+02 1.275e+03, threshold=6.554e+02, percent-clipped=2.0 2023-05-16 12:35:09,527 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=204163.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 12:35:15,190 INFO [finetune.py:992] (0/2) Epoch 9, batch 6300, loss[loss=0.1649, simple_loss=0.2588, pruned_loss=0.03557, over 10741.00 frames. ], tot_loss[loss=0.1715, simple_loss=0.2596, pruned_loss=0.0417, over 2376066.99 frames. ], batch size: 69, lr: 4.15e-03, grad_scale: 16.0 2023-05-16 12:35:15,270 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=204171.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 12:35:24,487 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.8896, 3.4852, 5.3047, 2.7927, 2.9213, 3.8443, 3.4302, 4.0058], device='cuda:0'), covar=tensor([0.0476, 0.1214, 0.0330, 0.1234, 0.1972, 0.1466, 0.1342, 0.1175], device='cuda:0'), in_proj_covar=tensor([0.0233, 0.0233, 0.0241, 0.0181, 0.0235, 0.0286, 0.0221, 0.0257], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-16 12:35:38,016 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=204203.0, num_to_drop=1, layers_to_drop={3} 2023-05-16 12:35:51,311 INFO [finetune.py:992] (0/2) Epoch 9, batch 6350, loss[loss=0.2014, simple_loss=0.2884, pruned_loss=0.05725, over 12103.00 frames. ], tot_loss[loss=0.1717, simple_loss=0.2596, pruned_loss=0.0419, over 2371788.25 frames. ], batch size: 42, lr: 4.15e-03, grad_scale: 16.0 2023-05-16 12:35:53,681 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=204224.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 12:36:15,886 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.960e+02 2.766e+02 3.267e+02 4.031e+02 1.123e+03, threshold=6.534e+02, percent-clipped=2.0 2023-05-16 12:36:16,720 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=204256.0, num_to_drop=1, layers_to_drop={2} 2023-05-16 12:36:22,400 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.6537, 2.9832, 3.6270, 4.6057, 3.7687, 4.5288, 3.8645, 3.4521], device='cuda:0'), covar=tensor([0.0033, 0.0300, 0.0149, 0.0032, 0.0143, 0.0058, 0.0106, 0.0295], device='cuda:0'), in_proj_covar=tensor([0.0084, 0.0123, 0.0105, 0.0075, 0.0102, 0.0115, 0.0092, 0.0137], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 12:36:27,212 INFO [finetune.py:992] (0/2) Epoch 9, batch 6400, loss[loss=0.2025, simple_loss=0.2952, pruned_loss=0.05491, over 10263.00 frames. ], tot_loss[loss=0.1713, simple_loss=0.2594, pruned_loss=0.04163, over 2380666.49 frames. ], batch size: 68, lr: 4.15e-03, grad_scale: 16.0 2023-05-16 12:37:02,474 INFO [finetune.py:992] (0/2) Epoch 9, batch 6450, loss[loss=0.1626, simple_loss=0.2496, pruned_loss=0.03782, over 12174.00 frames. ], tot_loss[loss=0.1698, simple_loss=0.2577, pruned_loss=0.04093, over 2384980.73 frames. ], batch size: 31, lr: 4.15e-03, grad_scale: 16.0 2023-05-16 12:37:26,650 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.758e+02 2.828e+02 3.342e+02 4.052e+02 6.928e+02, threshold=6.683e+02, percent-clipped=1.0 2023-05-16 12:37:33,441 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.40 vs. limit=2.0 2023-05-16 12:37:39,322 INFO [finetune.py:992] (0/2) Epoch 9, batch 6500, loss[loss=0.1875, simple_loss=0.2817, pruned_loss=0.04668, over 11592.00 frames. ], tot_loss[loss=0.1711, simple_loss=0.2593, pruned_loss=0.0415, over 2370711.64 frames. ], batch size: 48, lr: 4.15e-03, grad_scale: 16.0 2023-05-16 12:37:40,800 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.1794, 4.8784, 5.1562, 5.1124, 4.3667, 4.4429, 4.5746, 4.8882], device='cuda:0'), covar=tensor([0.0875, 0.1060, 0.0848, 0.0795, 0.3066, 0.2178, 0.0658, 0.1553], device='cuda:0'), in_proj_covar=tensor([0.0526, 0.0695, 0.0605, 0.0615, 0.0848, 0.0738, 0.0546, 0.0481], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0005, 0.0004, 0.0004, 0.0005, 0.0005, 0.0004, 0.0003], device='cuda:0') 2023-05-16 12:37:55,078 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.0890, 5.0093, 4.8735, 4.8759, 4.5631, 5.0078, 5.0790, 5.1561], device='cuda:0'), covar=tensor([0.0181, 0.0136, 0.0179, 0.0305, 0.0707, 0.0342, 0.0126, 0.0200], device='cuda:0'), in_proj_covar=tensor([0.0192, 0.0192, 0.0185, 0.0241, 0.0238, 0.0212, 0.0169, 0.0219], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0005, 0.0004, 0.0004, 0.0003, 0.0005], device='cuda:0') 2023-05-16 12:37:57,274 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=2.39 vs. limit=5.0 2023-05-16 12:38:02,538 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=204403.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 12:38:15,128 INFO [finetune.py:992] (0/2) Epoch 9, batch 6550, loss[loss=0.1599, simple_loss=0.2523, pruned_loss=0.03378, over 12030.00 frames. ], tot_loss[loss=0.1716, simple_loss=0.2596, pruned_loss=0.04176, over 2376563.87 frames. ], batch size: 31, lr: 4.15e-03, grad_scale: 16.0 2023-05-16 12:38:15,992 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=204422.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 12:38:18,895 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.2112, 4.0998, 4.0339, 4.4752, 2.9224, 4.0473, 2.5720, 4.2275], device='cuda:0'), covar=tensor([0.1568, 0.0712, 0.1051, 0.0623, 0.1204, 0.0568, 0.1886, 0.1202], device='cuda:0'), in_proj_covar=tensor([0.0225, 0.0265, 0.0296, 0.0358, 0.0238, 0.0239, 0.0261, 0.0379], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 12:38:19,535 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.8447, 4.4229, 4.4530, 4.5392, 4.4021, 4.7187, 4.6009, 2.5412], device='cuda:0'), covar=tensor([0.0136, 0.0093, 0.0134, 0.0105, 0.0078, 0.0117, 0.0100, 0.0925], device='cuda:0'), in_proj_covar=tensor([0.0065, 0.0076, 0.0079, 0.0071, 0.0058, 0.0088, 0.0078, 0.0095], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-16 12:38:21,260 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.95 vs. limit=2.0 2023-05-16 12:38:23,773 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=204433.0, num_to_drop=1, layers_to_drop={1} 2023-05-16 12:38:27,434 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.6050, 4.2901, 4.4122, 4.4521, 4.2136, 4.5074, 4.4665, 2.5980], device='cuda:0'), covar=tensor([0.0115, 0.0078, 0.0098, 0.0073, 0.0061, 0.0093, 0.0076, 0.0817], device='cuda:0'), in_proj_covar=tensor([0.0066, 0.0076, 0.0079, 0.0071, 0.0058, 0.0088, 0.0078, 0.0095], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-16 12:38:36,446 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=204451.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 12:38:39,190 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.736e+02 2.755e+02 3.300e+02 4.124e+02 6.976e+02, threshold=6.601e+02, percent-clipped=1.0 2023-05-16 12:38:49,886 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=204470.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 12:38:50,584 INFO [finetune.py:992] (0/2) Epoch 9, batch 6600, loss[loss=0.1511, simple_loss=0.2321, pruned_loss=0.03502, over 11778.00 frames. ], tot_loss[loss=0.1716, simple_loss=0.2598, pruned_loss=0.04172, over 2375101.29 frames. ], batch size: 26, lr: 4.15e-03, grad_scale: 16.0 2023-05-16 12:38:50,879 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.6735, 2.8775, 4.4910, 4.6504, 2.9085, 2.6441, 2.9756, 2.1099], device='cuda:0'), covar=tensor([0.1433, 0.2786, 0.0440, 0.0388, 0.1221, 0.2155, 0.2490, 0.3819], device='cuda:0'), in_proj_covar=tensor([0.0285, 0.0381, 0.0269, 0.0294, 0.0264, 0.0294, 0.0367, 0.0365], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-16 12:38:55,189 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.6873, 2.9836, 3.7027, 4.6610, 3.8710, 4.5549, 3.9471, 3.1136], device='cuda:0'), covar=tensor([0.0031, 0.0321, 0.0147, 0.0032, 0.0114, 0.0076, 0.0097, 0.0390], device='cuda:0'), in_proj_covar=tensor([0.0084, 0.0123, 0.0105, 0.0075, 0.0102, 0.0115, 0.0092, 0.0137], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 12:39:01,013 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.2086, 4.0848, 3.9782, 4.3906, 2.8704, 3.9526, 2.6617, 4.1219], device='cuda:0'), covar=tensor([0.1580, 0.0684, 0.0942, 0.0676, 0.1145, 0.0585, 0.1781, 0.1106], device='cuda:0'), in_proj_covar=tensor([0.0225, 0.0264, 0.0295, 0.0356, 0.0237, 0.0238, 0.0260, 0.0378], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 12:39:14,363 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=204503.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 12:39:26,382 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=204519.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 12:39:27,724 INFO [finetune.py:992] (0/2) Epoch 9, batch 6650, loss[loss=0.1678, simple_loss=0.265, pruned_loss=0.03535, over 12354.00 frames. ], tot_loss[loss=0.1705, simple_loss=0.2591, pruned_loss=0.041, over 2378654.88 frames. ], batch size: 35, lr: 4.15e-03, grad_scale: 16.0 2023-05-16 12:39:41,122 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.0309, 4.6754, 4.7786, 4.7915, 4.6479, 4.9539, 4.8065, 2.6287], device='cuda:0'), covar=tensor([0.0129, 0.0075, 0.0097, 0.0075, 0.0062, 0.0093, 0.0086, 0.0829], device='cuda:0'), in_proj_covar=tensor([0.0066, 0.0077, 0.0080, 0.0072, 0.0059, 0.0089, 0.0079, 0.0096], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 12:39:48,924 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=204551.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 12:39:51,779 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.759e+02 2.747e+02 3.214e+02 3.903e+02 6.374e+02, threshold=6.428e+02, percent-clipped=0.0 2023-05-16 12:39:52,621 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=204556.0, num_to_drop=1, layers_to_drop={2} 2023-05-16 12:40:00,439 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.3883, 2.2517, 3.1364, 4.3823, 2.2911, 4.3302, 4.4040, 4.5235], device='cuda:0'), covar=tensor([0.0134, 0.1364, 0.0494, 0.0130, 0.1260, 0.0237, 0.0138, 0.0095], device='cuda:0'), in_proj_covar=tensor([0.0120, 0.0203, 0.0183, 0.0116, 0.0188, 0.0177, 0.0174, 0.0118], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 12:40:03,100 INFO [finetune.py:992] (0/2) Epoch 9, batch 6700, loss[loss=0.2169, simple_loss=0.2968, pruned_loss=0.06846, over 12045.00 frames. ], tot_loss[loss=0.1703, simple_loss=0.2588, pruned_loss=0.04091, over 2384753.95 frames. ], batch size: 42, lr: 4.15e-03, grad_scale: 16.0 2023-05-16 12:40:26,681 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=204604.0, num_to_drop=1, layers_to_drop={1} 2023-05-16 12:40:28,949 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.4605, 2.5304, 4.2981, 4.6493, 3.2761, 2.5283, 2.8669, 1.9087], device='cuda:0'), covar=tensor([0.1745, 0.3545, 0.0511, 0.0354, 0.1005, 0.2470, 0.2857, 0.4885], device='cuda:0'), in_proj_covar=tensor([0.0285, 0.0380, 0.0270, 0.0293, 0.0263, 0.0294, 0.0367, 0.0365], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-16 12:40:38,361 INFO [finetune.py:992] (0/2) Epoch 9, batch 6750, loss[loss=0.1605, simple_loss=0.2422, pruned_loss=0.03941, over 12095.00 frames. ], tot_loss[loss=0.1699, simple_loss=0.2584, pruned_loss=0.04069, over 2387116.60 frames. ], batch size: 32, lr: 4.15e-03, grad_scale: 16.0 2023-05-16 12:40:49,197 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.2562, 2.0322, 3.0873, 4.2611, 1.9573, 4.3148, 4.2716, 4.3602], device='cuda:0'), covar=tensor([0.0142, 0.1406, 0.0480, 0.0136, 0.1469, 0.0190, 0.0141, 0.0096], device='cuda:0'), in_proj_covar=tensor([0.0120, 0.0202, 0.0183, 0.0116, 0.0187, 0.0176, 0.0173, 0.0118], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 12:41:03,407 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.015e+02 2.632e+02 3.211e+02 3.961e+02 7.853e+02, threshold=6.423e+02, percent-clipped=1.0 2023-05-16 12:41:15,433 INFO [finetune.py:992] (0/2) Epoch 9, batch 6800, loss[loss=0.1483, simple_loss=0.2363, pruned_loss=0.03011, over 12343.00 frames. ], tot_loss[loss=0.1698, simple_loss=0.2583, pruned_loss=0.04061, over 2386898.66 frames. ], batch size: 31, lr: 4.15e-03, grad_scale: 16.0 2023-05-16 12:41:27,646 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.7275, 2.8979, 3.6885, 4.6691, 3.9539, 4.6443, 4.0864, 3.2652], device='cuda:0'), covar=tensor([0.0028, 0.0328, 0.0154, 0.0031, 0.0127, 0.0067, 0.0083, 0.0316], device='cuda:0'), in_proj_covar=tensor([0.0085, 0.0124, 0.0106, 0.0076, 0.0103, 0.0116, 0.0093, 0.0138], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 12:41:50,815 INFO [finetune.py:992] (0/2) Epoch 9, batch 6850, loss[loss=0.1727, simple_loss=0.2627, pruned_loss=0.04132, over 11655.00 frames. ], tot_loss[loss=0.1708, simple_loss=0.2594, pruned_loss=0.04106, over 2373386.07 frames. ], batch size: 48, lr: 4.15e-03, grad_scale: 16.0 2023-05-16 12:41:59,662 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=204733.0, num_to_drop=1, layers_to_drop={1} 2023-05-16 12:42:15,222 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.976e+02 2.697e+02 3.191e+02 3.818e+02 1.361e+03, threshold=6.382e+02, percent-clipped=3.0 2023-05-16 12:42:26,682 INFO [finetune.py:992] (0/2) Epoch 9, batch 6900, loss[loss=0.1657, simple_loss=0.2582, pruned_loss=0.03661, over 12310.00 frames. ], tot_loss[loss=0.1713, simple_loss=0.2599, pruned_loss=0.04129, over 2371825.90 frames. ], batch size: 34, lr: 4.15e-03, grad_scale: 16.0 2023-05-16 12:42:34,009 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=204781.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 12:42:50,141 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.7138, 2.0928, 2.9348, 3.7331, 2.1845, 3.8097, 3.6796, 3.8359], device='cuda:0'), covar=tensor([0.0161, 0.1261, 0.0489, 0.0166, 0.1168, 0.0272, 0.0233, 0.0124], device='cuda:0'), in_proj_covar=tensor([0.0120, 0.0203, 0.0183, 0.0116, 0.0188, 0.0177, 0.0173, 0.0119], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 12:43:02,373 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=204819.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 12:43:03,631 INFO [finetune.py:992] (0/2) Epoch 9, batch 6950, loss[loss=0.1636, simple_loss=0.25, pruned_loss=0.03863, over 12243.00 frames. ], tot_loss[loss=0.1709, simple_loss=0.2595, pruned_loss=0.04118, over 2368400.13 frames. ], batch size: 32, lr: 4.15e-03, grad_scale: 16.0 2023-05-16 12:43:03,837 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=204821.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 12:43:27,342 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.2836, 6.1230, 5.6778, 5.5945, 6.1648, 5.4591, 5.7133, 5.6975], device='cuda:0'), covar=tensor([0.1448, 0.0908, 0.1087, 0.2116, 0.0982, 0.2261, 0.1643, 0.1245], device='cuda:0'), in_proj_covar=tensor([0.0359, 0.0484, 0.0383, 0.0434, 0.0460, 0.0433, 0.0390, 0.0378], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 12:43:27,951 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.024e+02 2.567e+02 3.121e+02 3.938e+02 7.292e+02, threshold=6.242e+02, percent-clipped=1.0 2023-05-16 12:43:36,584 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=204867.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 12:43:39,312 INFO [finetune.py:992] (0/2) Epoch 9, batch 7000, loss[loss=0.1518, simple_loss=0.2432, pruned_loss=0.03019, over 12092.00 frames. ], tot_loss[loss=0.1706, simple_loss=0.259, pruned_loss=0.04112, over 2372498.84 frames. ], batch size: 32, lr: 4.15e-03, grad_scale: 16.0 2023-05-16 12:43:47,419 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=204882.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 12:44:06,573 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.3442, 4.9445, 5.1162, 5.1283, 4.7854, 5.1947, 5.0621, 3.1481], device='cuda:0'), covar=tensor([0.0091, 0.0068, 0.0077, 0.0061, 0.0057, 0.0075, 0.0083, 0.0539], device='cuda:0'), in_proj_covar=tensor([0.0066, 0.0076, 0.0080, 0.0072, 0.0059, 0.0089, 0.0079, 0.0095], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 12:44:15,048 INFO [finetune.py:992] (0/2) Epoch 9, batch 7050, loss[loss=0.1507, simple_loss=0.2354, pruned_loss=0.03301, over 12091.00 frames. ], tot_loss[loss=0.1701, simple_loss=0.2586, pruned_loss=0.04081, over 2375819.62 frames. ], batch size: 32, lr: 4.15e-03, grad_scale: 16.0 2023-05-16 12:44:39,899 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.000e+02 2.664e+02 3.244e+02 3.963e+02 8.511e+02, threshold=6.488e+02, percent-clipped=4.0 2023-05-16 12:44:42,394 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2023-05-16 12:44:44,722 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=4.86 vs. limit=5.0 2023-05-16 12:44:52,090 INFO [finetune.py:992] (0/2) Epoch 9, batch 7100, loss[loss=0.1675, simple_loss=0.2565, pruned_loss=0.03927, over 12149.00 frames. ], tot_loss[loss=0.1702, simple_loss=0.2587, pruned_loss=0.04082, over 2378367.14 frames. ], batch size: 34, lr: 4.15e-03, grad_scale: 16.0 2023-05-16 12:45:27,683 INFO [finetune.py:992] (0/2) Epoch 9, batch 7150, loss[loss=0.186, simple_loss=0.2809, pruned_loss=0.04557, over 12037.00 frames. ], tot_loss[loss=0.1713, simple_loss=0.2599, pruned_loss=0.04139, over 2371651.75 frames. ], batch size: 42, lr: 4.15e-03, grad_scale: 16.0 2023-05-16 12:45:51,582 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.953e+02 2.688e+02 3.284e+02 3.844e+02 9.605e+02, threshold=6.568e+02, percent-clipped=3.0 2023-05-16 12:46:01,133 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.7190, 2.6743, 4.4121, 4.5207, 2.6729, 2.5866, 2.9535, 2.1607], device='cuda:0'), covar=tensor([0.1452, 0.2958, 0.0481, 0.0445, 0.1380, 0.2224, 0.2665, 0.3869], device='cuda:0'), in_proj_covar=tensor([0.0284, 0.0379, 0.0269, 0.0293, 0.0263, 0.0294, 0.0366, 0.0364], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-16 12:46:02,884 INFO [finetune.py:992] (0/2) Epoch 9, batch 7200, loss[loss=0.2306, simple_loss=0.3134, pruned_loss=0.07392, over 10383.00 frames. ], tot_loss[loss=0.1722, simple_loss=0.2606, pruned_loss=0.04186, over 2368653.28 frames. ], batch size: 68, lr: 4.14e-03, grad_scale: 16.0 2023-05-16 12:46:31,289 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=205109.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 12:46:39,388 INFO [finetune.py:992] (0/2) Epoch 9, batch 7250, loss[loss=0.174, simple_loss=0.2596, pruned_loss=0.04415, over 12246.00 frames. ], tot_loss[loss=0.1712, simple_loss=0.2598, pruned_loss=0.04135, over 2370750.94 frames. ], batch size: 32, lr: 4.14e-03, grad_scale: 16.0 2023-05-16 12:46:45,888 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.3414, 4.6036, 2.9321, 2.6928, 3.9488, 2.5599, 3.9912, 3.1488], device='cuda:0'), covar=tensor([0.0589, 0.0459, 0.1029, 0.1366, 0.0264, 0.1253, 0.0402, 0.0775], device='cuda:0'), in_proj_covar=tensor([0.0188, 0.0256, 0.0177, 0.0199, 0.0140, 0.0181, 0.0197, 0.0175], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 12:46:50,208 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.0419, 3.6323, 5.4175, 2.8043, 3.0683, 3.9102, 3.5763, 4.0175], device='cuda:0'), covar=tensor([0.0398, 0.1035, 0.0266, 0.1177, 0.1814, 0.1544, 0.1224, 0.1076], device='cuda:0'), in_proj_covar=tensor([0.0233, 0.0232, 0.0240, 0.0181, 0.0233, 0.0285, 0.0221, 0.0257], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-16 12:47:02,990 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.2858, 2.8155, 3.8948, 3.1741, 3.6817, 3.3318, 2.7499, 3.7117], device='cuda:0'), covar=tensor([0.0121, 0.0295, 0.0121, 0.0226, 0.0135, 0.0160, 0.0302, 0.0116], device='cuda:0'), in_proj_covar=tensor([0.0174, 0.0201, 0.0183, 0.0182, 0.0214, 0.0159, 0.0194, 0.0179], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 12:47:03,424 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.884e+02 2.817e+02 3.407e+02 3.876e+02 9.349e+02, threshold=6.813e+02, percent-clipped=5.0 2023-05-16 12:47:10,726 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.2972, 4.8247, 5.3257, 4.5829, 4.8748, 4.6558, 5.3332, 5.0426], device='cuda:0'), covar=tensor([0.0302, 0.0364, 0.0247, 0.0256, 0.0403, 0.0297, 0.0243, 0.0239], device='cuda:0'), in_proj_covar=tensor([0.0253, 0.0262, 0.0284, 0.0255, 0.0255, 0.0258, 0.0234, 0.0213], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 12:47:10,793 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.3100, 4.6991, 2.9450, 2.7187, 3.9841, 2.6280, 3.9873, 3.1276], device='cuda:0'), covar=tensor([0.0725, 0.0413, 0.1204, 0.1505, 0.0294, 0.1341, 0.0483, 0.0842], device='cuda:0'), in_proj_covar=tensor([0.0190, 0.0259, 0.0179, 0.0201, 0.0142, 0.0183, 0.0200, 0.0176], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 12:47:14,314 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=205170.0, num_to_drop=1, layers_to_drop={0} 2023-05-16 12:47:14,819 INFO [finetune.py:992] (0/2) Epoch 9, batch 7300, loss[loss=0.149, simple_loss=0.2294, pruned_loss=0.03432, over 12352.00 frames. ], tot_loss[loss=0.1714, simple_loss=0.26, pruned_loss=0.04142, over 2369704.35 frames. ], batch size: 30, lr: 4.14e-03, grad_scale: 16.0 2023-05-16 12:47:19,183 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=205177.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 12:47:34,156 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=205198.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 12:47:50,521 INFO [finetune.py:992] (0/2) Epoch 9, batch 7350, loss[loss=0.1487, simple_loss=0.2327, pruned_loss=0.03237, over 12117.00 frames. ], tot_loss[loss=0.1714, simple_loss=0.2604, pruned_loss=0.04124, over 2378276.87 frames. ], batch size: 30, lr: 4.14e-03, grad_scale: 16.0 2023-05-16 12:48:16,020 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.904e+02 2.746e+02 3.298e+02 4.043e+02 7.027e+02, threshold=6.595e+02, percent-clipped=1.0 2023-05-16 12:48:19,148 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=205259.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 12:48:27,508 INFO [finetune.py:992] (0/2) Epoch 9, batch 7400, loss[loss=0.231, simple_loss=0.3081, pruned_loss=0.07693, over 8419.00 frames. ], tot_loss[loss=0.1725, simple_loss=0.2614, pruned_loss=0.04174, over 2378099.34 frames. ], batch size: 98, lr: 4.14e-03, grad_scale: 16.0 2023-05-16 12:48:41,144 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.6323, 4.1604, 3.7879, 4.3448, 3.5205, 3.9380, 2.6368, 4.3704], device='cuda:0'), covar=tensor([0.1169, 0.0623, 0.1184, 0.0838, 0.0788, 0.0511, 0.1561, 0.1126], device='cuda:0'), in_proj_covar=tensor([0.0223, 0.0263, 0.0293, 0.0355, 0.0235, 0.0237, 0.0258, 0.0378], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 12:49:02,820 INFO [finetune.py:992] (0/2) Epoch 9, batch 7450, loss[loss=0.1895, simple_loss=0.2748, pruned_loss=0.05206, over 12121.00 frames. ], tot_loss[loss=0.1725, simple_loss=0.2614, pruned_loss=0.04185, over 2364806.83 frames. ], batch size: 39, lr: 4.14e-03, grad_scale: 16.0 2023-05-16 12:49:09,564 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.0857, 3.4198, 3.5079, 4.0123, 2.8982, 3.4245, 2.6877, 3.5756], device='cuda:0'), covar=tensor([0.1706, 0.0900, 0.0830, 0.0677, 0.0973, 0.0694, 0.1533, 0.0908], device='cuda:0'), in_proj_covar=tensor([0.0223, 0.0263, 0.0292, 0.0353, 0.0235, 0.0236, 0.0257, 0.0376], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 12:49:26,870 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.913e+02 2.873e+02 3.463e+02 4.047e+02 7.614e+02, threshold=6.926e+02, percent-clipped=2.0 2023-05-16 12:49:38,171 INFO [finetune.py:992] (0/2) Epoch 9, batch 7500, loss[loss=0.1769, simple_loss=0.2734, pruned_loss=0.04019, over 12363.00 frames. ], tot_loss[loss=0.1722, simple_loss=0.261, pruned_loss=0.0417, over 2361222.91 frames. ], batch size: 35, lr: 4.14e-03, grad_scale: 16.0 2023-05-16 12:50:05,745 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.4674, 5.0889, 5.4921, 4.7552, 5.0756, 4.8892, 5.4876, 5.0883], device='cuda:0'), covar=tensor([0.0253, 0.0338, 0.0256, 0.0259, 0.0364, 0.0341, 0.0217, 0.0271], device='cuda:0'), in_proj_covar=tensor([0.0251, 0.0259, 0.0282, 0.0254, 0.0254, 0.0256, 0.0233, 0.0212], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 12:50:15,124 INFO [finetune.py:992] (0/2) Epoch 9, batch 7550, loss[loss=0.1699, simple_loss=0.2647, pruned_loss=0.03753, over 12138.00 frames. ], tot_loss[loss=0.1724, simple_loss=0.261, pruned_loss=0.04187, over 2368117.42 frames. ], batch size: 36, lr: 4.14e-03, grad_scale: 16.0 2023-05-16 12:50:39,122 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.857e+02 2.733e+02 3.190e+02 3.709e+02 8.114e+02, threshold=6.380e+02, percent-clipped=1.0 2023-05-16 12:50:46,047 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=205465.0, num_to_drop=1, layers_to_drop={1} 2023-05-16 12:50:50,139 INFO [finetune.py:992] (0/2) Epoch 9, batch 7600, loss[loss=0.1591, simple_loss=0.2516, pruned_loss=0.03332, over 12368.00 frames. ], tot_loss[loss=0.1717, simple_loss=0.2599, pruned_loss=0.04177, over 2368650.94 frames. ], batch size: 36, lr: 4.14e-03, grad_scale: 16.0 2023-05-16 12:50:54,524 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=205477.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 12:50:55,688 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.43 vs. limit=2.0 2023-05-16 12:51:26,107 INFO [finetune.py:992] (0/2) Epoch 9, batch 7650, loss[loss=0.1605, simple_loss=0.2541, pruned_loss=0.03345, over 12138.00 frames. ], tot_loss[loss=0.172, simple_loss=0.2607, pruned_loss=0.04163, over 2368396.42 frames. ], batch size: 38, lr: 4.14e-03, grad_scale: 16.0 2023-05-16 12:51:28,884 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=205525.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 12:51:47,823 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.46 vs. limit=2.0 2023-05-16 12:51:50,151 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=205554.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 12:51:50,747 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.107e+02 2.771e+02 3.229e+02 3.973e+02 7.489e+02, threshold=6.459e+02, percent-clipped=3.0 2023-05-16 12:52:02,280 INFO [finetune.py:992] (0/2) Epoch 9, batch 7700, loss[loss=0.2011, simple_loss=0.2885, pruned_loss=0.05679, over 12133.00 frames. ], tot_loss[loss=0.172, simple_loss=0.2606, pruned_loss=0.04167, over 2367654.83 frames. ], batch size: 38, lr: 4.14e-03, grad_scale: 16.0 2023-05-16 12:52:27,008 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.2844, 4.1585, 2.7279, 2.4775, 3.6412, 2.4917, 3.7373, 3.0343], device='cuda:0'), covar=tensor([0.0672, 0.0616, 0.1205, 0.1517, 0.0338, 0.1357, 0.0480, 0.0811], device='cuda:0'), in_proj_covar=tensor([0.0189, 0.0259, 0.0177, 0.0200, 0.0141, 0.0182, 0.0199, 0.0176], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 12:52:38,366 INFO [finetune.py:992] (0/2) Epoch 9, batch 7750, loss[loss=0.1991, simple_loss=0.2851, pruned_loss=0.0566, over 12081.00 frames. ], tot_loss[loss=0.1717, simple_loss=0.26, pruned_loss=0.04176, over 2365333.23 frames. ], batch size: 42, lr: 4.14e-03, grad_scale: 16.0 2023-05-16 12:52:38,636 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.8285, 3.2216, 5.1184, 2.6106, 2.6353, 3.7939, 3.3542, 3.8569], device='cuda:0'), covar=tensor([0.0342, 0.1100, 0.0285, 0.1084, 0.1931, 0.1349, 0.1186, 0.1021], device='cuda:0'), in_proj_covar=tensor([0.0233, 0.0232, 0.0241, 0.0180, 0.0233, 0.0287, 0.0221, 0.0258], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-16 12:52:39,979 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.4863, 4.6722, 3.0249, 2.7776, 4.0232, 2.6123, 4.0287, 3.2545], device='cuda:0'), covar=tensor([0.0646, 0.0522, 0.1115, 0.1426, 0.0319, 0.1352, 0.0464, 0.0818], device='cuda:0'), in_proj_covar=tensor([0.0189, 0.0259, 0.0177, 0.0201, 0.0141, 0.0182, 0.0199, 0.0176], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 12:52:51,518 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-05-16 12:53:02,428 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.743e+02 2.764e+02 3.338e+02 3.986e+02 7.043e+02, threshold=6.676e+02, percent-clipped=1.0 2023-05-16 12:53:02,615 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=205655.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 12:53:14,377 INFO [finetune.py:992] (0/2) Epoch 9, batch 7800, loss[loss=0.1413, simple_loss=0.2302, pruned_loss=0.02616, over 12006.00 frames. ], tot_loss[loss=0.1725, simple_loss=0.2608, pruned_loss=0.04213, over 2368628.61 frames. ], batch size: 28, lr: 4.14e-03, grad_scale: 16.0 2023-05-16 12:53:39,839 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.4850, 4.7531, 3.0664, 2.6993, 4.1105, 2.7677, 4.0959, 3.3060], device='cuda:0'), covar=tensor([0.0627, 0.0450, 0.1080, 0.1488, 0.0225, 0.1244, 0.0442, 0.0822], device='cuda:0'), in_proj_covar=tensor([0.0190, 0.0259, 0.0177, 0.0201, 0.0141, 0.0182, 0.0200, 0.0176], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 12:53:46,732 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=205716.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 12:53:50,033 INFO [finetune.py:992] (0/2) Epoch 9, batch 7850, loss[loss=0.1802, simple_loss=0.2654, pruned_loss=0.04749, over 12151.00 frames. ], tot_loss[loss=0.1714, simple_loss=0.2597, pruned_loss=0.04158, over 2378967.89 frames. ], batch size: 36, lr: 4.14e-03, grad_scale: 16.0 2023-05-16 12:54:06,961 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=205744.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 12:54:12,721 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.2376, 4.5382, 4.0566, 4.7942, 4.4638, 2.9221, 4.2529, 3.0918], device='cuda:0'), covar=tensor([0.0776, 0.0790, 0.1355, 0.0497, 0.0989, 0.1553, 0.1025, 0.3239], device='cuda:0'), in_proj_covar=tensor([0.0298, 0.0379, 0.0353, 0.0292, 0.0362, 0.0264, 0.0342, 0.0362], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 12:54:14,491 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.937e+02 2.844e+02 3.312e+02 4.065e+02 1.026e+03, threshold=6.624e+02, percent-clipped=4.0 2023-05-16 12:54:21,863 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=205765.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 12:54:26,058 INFO [finetune.py:992] (0/2) Epoch 9, batch 7900, loss[loss=0.1678, simple_loss=0.2619, pruned_loss=0.03682, over 12315.00 frames. ], tot_loss[loss=0.1707, simple_loss=0.2594, pruned_loss=0.04102, over 2383751.39 frames. ], batch size: 34, lr: 4.14e-03, grad_scale: 16.0 2023-05-16 12:54:38,016 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.5450, 5.3520, 5.4634, 5.5193, 5.1395, 5.1374, 4.9733, 5.4646], device='cuda:0'), covar=tensor([0.0664, 0.0563, 0.0689, 0.0517, 0.1765, 0.1302, 0.0462, 0.0937], device='cuda:0'), in_proj_covar=tensor([0.0523, 0.0689, 0.0598, 0.0612, 0.0837, 0.0729, 0.0537, 0.0477], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0005, 0.0004, 0.0004, 0.0005, 0.0005, 0.0003, 0.0003], device='cuda:0') 2023-05-16 12:54:51,129 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=205805.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 12:54:56,708 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=205813.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 12:55:02,870 INFO [finetune.py:992] (0/2) Epoch 9, batch 7950, loss[loss=0.1535, simple_loss=0.2307, pruned_loss=0.0381, over 12179.00 frames. ], tot_loss[loss=0.1709, simple_loss=0.2595, pruned_loss=0.04119, over 2377021.24 frames. ], batch size: 29, lr: 4.14e-03, grad_scale: 16.0 2023-05-16 12:55:10,090 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.3472, 4.4249, 4.2423, 4.6454, 3.4449, 3.9745, 2.6306, 4.3579], device='cuda:0'), covar=tensor([0.1555, 0.0536, 0.0860, 0.0584, 0.0931, 0.0562, 0.1714, 0.1129], device='cuda:0'), in_proj_covar=tensor([0.0224, 0.0264, 0.0296, 0.0356, 0.0236, 0.0239, 0.0259, 0.0378], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 12:55:26,399 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=205854.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 12:55:26,947 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.901e+02 2.807e+02 3.233e+02 3.735e+02 6.641e+02, threshold=6.465e+02, percent-clipped=1.0 2023-05-16 12:55:33,663 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.42 vs. limit=2.0 2023-05-16 12:55:38,083 INFO [finetune.py:992] (0/2) Epoch 9, batch 8000, loss[loss=0.1693, simple_loss=0.2587, pruned_loss=0.03998, over 12108.00 frames. ], tot_loss[loss=0.1711, simple_loss=0.2593, pruned_loss=0.04141, over 2377277.67 frames. ], batch size: 33, lr: 4.14e-03, grad_scale: 16.0 2023-05-16 12:55:51,910 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=205890.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 12:55:57,119 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=2.59 vs. limit=5.0 2023-05-16 12:56:00,345 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=205902.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 12:56:13,912 INFO [finetune.py:992] (0/2) Epoch 9, batch 8050, loss[loss=0.175, simple_loss=0.2644, pruned_loss=0.04285, over 12153.00 frames. ], tot_loss[loss=0.1705, simple_loss=0.2589, pruned_loss=0.04108, over 2376312.32 frames. ], batch size: 36, lr: 4.14e-03, grad_scale: 8.0 2023-05-16 12:56:27,834 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([6.0743, 6.0797, 5.8648, 5.3216, 5.2221, 6.0571, 5.5967, 5.3572], device='cuda:0'), covar=tensor([0.0682, 0.0803, 0.0621, 0.1651, 0.0559, 0.0617, 0.1473, 0.0970], device='cuda:0'), in_proj_covar=tensor([0.0606, 0.0536, 0.0499, 0.0619, 0.0402, 0.0697, 0.0750, 0.0552], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0004, 0.0004, 0.0003], device='cuda:0') 2023-05-16 12:56:35,788 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=205951.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 12:56:40,446 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.919e+02 2.735e+02 3.249e+02 3.846e+02 1.067e+03, threshold=6.497e+02, percent-clipped=3.0 2023-05-16 12:56:51,077 INFO [finetune.py:992] (0/2) Epoch 9, batch 8100, loss[loss=0.1483, simple_loss=0.2428, pruned_loss=0.02693, over 12030.00 frames. ], tot_loss[loss=0.1715, simple_loss=0.2593, pruned_loss=0.04185, over 2370660.20 frames. ], batch size: 31, lr: 4.14e-03, grad_scale: 8.0 2023-05-16 12:57:11,648 INFO [checkpoint.py:75] (0/2) Saving checkpoint to pruned_transducer_stateless7/exp_giga_finetune/checkpoint-106000.pt 2023-05-16 12:57:22,296 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=206011.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 12:57:29,337 INFO [finetune.py:992] (0/2) Epoch 9, batch 8150, loss[loss=0.1917, simple_loss=0.2601, pruned_loss=0.06164, over 11605.00 frames. ], tot_loss[loss=0.1709, simple_loss=0.2586, pruned_loss=0.04164, over 2371981.45 frames. ], batch size: 48, lr: 4.14e-03, grad_scale: 8.0 2023-05-16 12:57:54,550 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.870e+02 2.743e+02 3.254e+02 3.827e+02 7.721e+02, threshold=6.508e+02, percent-clipped=1.0 2023-05-16 12:57:54,742 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=206056.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 12:58:05,123 INFO [finetune.py:992] (0/2) Epoch 9, batch 8200, loss[loss=0.2565, simple_loss=0.3188, pruned_loss=0.09706, over 7875.00 frames. ], tot_loss[loss=0.172, simple_loss=0.2599, pruned_loss=0.04209, over 2364473.40 frames. ], batch size: 98, lr: 4.13e-03, grad_scale: 8.0 2023-05-16 12:58:22,369 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.3661, 2.5972, 3.5829, 4.3312, 3.7591, 4.3157, 3.7998, 3.1349], device='cuda:0'), covar=tensor([0.0033, 0.0359, 0.0140, 0.0042, 0.0108, 0.0066, 0.0098, 0.0315], device='cuda:0'), in_proj_covar=tensor([0.0084, 0.0124, 0.0106, 0.0076, 0.0102, 0.0115, 0.0093, 0.0137], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 12:58:27,336 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=206100.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 12:58:28,174 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.5339, 4.8187, 3.1829, 2.6174, 4.2145, 2.9631, 4.1469, 3.3640], device='cuda:0'), covar=tensor([0.0704, 0.0550, 0.1083, 0.1651, 0.0277, 0.1209, 0.0444, 0.0864], device='cuda:0'), in_proj_covar=tensor([0.0190, 0.0260, 0.0179, 0.0201, 0.0142, 0.0182, 0.0199, 0.0176], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 12:58:39,556 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=206117.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 12:58:42,272 INFO [finetune.py:992] (0/2) Epoch 9, batch 8250, loss[loss=0.1571, simple_loss=0.2471, pruned_loss=0.03353, over 12105.00 frames. ], tot_loss[loss=0.1712, simple_loss=0.2593, pruned_loss=0.04149, over 2376249.46 frames. ], batch size: 33, lr: 4.13e-03, grad_scale: 8.0 2023-05-16 12:58:46,005 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.7186, 5.6359, 5.6214, 4.9032, 5.0484, 5.8166, 4.8585, 5.1081], device='cuda:0'), covar=tensor([0.1308, 0.1568, 0.1130, 0.2969, 0.1149, 0.1299, 0.3764, 0.1970], device='cuda:0'), in_proj_covar=tensor([0.0607, 0.0536, 0.0500, 0.0619, 0.0401, 0.0698, 0.0751, 0.0552], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0004, 0.0004, 0.0003], device='cuda:0') 2023-05-16 12:59:07,281 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.952e+02 2.848e+02 3.348e+02 4.019e+02 6.771e+02, threshold=6.696e+02, percent-clipped=1.0 2023-05-16 12:59:17,911 INFO [finetune.py:992] (0/2) Epoch 9, batch 8300, loss[loss=0.2022, simple_loss=0.2924, pruned_loss=0.05596, over 11597.00 frames. ], tot_loss[loss=0.1718, simple_loss=0.2599, pruned_loss=0.04182, over 2374164.98 frames. ], batch size: 48, lr: 4.13e-03, grad_scale: 8.0 2023-05-16 12:59:31,109 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.5818, 4.2368, 4.5570, 4.0030, 4.3041, 4.0658, 4.5907, 4.1946], device='cuda:0'), covar=tensor([0.0341, 0.0465, 0.0347, 0.0305, 0.0379, 0.0385, 0.0249, 0.0610], device='cuda:0'), in_proj_covar=tensor([0.0248, 0.0258, 0.0280, 0.0250, 0.0250, 0.0252, 0.0229, 0.0208], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 12:59:36,833 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=206197.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 12:59:41,745 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.35 vs. limit=2.0 2023-05-16 12:59:53,886 INFO [finetune.py:992] (0/2) Epoch 9, batch 8350, loss[loss=0.1587, simple_loss=0.2485, pruned_loss=0.03448, over 12034.00 frames. ], tot_loss[loss=0.1707, simple_loss=0.2592, pruned_loss=0.04111, over 2382407.64 frames. ], batch size: 37, lr: 4.13e-03, grad_scale: 8.0 2023-05-16 13:00:13,308 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=206246.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 13:00:16,573 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.38 vs. limit=2.0 2023-05-16 13:00:20,441 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.777e+02 2.699e+02 3.145e+02 4.092e+02 1.329e+03, threshold=6.291e+02, percent-clipped=4.0 2023-05-16 13:00:22,067 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=206258.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 13:00:22,721 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([6.0994, 6.0721, 5.8473, 5.2333, 5.2309, 6.0291, 5.5680, 5.3627], device='cuda:0'), covar=tensor([0.0685, 0.0874, 0.0603, 0.1692, 0.0661, 0.0748, 0.1809, 0.1023], device='cuda:0'), in_proj_covar=tensor([0.0612, 0.0540, 0.0505, 0.0625, 0.0406, 0.0704, 0.0759, 0.0558], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0004, 0.0004, 0.0003], device='cuda:0') 2023-05-16 13:00:30,043 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=3.26 vs. limit=5.0 2023-05-16 13:00:30,941 INFO [finetune.py:992] (0/2) Epoch 9, batch 8400, loss[loss=0.186, simple_loss=0.2776, pruned_loss=0.04724, over 10578.00 frames. ], tot_loss[loss=0.1713, simple_loss=0.2599, pruned_loss=0.04139, over 2378105.88 frames. ], batch size: 68, lr: 4.13e-03, grad_scale: 8.0 2023-05-16 13:00:31,752 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.0701, 4.9886, 4.8707, 4.9855, 4.1524, 5.0266, 4.9940, 5.2280], device='cuda:0'), covar=tensor([0.0380, 0.0164, 0.0216, 0.0335, 0.1310, 0.0530, 0.0205, 0.0229], device='cuda:0'), in_proj_covar=tensor([0.0195, 0.0195, 0.0188, 0.0243, 0.0241, 0.0215, 0.0171, 0.0221], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0005, 0.0004, 0.0005, 0.0003, 0.0005], device='cuda:0') 2023-05-16 13:00:49,776 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.1629, 3.4676, 3.5968, 4.0005, 2.7872, 3.3508, 2.4868, 3.4499], device='cuda:0'), covar=tensor([0.1571, 0.0868, 0.0923, 0.0703, 0.1070, 0.0737, 0.1712, 0.1080], device='cuda:0'), in_proj_covar=tensor([0.0226, 0.0267, 0.0298, 0.0359, 0.0238, 0.0241, 0.0261, 0.0380], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 13:00:51,390 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.39 vs. limit=2.0 2023-05-16 13:00:59,778 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=206311.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 13:01:01,910 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.7791, 5.4526, 5.0672, 5.0783, 5.5247, 4.9196, 4.9962, 5.0807], device='cuda:0'), covar=tensor([0.1265, 0.0852, 0.1013, 0.1667, 0.0930, 0.1992, 0.1771, 0.1120], device='cuda:0'), in_proj_covar=tensor([0.0358, 0.0484, 0.0381, 0.0433, 0.0463, 0.0434, 0.0391, 0.0378], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 13:01:06,765 INFO [finetune.py:992] (0/2) Epoch 9, batch 8450, loss[loss=0.1818, simple_loss=0.2697, pruned_loss=0.04692, over 12119.00 frames. ], tot_loss[loss=0.1715, simple_loss=0.2598, pruned_loss=0.04164, over 2376138.43 frames. ], batch size: 38, lr: 4.13e-03, grad_scale: 8.0 2023-05-16 13:01:29,311 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.37 vs. limit=2.0 2023-05-16 13:01:31,346 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.934e+02 2.753e+02 3.191e+02 4.256e+02 8.777e+02, threshold=6.383e+02, percent-clipped=2.0 2023-05-16 13:01:33,614 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=206359.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 13:01:40,305 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.9696, 3.0445, 5.3773, 2.5397, 2.4369, 4.1218, 3.3349, 4.0831], device='cuda:0'), covar=tensor([0.0354, 0.1284, 0.0204, 0.1188, 0.2019, 0.1141, 0.1281, 0.0864], device='cuda:0'), in_proj_covar=tensor([0.0231, 0.0228, 0.0240, 0.0178, 0.0231, 0.0285, 0.0219, 0.0256], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:0') 2023-05-16 13:01:42,110 INFO [finetune.py:992] (0/2) Epoch 9, batch 8500, loss[loss=0.2575, simple_loss=0.3275, pruned_loss=0.09371, over 8627.00 frames. ], tot_loss[loss=0.1715, simple_loss=0.2596, pruned_loss=0.04166, over 2367798.23 frames. ], batch size: 98, lr: 4.13e-03, grad_scale: 8.0 2023-05-16 13:02:04,478 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=206400.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 13:02:10,459 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=3.26 vs. limit=5.0 2023-05-16 13:02:13,036 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=206412.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 13:02:19,388 INFO [finetune.py:992] (0/2) Epoch 9, batch 8550, loss[loss=0.1686, simple_loss=0.2609, pruned_loss=0.03819, over 12362.00 frames. ], tot_loss[loss=0.1713, simple_loss=0.2593, pruned_loss=0.04167, over 2363038.69 frames. ], batch size: 35, lr: 4.13e-03, grad_scale: 8.0 2023-05-16 13:02:38,229 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=206448.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 13:02:43,928 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.908e+02 2.816e+02 3.212e+02 3.831e+02 6.403e+02, threshold=6.425e+02, percent-clipped=1.0 2023-05-16 13:02:54,504 INFO [finetune.py:992] (0/2) Epoch 9, batch 8600, loss[loss=0.1599, simple_loss=0.2506, pruned_loss=0.03462, over 11160.00 frames. ], tot_loss[loss=0.1709, simple_loss=0.2587, pruned_loss=0.04157, over 2366631.85 frames. ], batch size: 55, lr: 4.13e-03, grad_scale: 8.0 2023-05-16 13:03:00,645 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=4.62 vs. limit=5.0 2023-05-16 13:03:30,180 INFO [finetune.py:992] (0/2) Epoch 9, batch 8650, loss[loss=0.1981, simple_loss=0.2888, pruned_loss=0.05377, over 12071.00 frames. ], tot_loss[loss=0.1709, simple_loss=0.259, pruned_loss=0.04141, over 2374509.16 frames. ], batch size: 42, lr: 4.13e-03, grad_scale: 8.0 2023-05-16 13:03:39,695 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.9638, 3.0468, 4.8194, 5.1477, 3.1530, 2.8913, 3.2224, 2.3156], device='cuda:0'), covar=tensor([0.1365, 0.2728, 0.0412, 0.0297, 0.1097, 0.2006, 0.2455, 0.3838], device='cuda:0'), in_proj_covar=tensor([0.0285, 0.0380, 0.0270, 0.0292, 0.0262, 0.0295, 0.0367, 0.0365], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-16 13:03:49,471 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=206546.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 13:03:54,464 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=206553.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 13:03:56,478 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.891e+02 2.768e+02 3.154e+02 3.662e+02 7.102e+02, threshold=6.307e+02, percent-clipped=1.0 2023-05-16 13:03:58,143 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.4187, 4.6349, 2.9041, 2.5497, 3.9104, 2.5182, 3.9652, 3.1843], device='cuda:0'), covar=tensor([0.0649, 0.0523, 0.1132, 0.1549, 0.0316, 0.1332, 0.0499, 0.0791], device='cuda:0'), in_proj_covar=tensor([0.0189, 0.0258, 0.0178, 0.0200, 0.0142, 0.0181, 0.0198, 0.0175], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 13:04:06,253 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.86 vs. limit=2.0 2023-05-16 13:04:07,007 INFO [finetune.py:992] (0/2) Epoch 9, batch 8700, loss[loss=0.1764, simple_loss=0.2609, pruned_loss=0.04591, over 11570.00 frames. ], tot_loss[loss=0.1705, simple_loss=0.2589, pruned_loss=0.04109, over 2383184.26 frames. ], batch size: 48, lr: 4.13e-03, grad_scale: 8.0 2023-05-16 13:04:23,457 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=206594.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 13:04:43,004 INFO [finetune.py:992] (0/2) Epoch 9, batch 8750, loss[loss=0.1474, simple_loss=0.238, pruned_loss=0.02844, over 12117.00 frames. ], tot_loss[loss=0.1703, simple_loss=0.2588, pruned_loss=0.0409, over 2385916.07 frames. ], batch size: 33, lr: 4.13e-03, grad_scale: 8.0 2023-05-16 13:04:52,126 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.3698, 4.5281, 2.8380, 2.5642, 3.8700, 2.4887, 3.9301, 3.0509], device='cuda:0'), covar=tensor([0.0649, 0.0556, 0.1150, 0.1535, 0.0288, 0.1333, 0.0529, 0.0844], device='cuda:0'), in_proj_covar=tensor([0.0190, 0.0259, 0.0178, 0.0201, 0.0142, 0.0182, 0.0199, 0.0176], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 13:05:07,595 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.768e+02 2.751e+02 3.304e+02 4.035e+02 8.012e+02, threshold=6.608e+02, percent-clipped=2.0 2023-05-16 13:05:14,518 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.96 vs. limit=2.0 2023-05-16 13:05:18,982 INFO [finetune.py:992] (0/2) Epoch 9, batch 8800, loss[loss=0.1661, simple_loss=0.2541, pruned_loss=0.03909, over 12097.00 frames. ], tot_loss[loss=0.1698, simple_loss=0.2584, pruned_loss=0.04057, over 2390282.45 frames. ], batch size: 33, lr: 4.13e-03, grad_scale: 8.0 2023-05-16 13:05:21,565 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 2023-05-16 13:05:32,582 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.4916, 5.0329, 5.5039, 4.7947, 5.1229, 4.9054, 5.5100, 5.1794], device='cuda:0'), covar=tensor([0.0230, 0.0313, 0.0200, 0.0225, 0.0341, 0.0288, 0.0170, 0.0212], device='cuda:0'), in_proj_covar=tensor([0.0251, 0.0261, 0.0281, 0.0253, 0.0254, 0.0255, 0.0233, 0.0212], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 13:05:39,894 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.3218, 4.8560, 5.3186, 4.6784, 4.9609, 4.7644, 5.3240, 5.0441], device='cuda:0'), covar=tensor([0.0250, 0.0359, 0.0231, 0.0231, 0.0353, 0.0287, 0.0201, 0.0256], device='cuda:0'), in_proj_covar=tensor([0.0251, 0.0260, 0.0281, 0.0252, 0.0254, 0.0255, 0.0232, 0.0211], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 13:05:43,954 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.2383, 6.0567, 5.6158, 5.5478, 6.1076, 5.4620, 5.5017, 5.6956], device='cuda:0'), covar=tensor([0.1467, 0.0903, 0.1047, 0.2064, 0.1060, 0.2054, 0.2121, 0.1131], device='cuda:0'), in_proj_covar=tensor([0.0356, 0.0482, 0.0382, 0.0430, 0.0458, 0.0431, 0.0390, 0.0375], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 13:05:48,994 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=206712.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 13:05:55,252 INFO [finetune.py:992] (0/2) Epoch 9, batch 8850, loss[loss=0.1565, simple_loss=0.2469, pruned_loss=0.03302, over 12291.00 frames. ], tot_loss[loss=0.1712, simple_loss=0.2596, pruned_loss=0.04133, over 2373313.03 frames. ], batch size: 34, lr: 4.13e-03, grad_scale: 8.0 2023-05-16 13:06:04,754 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.3401, 5.1663, 5.2832, 5.2910, 4.8927, 5.0059, 4.7923, 5.2461], device='cuda:0'), covar=tensor([0.0687, 0.0609, 0.0767, 0.0666, 0.2035, 0.1279, 0.0602, 0.0925], device='cuda:0'), in_proj_covar=tensor([0.0528, 0.0695, 0.0597, 0.0610, 0.0839, 0.0728, 0.0542, 0.0476], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0005, 0.0004, 0.0004, 0.0005, 0.0005, 0.0004, 0.0003], device='cuda:0') 2023-05-16 13:06:15,711 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.31 vs. limit=2.0 2023-05-16 13:06:20,285 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.931e+02 2.742e+02 3.107e+02 3.864e+02 8.527e+02, threshold=6.215e+02, percent-clipped=1.0 2023-05-16 13:06:23,218 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=206760.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 13:06:30,904 INFO [finetune.py:992] (0/2) Epoch 9, batch 8900, loss[loss=0.1773, simple_loss=0.2714, pruned_loss=0.04158, over 12316.00 frames. ], tot_loss[loss=0.1707, simple_loss=0.259, pruned_loss=0.04126, over 2377445.69 frames. ], batch size: 34, lr: 4.13e-03, grad_scale: 8.0 2023-05-16 13:06:56,198 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.6132, 2.4842, 3.3114, 4.6275, 2.0125, 4.6196, 4.5879, 4.7042], device='cuda:0'), covar=tensor([0.0113, 0.1258, 0.0405, 0.0119, 0.1416, 0.0134, 0.0123, 0.0080], device='cuda:0'), in_proj_covar=tensor([0.0120, 0.0202, 0.0185, 0.0116, 0.0189, 0.0176, 0.0173, 0.0118], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 13:07:03,075 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.7509, 3.7435, 3.2639, 3.2113, 2.9574, 2.8221, 3.6843, 2.5469], device='cuda:0'), covar=tensor([0.0305, 0.0095, 0.0170, 0.0185, 0.0344, 0.0348, 0.0132, 0.0398], device='cuda:0'), in_proj_covar=tensor([0.0189, 0.0161, 0.0157, 0.0186, 0.0201, 0.0199, 0.0167, 0.0198], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 13:07:07,237 INFO [finetune.py:992] (0/2) Epoch 9, batch 8950, loss[loss=0.2348, simple_loss=0.3126, pruned_loss=0.07853, over 7983.00 frames. ], tot_loss[loss=0.171, simple_loss=0.2591, pruned_loss=0.0414, over 2372731.24 frames. ], batch size: 98, lr: 4.13e-03, grad_scale: 8.0 2023-05-16 13:07:07,752 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.44 vs. limit=2.0 2023-05-16 13:07:17,261 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=206834.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 13:07:30,457 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=206853.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 13:07:32,401 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.908e+02 2.757e+02 3.302e+02 3.769e+02 8.409e+02, threshold=6.604e+02, percent-clipped=4.0 2023-05-16 13:07:43,051 INFO [finetune.py:992] (0/2) Epoch 9, batch 9000, loss[loss=0.1846, simple_loss=0.278, pruned_loss=0.04559, over 12335.00 frames. ], tot_loss[loss=0.1701, simple_loss=0.2583, pruned_loss=0.0409, over 2375863.91 frames. ], batch size: 36, lr: 4.13e-03, grad_scale: 8.0 2023-05-16 13:07:43,052 INFO [finetune.py:1017] (0/2) Computing validation loss 2023-05-16 13:07:50,810 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.5135, 5.4601, 5.3858, 4.8299, 5.1096, 5.4582, 5.0102, 5.0436], device='cuda:0'), covar=tensor([0.0715, 0.0978, 0.0626, 0.1818, 0.0544, 0.0839, 0.1594, 0.0958], device='cuda:0'), in_proj_covar=tensor([0.0604, 0.0539, 0.0501, 0.0622, 0.0409, 0.0699, 0.0757, 0.0558], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0004, 0.0004, 0.0003], device='cuda:0') 2023-05-16 13:08:01,993 INFO [finetune.py:1026] (0/2) Epoch 9, validation: loss=0.3325, simple_loss=0.4032, pruned_loss=0.1308, over 1020973.00 frames. 2023-05-16 13:08:01,994 INFO [finetune.py:1027] (0/2) Maximum memory allocated so far is 12761MB 2023-05-16 13:08:18,884 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=206895.0, num_to_drop=1, layers_to_drop={0} 2023-05-16 13:08:19,623 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.0660, 4.9101, 4.8517, 4.8156, 4.5535, 4.9543, 4.9487, 5.1991], device='cuda:0'), covar=tensor([0.0177, 0.0149, 0.0195, 0.0358, 0.0705, 0.0293, 0.0144, 0.0168], device='cuda:0'), in_proj_covar=tensor([0.0196, 0.0197, 0.0189, 0.0244, 0.0242, 0.0215, 0.0173, 0.0222], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0005, 0.0004, 0.0005, 0.0003, 0.0005], device='cuda:0') 2023-05-16 13:08:23,059 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=206901.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 13:08:37,871 INFO [finetune.py:992] (0/2) Epoch 9, batch 9050, loss[loss=0.1699, simple_loss=0.2632, pruned_loss=0.03832, over 12074.00 frames. ], tot_loss[loss=0.17, simple_loss=0.2582, pruned_loss=0.04088, over 2382153.72 frames. ], batch size: 40, lr: 4.13e-03, grad_scale: 8.0 2023-05-16 13:09:03,322 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.727e+02 2.978e+02 3.533e+02 4.335e+02 8.865e+02, threshold=7.066e+02, percent-clipped=4.0 2023-05-16 13:09:13,739 INFO [finetune.py:992] (0/2) Epoch 9, batch 9100, loss[loss=0.1649, simple_loss=0.246, pruned_loss=0.04186, over 12294.00 frames. ], tot_loss[loss=0.1702, simple_loss=0.2584, pruned_loss=0.04102, over 2384706.14 frames. ], batch size: 28, lr: 4.13e-03, grad_scale: 8.0 2023-05-16 13:09:48,943 INFO [finetune.py:992] (0/2) Epoch 9, batch 9150, loss[loss=0.1662, simple_loss=0.2667, pruned_loss=0.03286, over 12268.00 frames. ], tot_loss[loss=0.1714, simple_loss=0.2595, pruned_loss=0.04163, over 2378597.69 frames. ], batch size: 37, lr: 4.12e-03, grad_scale: 8.0 2023-05-16 13:10:14,432 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.033e+02 2.676e+02 3.136e+02 3.998e+02 8.990e+02, threshold=6.272e+02, percent-clipped=3.0 2023-05-16 13:10:23,282 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.8291, 3.3574, 5.1400, 2.7827, 2.7142, 3.8371, 3.4303, 3.9060], device='cuda:0'), covar=tensor([0.0422, 0.1188, 0.0362, 0.1181, 0.1986, 0.1419, 0.1245, 0.1161], device='cuda:0'), in_proj_covar=tensor([0.0235, 0.0232, 0.0243, 0.0180, 0.0235, 0.0289, 0.0222, 0.0259], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-16 13:10:25,182 INFO [finetune.py:992] (0/2) Epoch 9, batch 9200, loss[loss=0.1735, simple_loss=0.2697, pruned_loss=0.03865, over 12121.00 frames. ], tot_loss[loss=0.1719, simple_loss=0.26, pruned_loss=0.04188, over 2371863.97 frames. ], batch size: 33, lr: 4.12e-03, grad_scale: 8.0 2023-05-16 13:11:01,769 INFO [finetune.py:992] (0/2) Epoch 9, batch 9250, loss[loss=0.179, simple_loss=0.265, pruned_loss=0.04646, over 12353.00 frames. ], tot_loss[loss=0.171, simple_loss=0.2595, pruned_loss=0.04128, over 2379592.22 frames. ], batch size: 31, lr: 4.12e-03, grad_scale: 8.0 2023-05-16 13:11:26,267 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.903e+02 2.738e+02 3.250e+02 4.161e+02 6.723e+02, threshold=6.501e+02, percent-clipped=1.0 2023-05-16 13:11:32,186 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.6526, 2.8646, 4.3222, 4.5168, 2.8328, 2.6279, 2.9010, 2.0546], device='cuda:0'), covar=tensor([0.1444, 0.2640, 0.0505, 0.0414, 0.1230, 0.2245, 0.2441, 0.3862], device='cuda:0'), in_proj_covar=tensor([0.0285, 0.0380, 0.0271, 0.0294, 0.0263, 0.0294, 0.0368, 0.0365], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-16 13:11:35,711 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.0759, 4.7041, 4.7797, 4.9298, 4.8696, 4.9892, 4.9024, 2.5819], device='cuda:0'), covar=tensor([0.0081, 0.0069, 0.0093, 0.0064, 0.0039, 0.0108, 0.0056, 0.0767], device='cuda:0'), in_proj_covar=tensor([0.0065, 0.0075, 0.0078, 0.0070, 0.0058, 0.0088, 0.0077, 0.0093], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-16 13:11:36,824 INFO [finetune.py:992] (0/2) Epoch 9, batch 9300, loss[loss=0.2872, simple_loss=0.343, pruned_loss=0.1157, over 7736.00 frames. ], tot_loss[loss=0.1715, simple_loss=0.2601, pruned_loss=0.04149, over 2376408.66 frames. ], batch size: 98, lr: 4.12e-03, grad_scale: 8.0 2023-05-16 13:11:40,585 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=207176.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 13:11:50,478 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=207190.0, num_to_drop=1, layers_to_drop={0} 2023-05-16 13:12:02,129 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.0422, 4.9152, 4.8166, 4.8130, 4.5328, 5.0029, 5.0108, 5.2005], device='cuda:0'), covar=tensor([0.0189, 0.0136, 0.0176, 0.0303, 0.0722, 0.0312, 0.0131, 0.0164], device='cuda:0'), in_proj_covar=tensor([0.0197, 0.0197, 0.0190, 0.0244, 0.0242, 0.0215, 0.0173, 0.0223], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0005, 0.0004, 0.0005, 0.0004, 0.0005], device='cuda:0') 2023-05-16 13:12:13,445 INFO [finetune.py:992] (0/2) Epoch 9, batch 9350, loss[loss=0.1728, simple_loss=0.2669, pruned_loss=0.03935, over 12365.00 frames. ], tot_loss[loss=0.1713, simple_loss=0.2597, pruned_loss=0.0414, over 2372063.31 frames. ], batch size: 35, lr: 4.12e-03, grad_scale: 8.0 2023-05-16 13:12:14,586 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.70 vs. limit=2.0 2023-05-16 13:12:25,762 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=207237.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 13:12:39,157 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.598e+02 2.762e+02 3.159e+02 4.154e+02 6.329e+02, threshold=6.318e+02, percent-clipped=0.0 2023-05-16 13:12:49,862 INFO [finetune.py:992] (0/2) Epoch 9, batch 9400, loss[loss=0.1743, simple_loss=0.2612, pruned_loss=0.04369, over 12257.00 frames. ], tot_loss[loss=0.1716, simple_loss=0.2602, pruned_loss=0.04151, over 2379990.41 frames. ], batch size: 32, lr: 4.12e-03, grad_scale: 8.0 2023-05-16 13:13:08,287 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.7976, 3.8548, 3.3156, 3.2135, 3.1195, 2.9236, 3.9007, 2.6247], device='cuda:0'), covar=tensor([0.0308, 0.0124, 0.0220, 0.0208, 0.0316, 0.0371, 0.0111, 0.0411], device='cuda:0'), in_proj_covar=tensor([0.0188, 0.0160, 0.0156, 0.0186, 0.0201, 0.0198, 0.0166, 0.0198], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 13:13:25,062 INFO [finetune.py:992] (0/2) Epoch 9, batch 9450, loss[loss=0.1738, simple_loss=0.2689, pruned_loss=0.03934, over 12331.00 frames. ], tot_loss[loss=0.1724, simple_loss=0.261, pruned_loss=0.04194, over 2377263.87 frames. ], batch size: 36, lr: 4.12e-03, grad_scale: 8.0 2023-05-16 13:13:50,121 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.085e+02 2.743e+02 3.330e+02 3.955e+02 7.509e+02, threshold=6.659e+02, percent-clipped=3.0 2023-05-16 13:13:59,680 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.8974, 5.5728, 5.1011, 5.0899, 5.6721, 4.9414, 5.0822, 5.1510], device='cuda:0'), covar=tensor([0.1148, 0.0945, 0.1167, 0.1854, 0.0959, 0.2054, 0.1940, 0.1193], device='cuda:0'), in_proj_covar=tensor([0.0359, 0.0488, 0.0385, 0.0435, 0.0464, 0.0438, 0.0390, 0.0380], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 13:14:00,987 INFO [finetune.py:992] (0/2) Epoch 9, batch 9500, loss[loss=0.2508, simple_loss=0.3257, pruned_loss=0.08802, over 8079.00 frames. ], tot_loss[loss=0.171, simple_loss=0.2593, pruned_loss=0.04137, over 2381979.40 frames. ], batch size: 98, lr: 4.12e-03, grad_scale: 8.0 2023-05-16 13:14:37,860 INFO [finetune.py:992] (0/2) Epoch 9, batch 9550, loss[loss=0.1884, simple_loss=0.284, pruned_loss=0.04642, over 12183.00 frames. ], tot_loss[loss=0.1709, simple_loss=0.2593, pruned_loss=0.04125, over 2375310.63 frames. ], batch size: 35, lr: 4.12e-03, grad_scale: 8.0 2023-05-16 13:14:51,001 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.56 vs. limit=2.0 2023-05-16 13:15:02,847 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.035e+02 2.805e+02 3.276e+02 4.126e+02 1.039e+03, threshold=6.552e+02, percent-clipped=2.0 2023-05-16 13:15:08,933 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=4.85 vs. limit=5.0 2023-05-16 13:15:13,493 INFO [finetune.py:992] (0/2) Epoch 9, batch 9600, loss[loss=0.1586, simple_loss=0.2426, pruned_loss=0.03733, over 12178.00 frames. ], tot_loss[loss=0.1714, simple_loss=0.2597, pruned_loss=0.04157, over 2371253.14 frames. ], batch size: 29, lr: 4.12e-03, grad_scale: 8.0 2023-05-16 13:15:27,394 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=207490.0, num_to_drop=1, layers_to_drop={1} 2023-05-16 13:15:49,495 INFO [finetune.py:992] (0/2) Epoch 9, batch 9650, loss[loss=0.1821, simple_loss=0.2729, pruned_loss=0.04567, over 12348.00 frames. ], tot_loss[loss=0.1724, simple_loss=0.261, pruned_loss=0.04192, over 2375298.77 frames. ], batch size: 35, lr: 4.12e-03, grad_scale: 8.0 2023-05-16 13:15:58,170 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=207532.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 13:16:02,198 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=207538.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 13:16:14,942 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.512e+02 2.667e+02 3.231e+02 3.855e+02 7.021e+02, threshold=6.462e+02, percent-clipped=2.0 2023-05-16 13:16:17,706 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.76 vs. limit=2.0 2023-05-16 13:16:25,737 INFO [finetune.py:992] (0/2) Epoch 9, batch 9700, loss[loss=0.1476, simple_loss=0.2344, pruned_loss=0.03044, over 12110.00 frames. ], tot_loss[loss=0.1716, simple_loss=0.26, pruned_loss=0.04161, over 2372538.60 frames. ], batch size: 33, lr: 4.12e-03, grad_scale: 8.0 2023-05-16 13:17:01,869 INFO [finetune.py:992] (0/2) Epoch 9, batch 9750, loss[loss=0.1712, simple_loss=0.2612, pruned_loss=0.04064, over 12098.00 frames. ], tot_loss[loss=0.1717, simple_loss=0.26, pruned_loss=0.04171, over 2368202.04 frames. ], batch size: 33, lr: 4.12e-03, grad_scale: 8.0 2023-05-16 13:17:26,415 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.908e+02 2.711e+02 3.133e+02 4.067e+02 8.021e+02, threshold=6.267e+02, percent-clipped=3.0 2023-05-16 13:17:37,213 INFO [finetune.py:992] (0/2) Epoch 9, batch 9800, loss[loss=0.1699, simple_loss=0.2581, pruned_loss=0.04087, over 12283.00 frames. ], tot_loss[loss=0.1705, simple_loss=0.259, pruned_loss=0.04106, over 2377736.70 frames. ], batch size: 33, lr: 4.12e-03, grad_scale: 8.0 2023-05-16 13:18:13,362 INFO [finetune.py:992] (0/2) Epoch 9, batch 9850, loss[loss=0.1793, simple_loss=0.2744, pruned_loss=0.04215, over 12175.00 frames. ], tot_loss[loss=0.1707, simple_loss=0.2592, pruned_loss=0.04113, over 2374202.32 frames. ], batch size: 36, lr: 4.12e-03, grad_scale: 8.0 2023-05-16 13:18:38,485 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.043e+02 2.674e+02 3.274e+02 3.799e+02 1.142e+03, threshold=6.547e+02, percent-clipped=1.0 2023-05-16 13:18:49,854 INFO [finetune.py:992] (0/2) Epoch 9, batch 9900, loss[loss=0.1804, simple_loss=0.2735, pruned_loss=0.0436, over 12183.00 frames. ], tot_loss[loss=0.1703, simple_loss=0.259, pruned_loss=0.0408, over 2378436.04 frames. ], batch size: 35, lr: 4.12e-03, grad_scale: 8.0 2023-05-16 13:18:55,168 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.6172, 3.1150, 5.1180, 2.5667, 2.6621, 3.7669, 3.0562, 3.8955], device='cuda:0'), covar=tensor([0.0513, 0.1334, 0.0285, 0.1243, 0.2027, 0.1422, 0.1459, 0.0986], device='cuda:0'), in_proj_covar=tensor([0.0234, 0.0232, 0.0244, 0.0180, 0.0234, 0.0288, 0.0221, 0.0258], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-16 13:19:26,959 INFO [finetune.py:992] (0/2) Epoch 9, batch 9950, loss[loss=0.1807, simple_loss=0.2723, pruned_loss=0.04454, over 12023.00 frames. ], tot_loss[loss=0.17, simple_loss=0.2585, pruned_loss=0.04077, over 2371013.31 frames. ], batch size: 40, lr: 4.12e-03, grad_scale: 8.0 2023-05-16 13:19:35,000 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=207832.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 13:19:48,494 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.5555, 2.1414, 3.8625, 4.5559, 3.9986, 4.3817, 4.1991, 3.2837], device='cuda:0'), covar=tensor([0.0042, 0.0548, 0.0118, 0.0030, 0.0101, 0.0088, 0.0078, 0.0346], device='cuda:0'), in_proj_covar=tensor([0.0087, 0.0126, 0.0107, 0.0078, 0.0105, 0.0118, 0.0095, 0.0140], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 13:19:51,787 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.069e+02 2.891e+02 3.446e+02 3.978e+02 7.477e+02, threshold=6.892e+02, percent-clipped=1.0 2023-05-16 13:20:02,354 INFO [finetune.py:992] (0/2) Epoch 9, batch 10000, loss[loss=0.1819, simple_loss=0.2779, pruned_loss=0.04296, over 10653.00 frames. ], tot_loss[loss=0.1708, simple_loss=0.2592, pruned_loss=0.04119, over 2365985.72 frames. ], batch size: 69, lr: 4.12e-03, grad_scale: 8.0 2023-05-16 13:20:08,831 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=207880.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 13:20:20,485 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.6702, 3.7529, 3.2998, 3.2855, 3.0364, 2.8780, 3.8344, 2.4558], device='cuda:0'), covar=tensor([0.0377, 0.0185, 0.0204, 0.0219, 0.0358, 0.0397, 0.0142, 0.0492], device='cuda:0'), in_proj_covar=tensor([0.0188, 0.0161, 0.0156, 0.0185, 0.0199, 0.0197, 0.0167, 0.0197], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 13:20:38,498 INFO [finetune.py:992] (0/2) Epoch 9, batch 10050, loss[loss=0.144, simple_loss=0.2232, pruned_loss=0.03238, over 12341.00 frames. ], tot_loss[loss=0.1711, simple_loss=0.2593, pruned_loss=0.04148, over 2363633.25 frames. ], batch size: 30, lr: 4.12e-03, grad_scale: 16.0 2023-05-16 13:20:55,771 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.84 vs. limit=2.0 2023-05-16 13:21:03,758 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.037e+02 2.712e+02 2.996e+02 4.038e+02 9.495e+02, threshold=5.992e+02, percent-clipped=2.0 2023-05-16 13:21:11,003 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.0483, 5.0243, 4.9009, 5.1652, 4.0313, 5.1607, 5.1582, 5.3342], device='cuda:0'), covar=tensor([0.0232, 0.0167, 0.0195, 0.0258, 0.1166, 0.0258, 0.0139, 0.0182], device='cuda:0'), in_proj_covar=tensor([0.0193, 0.0193, 0.0187, 0.0239, 0.0237, 0.0212, 0.0171, 0.0221], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0005, 0.0004, 0.0004, 0.0003, 0.0005], device='cuda:0') 2023-05-16 13:21:15,013 INFO [finetune.py:992] (0/2) Epoch 9, batch 10100, loss[loss=0.1653, simple_loss=0.252, pruned_loss=0.03926, over 12094.00 frames. ], tot_loss[loss=0.1712, simple_loss=0.2594, pruned_loss=0.04154, over 2361263.58 frames. ], batch size: 32, lr: 4.12e-03, grad_scale: 16.0 2023-05-16 13:21:17,510 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 2023-05-16 13:21:32,340 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.39 vs. limit=2.0 2023-05-16 13:21:35,761 INFO [checkpoint.py:75] (0/2) Saving checkpoint to pruned_transducer_stateless7/exp_giga_finetune/checkpoint-108000.pt 2023-05-16 13:21:47,803 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=4.67 vs. limit=5.0 2023-05-16 13:21:53,581 INFO [finetune.py:992] (0/2) Epoch 9, batch 10150, loss[loss=0.1919, simple_loss=0.2834, pruned_loss=0.05016, over 11590.00 frames. ], tot_loss[loss=0.1718, simple_loss=0.2597, pruned_loss=0.0419, over 2362108.43 frames. ], batch size: 48, lr: 4.11e-03, grad_scale: 16.0 2023-05-16 13:21:54,576 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.6574, 2.4584, 4.6654, 5.0198, 3.1199, 2.5693, 2.9180, 1.8893], device='cuda:0'), covar=tensor([0.1651, 0.4107, 0.0454, 0.0296, 0.1081, 0.2410, 0.2864, 0.5171], device='cuda:0'), in_proj_covar=tensor([0.0286, 0.0381, 0.0272, 0.0295, 0.0265, 0.0295, 0.0370, 0.0367], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-16 13:22:01,306 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=208032.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 13:22:19,190 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.079e+02 2.808e+02 3.338e+02 3.944e+02 5.870e+02, threshold=6.676e+02, percent-clipped=0.0 2023-05-16 13:22:30,123 INFO [finetune.py:992] (0/2) Epoch 9, batch 10200, loss[loss=0.1493, simple_loss=0.239, pruned_loss=0.02978, over 12085.00 frames. ], tot_loss[loss=0.1714, simple_loss=0.2595, pruned_loss=0.0416, over 2374165.01 frames. ], batch size: 32, lr: 4.11e-03, grad_scale: 16.0 2023-05-16 13:22:46,079 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=208093.0, num_to_drop=1, layers_to_drop={0} 2023-05-16 13:22:49,429 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.5777, 5.3481, 5.4628, 5.5425, 5.0952, 5.1888, 5.0181, 5.4367], device='cuda:0'), covar=tensor([0.0554, 0.0585, 0.0797, 0.0475, 0.1643, 0.1217, 0.0472, 0.0976], device='cuda:0'), in_proj_covar=tensor([0.0530, 0.0700, 0.0605, 0.0614, 0.0839, 0.0737, 0.0549, 0.0476], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0005, 0.0004, 0.0004, 0.0005, 0.0005, 0.0004, 0.0003], device='cuda:0') 2023-05-16 13:23:06,071 INFO [finetune.py:992] (0/2) Epoch 9, batch 10250, loss[loss=0.1442, simple_loss=0.2298, pruned_loss=0.02933, over 11783.00 frames. ], tot_loss[loss=0.1719, simple_loss=0.2603, pruned_loss=0.04175, over 2369932.15 frames. ], batch size: 26, lr: 4.11e-03, grad_scale: 16.0 2023-05-16 13:23:06,265 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.1653, 5.0984, 4.9564, 5.0721, 4.6801, 5.0430, 5.1945, 5.3606], device='cuda:0'), covar=tensor([0.0212, 0.0136, 0.0181, 0.0275, 0.0669, 0.0302, 0.0131, 0.0135], device='cuda:0'), in_proj_covar=tensor([0.0192, 0.0193, 0.0186, 0.0238, 0.0236, 0.0211, 0.0170, 0.0220], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0005, 0.0004, 0.0004, 0.0003, 0.0005], device='cuda:0') 2023-05-16 13:23:13,608 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.93 vs. limit=2.0 2023-05-16 13:23:28,620 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.83 vs. limit=2.0 2023-05-16 13:23:30,900 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.929e+02 2.613e+02 3.110e+02 3.784e+02 7.618e+02, threshold=6.219e+02, percent-clipped=3.0 2023-05-16 13:23:41,550 INFO [finetune.py:992] (0/2) Epoch 9, batch 10300, loss[loss=0.1694, simple_loss=0.2684, pruned_loss=0.03519, over 12180.00 frames. ], tot_loss[loss=0.172, simple_loss=0.2605, pruned_loss=0.04178, over 2372059.28 frames. ], batch size: 35, lr: 4.11e-03, grad_scale: 16.0 2023-05-16 13:24:17,382 INFO [finetune.py:992] (0/2) Epoch 9, batch 10350, loss[loss=0.1682, simple_loss=0.2662, pruned_loss=0.03515, over 12104.00 frames. ], tot_loss[loss=0.172, simple_loss=0.2604, pruned_loss=0.04183, over 2377247.10 frames. ], batch size: 33, lr: 4.11e-03, grad_scale: 16.0 2023-05-16 13:24:21,914 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.1041, 5.8998, 5.3967, 5.4063, 5.9824, 5.3329, 5.5299, 5.5285], device='cuda:0'), covar=tensor([0.1450, 0.0914, 0.1097, 0.1858, 0.0996, 0.2190, 0.1749, 0.1051], device='cuda:0'), in_proj_covar=tensor([0.0353, 0.0481, 0.0379, 0.0430, 0.0457, 0.0435, 0.0384, 0.0374], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 13:24:42,886 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.021e+02 2.816e+02 3.307e+02 4.221e+02 9.934e+02, threshold=6.615e+02, percent-clipped=3.0 2023-05-16 13:24:49,513 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=208265.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 13:24:53,616 INFO [finetune.py:992] (0/2) Epoch 9, batch 10400, loss[loss=0.1532, simple_loss=0.2504, pruned_loss=0.02797, over 12255.00 frames. ], tot_loss[loss=0.1713, simple_loss=0.2598, pruned_loss=0.04137, over 2380370.68 frames. ], batch size: 37, lr: 4.11e-03, grad_scale: 16.0 2023-05-16 13:25:10,320 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-05-16 13:25:29,376 INFO [finetune.py:992] (0/2) Epoch 9, batch 10450, loss[loss=0.1769, simple_loss=0.2725, pruned_loss=0.04068, over 12336.00 frames. ], tot_loss[loss=0.171, simple_loss=0.2596, pruned_loss=0.04123, over 2377659.03 frames. ], batch size: 36, lr: 4.11e-03, grad_scale: 16.0 2023-05-16 13:25:33,090 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=208326.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 13:25:54,848 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.882e+02 2.807e+02 3.294e+02 3.880e+02 7.412e+02, threshold=6.587e+02, percent-clipped=1.0 2023-05-16 13:26:05,561 INFO [finetune.py:992] (0/2) Epoch 9, batch 10500, loss[loss=0.1657, simple_loss=0.2608, pruned_loss=0.0353, over 11838.00 frames. ], tot_loss[loss=0.1711, simple_loss=0.2594, pruned_loss=0.04139, over 2366178.38 frames. ], batch size: 44, lr: 4.11e-03, grad_scale: 16.0 2023-05-16 13:26:09,435 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.4348, 3.9988, 3.9732, 4.3481, 3.0438, 3.8870, 2.5554, 4.0719], device='cuda:0'), covar=tensor([0.1452, 0.0682, 0.0979, 0.0720, 0.1049, 0.0604, 0.1819, 0.1281], device='cuda:0'), in_proj_covar=tensor([0.0220, 0.0261, 0.0290, 0.0348, 0.0231, 0.0236, 0.0254, 0.0371], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 13:26:18,481 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=208388.0, num_to_drop=1, layers_to_drop={2} 2023-05-16 13:26:42,089 INFO [finetune.py:992] (0/2) Epoch 9, batch 10550, loss[loss=0.2359, simple_loss=0.3118, pruned_loss=0.08004, over 8458.00 frames. ], tot_loss[loss=0.1713, simple_loss=0.2596, pruned_loss=0.04146, over 2357837.38 frames. ], batch size: 98, lr: 4.11e-03, grad_scale: 16.0 2023-05-16 13:27:06,869 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.104e+02 2.736e+02 3.181e+02 3.822e+02 7.078e+02, threshold=6.362e+02, percent-clipped=1.0 2023-05-16 13:27:14,269 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.7982, 3.4848, 5.2748, 2.6815, 2.7431, 3.8685, 3.2586, 3.9832], device='cuda:0'), covar=tensor([0.0444, 0.1095, 0.0258, 0.1228, 0.1878, 0.1385, 0.1343, 0.1127], device='cuda:0'), in_proj_covar=tensor([0.0236, 0.0235, 0.0248, 0.0181, 0.0237, 0.0292, 0.0224, 0.0263], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-16 13:27:17,670 INFO [finetune.py:992] (0/2) Epoch 9, batch 10600, loss[loss=0.1852, simple_loss=0.2795, pruned_loss=0.04542, over 12069.00 frames. ], tot_loss[loss=0.1709, simple_loss=0.2593, pruned_loss=0.04124, over 2361813.27 frames. ], batch size: 42, lr: 4.11e-03, grad_scale: 16.0 2023-05-16 13:27:18,062 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 2023-05-16 13:27:25,022 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.9309, 3.6952, 5.2775, 3.0038, 3.0075, 4.0179, 3.4485, 4.0901], device='cuda:0'), covar=tensor([0.0449, 0.1051, 0.0392, 0.1061, 0.1742, 0.1413, 0.1299, 0.1223], device='cuda:0'), in_proj_covar=tensor([0.0235, 0.0234, 0.0248, 0.0181, 0.0237, 0.0292, 0.0223, 0.0262], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-16 13:27:29,553 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=4.40 vs. limit=5.0 2023-05-16 13:27:53,867 INFO [finetune.py:992] (0/2) Epoch 9, batch 10650, loss[loss=0.2024, simple_loss=0.2942, pruned_loss=0.05532, over 12113.00 frames. ], tot_loss[loss=0.1713, simple_loss=0.2596, pruned_loss=0.04146, over 2355413.84 frames. ], batch size: 39, lr: 4.11e-03, grad_scale: 16.0 2023-05-16 13:27:56,081 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.4525, 2.6148, 3.1468, 4.3962, 2.1786, 4.3362, 4.4343, 4.5257], device='cuda:0'), covar=tensor([0.0126, 0.1151, 0.0488, 0.0160, 0.1419, 0.0223, 0.0154, 0.0103], device='cuda:0'), in_proj_covar=tensor([0.0121, 0.0203, 0.0186, 0.0117, 0.0189, 0.0177, 0.0174, 0.0119], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 13:28:07,356 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.2123, 4.8383, 4.9532, 5.0291, 4.8275, 5.0556, 5.0392, 2.9637], device='cuda:0'), covar=tensor([0.0078, 0.0055, 0.0075, 0.0057, 0.0045, 0.0076, 0.0050, 0.0681], device='cuda:0'), in_proj_covar=tensor([0.0066, 0.0076, 0.0079, 0.0072, 0.0059, 0.0089, 0.0078, 0.0095], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-16 13:28:19,138 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.969e+02 2.612e+02 3.191e+02 3.895e+02 7.455e+02, threshold=6.381e+02, percent-clipped=2.0 2023-05-16 13:28:29,720 INFO [finetune.py:992] (0/2) Epoch 9, batch 10700, loss[loss=0.1753, simple_loss=0.2661, pruned_loss=0.04228, over 12297.00 frames. ], tot_loss[loss=0.1716, simple_loss=0.2602, pruned_loss=0.04149, over 2356513.06 frames. ], batch size: 34, lr: 4.11e-03, grad_scale: 16.0 2023-05-16 13:28:33,551 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=208576.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 13:28:44,363 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2023-05-16 13:29:00,831 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.0561, 4.6689, 5.0294, 4.3507, 4.7422, 4.3367, 4.9687, 4.8410], device='cuda:0'), covar=tensor([0.0382, 0.0528, 0.0510, 0.0329, 0.0405, 0.0400, 0.0389, 0.0466], device='cuda:0'), in_proj_covar=tensor([0.0251, 0.0260, 0.0280, 0.0253, 0.0255, 0.0253, 0.0230, 0.0213], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 13:29:05,670 INFO [finetune.py:992] (0/2) Epoch 9, batch 10750, loss[loss=0.2019, simple_loss=0.2868, pruned_loss=0.05846, over 10492.00 frames. ], tot_loss[loss=0.1716, simple_loss=0.2601, pruned_loss=0.04156, over 2361054.97 frames. ], batch size: 68, lr: 4.11e-03, grad_scale: 16.0 2023-05-16 13:29:05,769 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=208621.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 13:29:05,966 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.7402, 2.7373, 4.1348, 4.3679, 2.9498, 2.6781, 2.7956, 2.2184], device='cuda:0'), covar=tensor([0.1430, 0.2793, 0.0545, 0.0435, 0.1118, 0.2109, 0.2682, 0.4028], device='cuda:0'), in_proj_covar=tensor([0.0289, 0.0385, 0.0275, 0.0299, 0.0268, 0.0299, 0.0373, 0.0371], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-16 13:29:17,147 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=208637.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 13:29:23,401 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.6183, 2.4089, 3.2870, 4.5267, 2.2743, 4.5221, 4.5650, 4.6709], device='cuda:0'), covar=tensor([0.0100, 0.1344, 0.0476, 0.0153, 0.1365, 0.0183, 0.0160, 0.0082], device='cuda:0'), in_proj_covar=tensor([0.0120, 0.0202, 0.0185, 0.0116, 0.0188, 0.0176, 0.0173, 0.0119], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 13:29:30,958 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.833e+02 2.664e+02 3.149e+02 3.821e+02 8.730e+02, threshold=6.297e+02, percent-clipped=2.0 2023-05-16 13:29:42,443 INFO [finetune.py:992] (0/2) Epoch 9, batch 10800, loss[loss=0.215, simple_loss=0.293, pruned_loss=0.06852, over 12164.00 frames. ], tot_loss[loss=0.1718, simple_loss=0.2599, pruned_loss=0.04181, over 2357493.11 frames. ], batch size: 36, lr: 4.11e-03, grad_scale: 16.0 2023-05-16 13:29:45,471 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.5159, 4.9930, 4.1795, 5.1642, 4.5653, 2.6059, 4.3348, 3.0519], device='cuda:0'), covar=tensor([0.0668, 0.0650, 0.1372, 0.0431, 0.1034, 0.1828, 0.1043, 0.3036], device='cuda:0'), in_proj_covar=tensor([0.0302, 0.0381, 0.0359, 0.0297, 0.0370, 0.0267, 0.0344, 0.0366], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 13:29:54,441 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=208688.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 13:30:17,704 INFO [finetune.py:992] (0/2) Epoch 9, batch 10850, loss[loss=0.2117, simple_loss=0.29, pruned_loss=0.06674, over 12151.00 frames. ], tot_loss[loss=0.1718, simple_loss=0.26, pruned_loss=0.04177, over 2371236.92 frames. ], batch size: 36, lr: 4.11e-03, grad_scale: 16.0 2023-05-16 13:30:25,735 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.9859, 4.7765, 4.7551, 4.8389, 4.6345, 4.8742, 4.8214, 2.8065], device='cuda:0'), covar=tensor([0.0099, 0.0063, 0.0090, 0.0065, 0.0063, 0.0088, 0.0073, 0.0711], device='cuda:0'), in_proj_covar=tensor([0.0065, 0.0075, 0.0078, 0.0071, 0.0058, 0.0088, 0.0077, 0.0093], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-16 13:30:28,592 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=208736.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 13:30:43,100 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.187e+02 2.779e+02 3.243e+02 4.269e+02 7.545e+02, threshold=6.485e+02, percent-clipped=7.0 2023-05-16 13:30:54,150 INFO [finetune.py:992] (0/2) Epoch 9, batch 10900, loss[loss=0.1636, simple_loss=0.2491, pruned_loss=0.039, over 12196.00 frames. ], tot_loss[loss=0.1715, simple_loss=0.2601, pruned_loss=0.04149, over 2378611.86 frames. ], batch size: 31, lr: 4.11e-03, grad_scale: 16.0 2023-05-16 13:31:09,374 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.33 vs. limit=2.0 2023-05-16 13:31:27,520 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.0273, 4.8946, 4.8539, 4.9091, 4.3865, 4.9702, 5.0431, 5.2217], device='cuda:0'), covar=tensor([0.0190, 0.0153, 0.0195, 0.0290, 0.0895, 0.0293, 0.0138, 0.0157], device='cuda:0'), in_proj_covar=tensor([0.0191, 0.0191, 0.0184, 0.0237, 0.0238, 0.0209, 0.0170, 0.0219], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0005, 0.0004, 0.0004, 0.0003, 0.0005], device='cuda:0') 2023-05-16 13:31:31,673 INFO [finetune.py:992] (0/2) Epoch 9, batch 10950, loss[loss=0.1601, simple_loss=0.2435, pruned_loss=0.03832, over 12250.00 frames. ], tot_loss[loss=0.1721, simple_loss=0.2608, pruned_loss=0.04165, over 2370206.21 frames. ], batch size: 32, lr: 4.11e-03, grad_scale: 16.0 2023-05-16 13:31:55,981 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.549e+02 2.795e+02 3.337e+02 4.140e+02 1.225e+03, threshold=6.674e+02, percent-clipped=5.0 2023-05-16 13:32:06,895 INFO [finetune.py:992] (0/2) Epoch 9, batch 11000, loss[loss=0.2079, simple_loss=0.2935, pruned_loss=0.06109, over 12137.00 frames. ], tot_loss[loss=0.1729, simple_loss=0.2614, pruned_loss=0.04218, over 2363862.36 frames. ], batch size: 34, lr: 4.11e-03, grad_scale: 16.0 2023-05-16 13:32:09,231 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.7508, 2.5461, 4.5786, 4.9843, 3.0959, 2.6994, 2.9359, 1.9808], device='cuda:0'), covar=tensor([0.1516, 0.3618, 0.0459, 0.0333, 0.1044, 0.2227, 0.2755, 0.4685], device='cuda:0'), in_proj_covar=tensor([0.0287, 0.0384, 0.0274, 0.0299, 0.0267, 0.0298, 0.0372, 0.0370], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-16 13:32:38,721 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.2098, 2.9326, 2.8230, 2.8475, 2.5959, 2.5764, 2.5530, 1.9989], device='cuda:0'), covar=tensor([0.0368, 0.0159, 0.0142, 0.0197, 0.0291, 0.0228, 0.0215, 0.0458], device='cuda:0'), in_proj_covar=tensor([0.0191, 0.0163, 0.0159, 0.0188, 0.0202, 0.0200, 0.0171, 0.0201], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 13:32:42,687 INFO [finetune.py:992] (0/2) Epoch 9, batch 11050, loss[loss=0.1997, simple_loss=0.2857, pruned_loss=0.05686, over 12101.00 frames. ], tot_loss[loss=0.1769, simple_loss=0.2654, pruned_loss=0.04417, over 2332204.33 frames. ], batch size: 38, lr: 4.11e-03, grad_scale: 16.0 2023-05-16 13:32:42,898 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=208921.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 13:32:51,042 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=208932.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 13:32:53,385 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.6528, 2.8588, 3.7870, 4.7015, 4.1275, 4.6489, 4.1284, 3.3553], device='cuda:0'), covar=tensor([0.0034, 0.0351, 0.0142, 0.0034, 0.0092, 0.0064, 0.0101, 0.0342], device='cuda:0'), in_proj_covar=tensor([0.0087, 0.0127, 0.0108, 0.0077, 0.0105, 0.0118, 0.0096, 0.0140], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 13:33:08,148 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.753e+02 3.079e+02 3.769e+02 4.599e+02 8.043e+02, threshold=7.539e+02, percent-clipped=5.0 2023-05-16 13:33:17,301 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=208969.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 13:33:18,545 INFO [finetune.py:992] (0/2) Epoch 9, batch 11100, loss[loss=0.2785, simple_loss=0.3549, pruned_loss=0.101, over 10309.00 frames. ], tot_loss[loss=0.181, simple_loss=0.2692, pruned_loss=0.04639, over 2295059.59 frames. ], batch size: 68, lr: 4.10e-03, grad_scale: 16.0 2023-05-16 13:33:55,076 INFO [finetune.py:992] (0/2) Epoch 9, batch 11150, loss[loss=0.3082, simple_loss=0.3596, pruned_loss=0.1283, over 6986.00 frames. ], tot_loss[loss=0.1868, simple_loss=0.2744, pruned_loss=0.0496, over 2243901.18 frames. ], batch size: 98, lr: 4.10e-03, grad_scale: 16.0 2023-05-16 13:34:19,850 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.408e+02 3.523e+02 4.250e+02 5.223e+02 9.989e+02, threshold=8.500e+02, percent-clipped=4.0 2023-05-16 13:34:30,300 INFO [finetune.py:992] (0/2) Epoch 9, batch 11200, loss[loss=0.3181, simple_loss=0.3735, pruned_loss=0.1313, over 7002.00 frames. ], tot_loss[loss=0.1946, simple_loss=0.2812, pruned_loss=0.05402, over 2177624.99 frames. ], batch size: 99, lr: 4.10e-03, grad_scale: 16.0 2023-05-16 13:35:06,101 INFO [finetune.py:992] (0/2) Epoch 9, batch 11250, loss[loss=0.2454, simple_loss=0.3384, pruned_loss=0.07617, over 12033.00 frames. ], tot_loss[loss=0.2037, simple_loss=0.2892, pruned_loss=0.05911, over 2107563.71 frames. ], batch size: 42, lr: 4.10e-03, grad_scale: 16.0 2023-05-16 13:35:12,352 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.7720, 3.0589, 2.4137, 2.1958, 2.5855, 2.2743, 2.9556, 2.5338], device='cuda:0'), covar=tensor([0.0464, 0.0424, 0.0777, 0.1269, 0.0243, 0.0997, 0.0398, 0.0667], device='cuda:0'), in_proj_covar=tensor([0.0187, 0.0251, 0.0175, 0.0195, 0.0138, 0.0177, 0.0193, 0.0173], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 13:35:14,622 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 2023-05-16 13:35:30,520 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.236e+02 3.451e+02 3.991e+02 4.600e+02 9.298e+02, threshold=7.983e+02, percent-clipped=1.0 2023-05-16 13:35:41,569 INFO [finetune.py:992] (0/2) Epoch 9, batch 11300, loss[loss=0.174, simple_loss=0.2651, pruned_loss=0.04146, over 12251.00 frames. ], tot_loss[loss=0.2096, simple_loss=0.2941, pruned_loss=0.06255, over 2048994.74 frames. ], batch size: 32, lr: 4.10e-03, grad_scale: 16.0 2023-05-16 13:36:03,313 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=209202.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 13:36:16,461 INFO [finetune.py:992] (0/2) Epoch 9, batch 11350, loss[loss=0.2897, simple_loss=0.3504, pruned_loss=0.1145, over 6796.00 frames. ], tot_loss[loss=0.2145, simple_loss=0.2985, pruned_loss=0.06529, over 2014921.39 frames. ], batch size: 98, lr: 4.10e-03, grad_scale: 16.0 2023-05-16 13:36:24,054 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=209232.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 13:36:40,354 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=209255.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 13:36:40,828 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.590e+02 3.481e+02 4.181e+02 4.952e+02 7.214e+02, threshold=8.362e+02, percent-clipped=0.0 2023-05-16 13:36:45,592 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=209263.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 13:36:48,840 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.7805, 5.4701, 5.1155, 5.0318, 5.5599, 4.9457, 5.1480, 5.1058], device='cuda:0'), covar=tensor([0.1163, 0.0856, 0.0902, 0.1598, 0.0836, 0.1896, 0.1559, 0.0929], device='cuda:0'), in_proj_covar=tensor([0.0347, 0.0469, 0.0373, 0.0422, 0.0448, 0.0426, 0.0378, 0.0364], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002], device='cuda:0') 2023-05-16 13:36:50,717 INFO [finetune.py:992] (0/2) Epoch 9, batch 11400, loss[loss=0.2349, simple_loss=0.3171, pruned_loss=0.07631, over 11718.00 frames. ], tot_loss[loss=0.2199, simple_loss=0.3029, pruned_loss=0.06845, over 1954768.28 frames. ], batch size: 44, lr: 4.10e-03, grad_scale: 16.0 2023-05-16 13:36:56,879 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=209280.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 13:37:09,661 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=209298.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 13:37:22,662 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=209316.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 13:37:25,614 INFO [finetune.py:992] (0/2) Epoch 9, batch 11450, loss[loss=0.2675, simple_loss=0.336, pruned_loss=0.09955, over 6656.00 frames. ], tot_loss[loss=0.2246, simple_loss=0.3065, pruned_loss=0.07137, over 1906994.35 frames. ], batch size: 99, lr: 4.10e-03, grad_scale: 16.0 2023-05-16 13:37:50,138 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.446e+02 3.472e+02 3.871e+02 4.791e+02 8.062e+02, threshold=7.742e+02, percent-clipped=0.0 2023-05-16 13:37:52,382 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=209359.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 13:38:00,954 INFO [finetune.py:992] (0/2) Epoch 9, batch 11500, loss[loss=0.2298, simple_loss=0.3134, pruned_loss=0.07311, over 10205.00 frames. ], tot_loss[loss=0.228, simple_loss=0.3086, pruned_loss=0.0737, over 1877043.07 frames. ], batch size: 68, lr: 4.10e-03, grad_scale: 16.0 2023-05-16 13:38:02,350 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.7969, 4.4401, 4.0312, 4.1387, 4.4734, 4.0091, 4.1566, 3.9752], device='cuda:0'), covar=tensor([0.1526, 0.0973, 0.1249, 0.1745, 0.0993, 0.2039, 0.1475, 0.1352], device='cuda:0'), in_proj_covar=tensor([0.0344, 0.0462, 0.0369, 0.0417, 0.0442, 0.0421, 0.0372, 0.0361], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002], device='cuda:0') 2023-05-16 13:38:10,480 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.9487, 2.1573, 2.6761, 3.0154, 2.2276, 3.0971, 2.9126, 3.1157], device='cuda:0'), covar=tensor([0.0139, 0.1015, 0.0415, 0.0159, 0.0954, 0.0254, 0.0283, 0.0109], device='cuda:0'), in_proj_covar=tensor([0.0116, 0.0197, 0.0179, 0.0112, 0.0183, 0.0171, 0.0166, 0.0115], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 13:38:35,485 INFO [finetune.py:992] (0/2) Epoch 9, batch 11550, loss[loss=0.2686, simple_loss=0.3311, pruned_loss=0.1031, over 6759.00 frames. ], tot_loss[loss=0.2321, simple_loss=0.3114, pruned_loss=0.0764, over 1842083.76 frames. ], batch size: 99, lr: 4.10e-03, grad_scale: 16.0 2023-05-16 13:38:58,447 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.8120, 2.1812, 2.7637, 2.6658, 2.9440, 2.8760, 2.8754, 2.3264], device='cuda:0'), covar=tensor([0.0083, 0.0372, 0.0173, 0.0078, 0.0112, 0.0093, 0.0126, 0.0390], device='cuda:0'), in_proj_covar=tensor([0.0084, 0.0122, 0.0103, 0.0074, 0.0100, 0.0113, 0.0092, 0.0135], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 13:38:59,550 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.537e+02 3.719e+02 4.229e+02 5.064e+02 1.141e+03, threshold=8.457e+02, percent-clipped=4.0 2023-05-16 13:39:03,158 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.7181, 3.5426, 3.5626, 3.7334, 3.3864, 3.7386, 3.7980, 3.7976], device='cuda:0'), covar=tensor([0.0186, 0.0150, 0.0164, 0.0274, 0.0535, 0.0294, 0.0155, 0.0201], device='cuda:0'), in_proj_covar=tensor([0.0176, 0.0174, 0.0170, 0.0218, 0.0217, 0.0192, 0.0157, 0.0203], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:0') 2023-05-16 13:39:06,516 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=209466.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 13:39:09,715 INFO [finetune.py:992] (0/2) Epoch 9, batch 11600, loss[loss=0.2695, simple_loss=0.3339, pruned_loss=0.1025, over 6964.00 frames. ], tot_loss[loss=0.2343, simple_loss=0.3126, pruned_loss=0.078, over 1823494.96 frames. ], batch size: 99, lr: 4.10e-03, grad_scale: 16.0 2023-05-16 13:39:19,319 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.7252, 3.5606, 3.5701, 3.7703, 3.4293, 3.7568, 3.7851, 3.8379], device='cuda:0'), covar=tensor([0.0197, 0.0153, 0.0186, 0.0276, 0.0531, 0.0323, 0.0159, 0.0208], device='cuda:0'), in_proj_covar=tensor([0.0177, 0.0175, 0.0171, 0.0219, 0.0219, 0.0193, 0.0158, 0.0203], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:0') 2023-05-16 13:39:32,185 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.34 vs. limit=2.0 2023-05-16 13:39:38,507 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=209510.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 13:39:41,582 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.0146, 2.2746, 3.4179, 2.8246, 3.2142, 3.1004, 2.2550, 3.3535], device='cuda:0'), covar=tensor([0.0107, 0.0362, 0.0086, 0.0234, 0.0134, 0.0148, 0.0358, 0.0094], device='cuda:0'), in_proj_covar=tensor([0.0163, 0.0193, 0.0172, 0.0173, 0.0199, 0.0150, 0.0185, 0.0168], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 13:39:46,477 INFO [finetune.py:992] (0/2) Epoch 9, batch 11650, loss[loss=0.1958, simple_loss=0.2828, pruned_loss=0.05439, over 12151.00 frames. ], tot_loss[loss=0.2336, simple_loss=0.3115, pruned_loss=0.07786, over 1803063.91 frames. ], batch size: 36, lr: 4.10e-03, grad_scale: 16.0 2023-05-16 13:39:51,079 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=209527.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 13:40:10,934 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.471e+02 3.255e+02 3.824e+02 4.485e+02 1.562e+03, threshold=7.649e+02, percent-clipped=1.0 2023-05-16 13:40:13,130 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=209558.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 13:40:21,777 INFO [finetune.py:992] (0/2) Epoch 9, batch 11700, loss[loss=0.2612, simple_loss=0.3237, pruned_loss=0.09941, over 6619.00 frames. ], tot_loss[loss=0.2348, simple_loss=0.3117, pruned_loss=0.07896, over 1760025.12 frames. ], batch size: 98, lr: 4.10e-03, grad_scale: 16.0 2023-05-16 13:40:22,005 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=209571.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 13:40:46,218 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=209606.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 13:40:49,319 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=209611.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 13:40:54,317 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.39 vs. limit=2.0 2023-05-16 13:40:56,651 INFO [finetune.py:992] (0/2) Epoch 9, batch 11750, loss[loss=0.2161, simple_loss=0.3068, pruned_loss=0.06274, over 10302.00 frames. ], tot_loss[loss=0.2366, simple_loss=0.3126, pruned_loss=0.08027, over 1736890.77 frames. ], batch size: 68, lr: 4.10e-03, grad_scale: 16.0 2023-05-16 13:41:07,037 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 2023-05-16 13:41:19,059 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=209654.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 13:41:20,969 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.757e+02 3.765e+02 4.358e+02 5.471e+02 1.515e+03, threshold=8.716e+02, percent-clipped=4.0 2023-05-16 13:41:25,937 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.6731, 2.8390, 4.3390, 4.5263, 2.9420, 2.7602, 2.8207, 2.0881], device='cuda:0'), covar=tensor([0.1501, 0.2951, 0.0458, 0.0343, 0.1220, 0.2262, 0.2806, 0.4281], device='cuda:0'), in_proj_covar=tensor([0.0279, 0.0370, 0.0264, 0.0287, 0.0258, 0.0290, 0.0362, 0.0361], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-16 13:41:28,578 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=209667.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 13:41:31,079 INFO [finetune.py:992] (0/2) Epoch 9, batch 11800, loss[loss=0.3305, simple_loss=0.3625, pruned_loss=0.1492, over 6253.00 frames. ], tot_loss[loss=0.2396, simple_loss=0.3149, pruned_loss=0.08219, over 1704554.55 frames. ], batch size: 99, lr: 4.10e-03, grad_scale: 16.0 2023-05-16 13:41:43,554 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=209688.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 13:41:52,559 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=3.54 vs. limit=5.0 2023-05-16 13:42:06,038 INFO [finetune.py:992] (0/2) Epoch 9, batch 11850, loss[loss=0.2053, simple_loss=0.2991, pruned_loss=0.05574, over 11038.00 frames. ], tot_loss[loss=0.242, simple_loss=0.3172, pruned_loss=0.08343, over 1695062.56 frames. ], batch size: 55, lr: 4.10e-03, grad_scale: 16.0 2023-05-16 13:42:25,411 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=209749.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 13:42:29,826 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.633e+02 3.558e+02 4.060e+02 4.711e+02 6.540e+02, threshold=8.119e+02, percent-clipped=0.0 2023-05-16 13:42:39,979 INFO [finetune.py:992] (0/2) Epoch 9, batch 11900, loss[loss=0.2087, simple_loss=0.3053, pruned_loss=0.05604, over 10239.00 frames. ], tot_loss[loss=0.2403, simple_loss=0.3165, pruned_loss=0.08207, over 1687240.38 frames. ], batch size: 68, lr: 4.10e-03, grad_scale: 16.0 2023-05-16 13:43:15,102 INFO [finetune.py:992] (0/2) Epoch 9, batch 11950, loss[loss=0.2296, simple_loss=0.3083, pruned_loss=0.07547, over 7228.00 frames. ], tot_loss[loss=0.2354, simple_loss=0.3128, pruned_loss=0.07895, over 1684221.15 frames. ], batch size: 99, lr: 4.10e-03, grad_scale: 16.0 2023-05-16 13:43:15,947 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=209822.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 13:43:39,982 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.202e+02 3.119e+02 3.617e+02 4.335e+02 9.745e+02, threshold=7.234e+02, percent-clipped=4.0 2023-05-16 13:43:41,522 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=209858.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 13:43:47,126 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=209866.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 13:43:50,516 INFO [finetune.py:992] (0/2) Epoch 9, batch 12000, loss[loss=0.1809, simple_loss=0.2803, pruned_loss=0.04071, over 10112.00 frames. ], tot_loss[loss=0.23, simple_loss=0.3089, pruned_loss=0.0755, over 1691411.63 frames. ], batch size: 68, lr: 4.10e-03, grad_scale: 16.0 2023-05-16 13:43:50,517 INFO [finetune.py:1017] (0/2) Computing validation loss 2023-05-16 13:44:04,986 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.3386, 4.0028, 3.6020, 3.4198, 3.2662, 3.2844, 3.1772, 2.4648], device='cuda:0'), covar=tensor([0.0473, 0.0060, 0.0091, 0.0152, 0.0232, 0.0210, 0.0169, 0.0502], device='cuda:0'), in_proj_covar=tensor([0.0181, 0.0151, 0.0149, 0.0178, 0.0190, 0.0191, 0.0160, 0.0191], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 13:44:08,688 INFO [finetune.py:1026] (0/2) Epoch 9, validation: loss=0.2842, simple_loss=0.3622, pruned_loss=0.103, over 1020973.00 frames. 2023-05-16 13:44:08,689 INFO [finetune.py:1027] (0/2) Maximum memory allocated so far is 12827MB 2023-05-16 13:44:33,570 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=209906.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 13:44:37,038 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=209911.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 13:44:43,493 INFO [finetune.py:992] (0/2) Epoch 9, batch 12050, loss[loss=0.2226, simple_loss=0.3096, pruned_loss=0.06783, over 10085.00 frames. ], tot_loss[loss=0.2246, simple_loss=0.3044, pruned_loss=0.0724, over 1679023.36 frames. ], batch size: 68, lr: 4.10e-03, grad_scale: 16.0 2023-05-16 13:45:05,435 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=209954.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 13:45:07,178 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.224e+02 2.944e+02 3.345e+02 3.905e+02 1.431e+03, threshold=6.691e+02, percent-clipped=3.0 2023-05-16 13:45:08,581 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=209959.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 13:45:10,475 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=209962.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 13:45:16,045 INFO [finetune.py:992] (0/2) Epoch 9, batch 12100, loss[loss=0.2065, simple_loss=0.2898, pruned_loss=0.06164, over 7018.00 frames. ], tot_loss[loss=0.2217, simple_loss=0.3026, pruned_loss=0.07037, over 1689313.30 frames. ], batch size: 97, lr: 4.09e-03, grad_scale: 16.0 2023-05-16 13:45:24,498 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.58 vs. limit=2.0 2023-05-16 13:45:34,585 INFO [checkpoint.py:75] (0/2) Saving checkpoint to pruned_transducer_stateless7/exp_giga_finetune/checkpoint-110000.pt 2023-05-16 13:45:39,073 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=210002.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 13:45:40,094 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.35 vs. limit=2.0 2023-05-16 13:45:45,598 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.2746, 3.4956, 3.1998, 3.5002, 3.3551, 2.5084, 3.2357, 2.7493], device='cuda:0'), covar=tensor([0.0972, 0.1181, 0.1541, 0.0756, 0.1371, 0.1770, 0.1175, 0.3450], device='cuda:0'), in_proj_covar=tensor([0.0281, 0.0351, 0.0326, 0.0267, 0.0339, 0.0251, 0.0316, 0.0339], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-16 13:45:50,084 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([1.9922, 2.2369, 2.1987, 2.1424, 1.9340, 1.9283, 2.1053, 1.5939], device='cuda:0'), covar=tensor([0.0298, 0.0220, 0.0187, 0.0205, 0.0352, 0.0247, 0.0173, 0.0424], device='cuda:0'), in_proj_covar=tensor([0.0180, 0.0150, 0.0147, 0.0176, 0.0189, 0.0189, 0.0158, 0.0190], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], device='cuda:0') 2023-05-16 13:45:51,166 INFO [finetune.py:992] (0/2) Epoch 9, batch 12150, loss[loss=0.2425, simple_loss=0.3113, pruned_loss=0.08685, over 7104.00 frames. ], tot_loss[loss=0.2229, simple_loss=0.3035, pruned_loss=0.07112, over 1686437.36 frames. ], batch size: 102, lr: 4.09e-03, grad_scale: 16.0 2023-05-16 13:46:05,598 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=210044.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 13:46:13,786 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.148e+02 3.271e+02 3.941e+02 4.882e+02 1.262e+03, threshold=7.882e+02, percent-clipped=10.0 2023-05-16 13:46:13,951 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=210057.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 13:46:22,206 INFO [finetune.py:992] (0/2) Epoch 9, batch 12200, loss[loss=0.2267, simple_loss=0.3124, pruned_loss=0.07051, over 6898.00 frames. ], tot_loss[loss=0.2248, simple_loss=0.3049, pruned_loss=0.07233, over 1668262.29 frames. ], batch size: 98, lr: 4.09e-03, grad_scale: 16.0 2023-05-16 13:46:29,298 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2023-05-16 13:46:43,803 INFO [checkpoint.py:75] (0/2) Saving checkpoint to pruned_transducer_stateless7/exp_giga_finetune/epoch-9.pt 2023-05-16 13:47:06,197 INFO [finetune.py:992] (0/2) Epoch 10, batch 0, loss[loss=0.1911, simple_loss=0.2835, pruned_loss=0.04934, over 12353.00 frames. ], tot_loss[loss=0.1911, simple_loss=0.2835, pruned_loss=0.04934, over 12353.00 frames. ], batch size: 35, lr: 4.09e-03, grad_scale: 16.0 2023-05-16 13:47:06,197 INFO [finetune.py:1017] (0/2) Computing validation loss 2023-05-16 13:47:23,506 INFO [finetune.py:1026] (0/2) Epoch 10, validation: loss=0.2866, simple_loss=0.3628, pruned_loss=0.1052, over 1020973.00 frames. 2023-05-16 13:47:23,507 INFO [finetune.py:1027] (0/2) Maximum memory allocated so far is 12827MB 2023-05-16 13:47:32,894 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=210118.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 13:47:35,580 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=210122.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 13:47:40,120 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=4.38 vs. limit=5.0 2023-05-16 13:47:58,746 INFO [finetune.py:992] (0/2) Epoch 10, batch 50, loss[loss=0.2535, simple_loss=0.3245, pruned_loss=0.09119, over 8025.00 frames. ], tot_loss[loss=0.1795, simple_loss=0.2678, pruned_loss=0.04553, over 525187.50 frames. ], batch size: 98, lr: 4.09e-03, grad_scale: 16.0 2023-05-16 13:48:00,019 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.719e+02 3.186e+02 3.722e+02 4.629e+02 6.830e+02, threshold=7.444e+02, percent-clipped=0.0 2023-05-16 13:48:06,467 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=210166.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 13:48:09,302 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=210170.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 13:48:22,003 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.3948, 4.7690, 3.1656, 2.7693, 4.1169, 2.7456, 4.1333, 3.4868], device='cuda:0'), covar=tensor([0.0709, 0.0628, 0.1034, 0.1579, 0.0297, 0.1423, 0.0544, 0.0789], device='cuda:0'), in_proj_covar=tensor([0.0180, 0.0237, 0.0170, 0.0191, 0.0132, 0.0174, 0.0184, 0.0169], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 13:48:27,746 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.0310, 4.9399, 4.7743, 4.8657, 4.5185, 4.9486, 4.9200, 5.1044], device='cuda:0'), covar=tensor([0.0172, 0.0125, 0.0191, 0.0329, 0.0675, 0.0234, 0.0138, 0.0182], device='cuda:0'), in_proj_covar=tensor([0.0167, 0.0166, 0.0162, 0.0209, 0.0206, 0.0183, 0.0151, 0.0194], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:0') 2023-05-16 13:48:34,914 INFO [finetune.py:992] (0/2) Epoch 10, batch 100, loss[loss=0.1739, simple_loss=0.2698, pruned_loss=0.03902, over 12288.00 frames. ], tot_loss[loss=0.1793, simple_loss=0.2685, pruned_loss=0.04509, over 938645.79 frames. ], batch size: 33, lr: 4.09e-03, grad_scale: 16.0 2023-05-16 13:48:41,395 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=210214.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 13:48:46,977 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.2448, 2.7677, 3.7824, 3.2820, 3.6216, 3.2822, 2.7057, 3.6228], device='cuda:0'), covar=tensor([0.0134, 0.0335, 0.0148, 0.0232, 0.0153, 0.0219, 0.0392, 0.0142], device='cuda:0'), in_proj_covar=tensor([0.0159, 0.0189, 0.0168, 0.0169, 0.0194, 0.0147, 0.0183, 0.0165], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 13:49:11,050 INFO [finetune.py:992] (0/2) Epoch 10, batch 150, loss[loss=0.1585, simple_loss=0.2503, pruned_loss=0.03337, over 12099.00 frames. ], tot_loss[loss=0.1774, simple_loss=0.2666, pruned_loss=0.04413, over 1258635.84 frames. ], batch size: 33, lr: 4.09e-03, grad_scale: 16.0 2023-05-16 13:49:12,471 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.846e+02 2.678e+02 3.225e+02 3.794e+02 7.055e+02, threshold=6.451e+02, percent-clipped=0.0 2023-05-16 13:49:16,096 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=210262.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 13:49:36,790 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.1714, 3.7992, 3.9871, 4.3320, 2.8386, 3.7378, 2.6328, 3.9922], device='cuda:0'), covar=tensor([0.1737, 0.0857, 0.0998, 0.0679, 0.1224, 0.0712, 0.1891, 0.1209], device='cuda:0'), in_proj_covar=tensor([0.0221, 0.0262, 0.0289, 0.0343, 0.0231, 0.0236, 0.0257, 0.0366], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 13:49:46,197 INFO [finetune.py:992] (0/2) Epoch 10, batch 200, loss[loss=0.1909, simple_loss=0.284, pruned_loss=0.04887, over 12075.00 frames. ], tot_loss[loss=0.1761, simple_loss=0.2651, pruned_loss=0.04356, over 1499921.77 frames. ], batch size: 42, lr: 4.09e-03, grad_scale: 16.0 2023-05-16 13:49:49,924 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=210310.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 13:50:14,361 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=210344.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 13:50:21,963 INFO [finetune.py:992] (0/2) Epoch 10, batch 250, loss[loss=0.1725, simple_loss=0.2609, pruned_loss=0.04203, over 12083.00 frames. ], tot_loss[loss=0.1763, simple_loss=0.2653, pruned_loss=0.04371, over 1701451.04 frames. ], batch size: 32, lr: 4.09e-03, grad_scale: 16.0 2023-05-16 13:50:23,373 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.066e+02 2.825e+02 3.486e+02 4.026e+02 8.795e+02, threshold=6.972e+02, percent-clipped=3.0 2023-05-16 13:50:44,176 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.5239, 4.2523, 4.2875, 4.3070, 4.2877, 4.4908, 4.4221, 2.4246], device='cuda:0'), covar=tensor([0.0129, 0.0084, 0.0120, 0.0091, 0.0067, 0.0097, 0.0110, 0.0865], device='cuda:0'), in_proj_covar=tensor([0.0064, 0.0074, 0.0077, 0.0070, 0.0057, 0.0087, 0.0075, 0.0094], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-16 13:50:49,017 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=210392.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 13:50:58,655 INFO [finetune.py:992] (0/2) Epoch 10, batch 300, loss[loss=0.1336, simple_loss=0.2222, pruned_loss=0.02254, over 12013.00 frames. ], tot_loss[loss=0.1758, simple_loss=0.2648, pruned_loss=0.04344, over 1851940.91 frames. ], batch size: 28, lr: 4.09e-03, grad_scale: 16.0 2023-05-16 13:51:04,411 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=210413.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 13:51:34,139 INFO [finetune.py:992] (0/2) Epoch 10, batch 350, loss[loss=0.174, simple_loss=0.2626, pruned_loss=0.04268, over 12290.00 frames. ], tot_loss[loss=0.174, simple_loss=0.2629, pruned_loss=0.04249, over 1969334.92 frames. ], batch size: 34, lr: 4.09e-03, grad_scale: 16.0 2023-05-16 13:51:35,573 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.046e+02 2.842e+02 3.213e+02 3.857e+02 1.790e+03, threshold=6.425e+02, percent-clipped=1.0 2023-05-16 13:51:49,786 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.5716, 5.3910, 5.4621, 5.5198, 5.1554, 5.2075, 4.9621, 5.4416], device='cuda:0'), covar=tensor([0.0549, 0.0515, 0.0770, 0.0548, 0.1862, 0.1306, 0.0521, 0.1146], device='cuda:0'), in_proj_covar=tensor([0.0489, 0.0644, 0.0561, 0.0569, 0.0772, 0.0682, 0.0506, 0.0444], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0004, 0.0004, 0.0005, 0.0004, 0.0003, 0.0003], device='cuda:0') 2023-05-16 13:51:49,822 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.0391, 4.6894, 4.8441, 4.8805, 4.7598, 5.0597, 4.8479, 2.7346], device='cuda:0'), covar=tensor([0.0136, 0.0092, 0.0118, 0.0079, 0.0057, 0.0094, 0.0103, 0.0845], device='cuda:0'), in_proj_covar=tensor([0.0064, 0.0074, 0.0077, 0.0070, 0.0057, 0.0087, 0.0076, 0.0094], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-16 13:52:10,242 INFO [finetune.py:992] (0/2) Epoch 10, batch 400, loss[loss=0.1844, simple_loss=0.2832, pruned_loss=0.04284, over 12367.00 frames. ], tot_loss[loss=0.1736, simple_loss=0.2621, pruned_loss=0.04251, over 2058458.58 frames. ], batch size: 38, lr: 4.09e-03, grad_scale: 16.0 2023-05-16 13:52:20,371 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.83 vs. limit=2.0 2023-05-16 13:52:45,814 INFO [finetune.py:992] (0/2) Epoch 10, batch 450, loss[loss=0.1634, simple_loss=0.2554, pruned_loss=0.03571, over 12026.00 frames. ], tot_loss[loss=0.1744, simple_loss=0.263, pruned_loss=0.04289, over 2135031.63 frames. ], batch size: 31, lr: 4.09e-03, grad_scale: 16.0 2023-05-16 13:52:47,257 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.061e+02 2.781e+02 3.270e+02 3.985e+02 2.182e+03, threshold=6.539e+02, percent-clipped=5.0 2023-05-16 13:53:21,555 INFO [finetune.py:992] (0/2) Epoch 10, batch 500, loss[loss=0.1788, simple_loss=0.2697, pruned_loss=0.04397, over 12288.00 frames. ], tot_loss[loss=0.1738, simple_loss=0.2626, pruned_loss=0.04245, over 2193557.40 frames. ], batch size: 33, lr: 4.09e-03, grad_scale: 16.0 2023-05-16 13:53:32,416 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.2944, 4.8561, 5.2762, 4.6178, 4.8610, 4.6561, 5.3060, 5.0258], device='cuda:0'), covar=tensor([0.0283, 0.0432, 0.0330, 0.0275, 0.0467, 0.0326, 0.0236, 0.0274], device='cuda:0'), in_proj_covar=tensor([0.0241, 0.0246, 0.0268, 0.0242, 0.0244, 0.0243, 0.0219, 0.0202], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 13:53:39,591 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.3544, 3.8957, 3.8523, 4.3291, 2.7585, 3.8863, 2.5876, 3.9661], device='cuda:0'), covar=tensor([0.1516, 0.0765, 0.1051, 0.0617, 0.1209, 0.0607, 0.1785, 0.1271], device='cuda:0'), in_proj_covar=tensor([0.0220, 0.0259, 0.0289, 0.0342, 0.0230, 0.0234, 0.0256, 0.0368], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 13:53:41,660 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=210632.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 13:53:57,445 INFO [finetune.py:992] (0/2) Epoch 10, batch 550, loss[loss=0.1619, simple_loss=0.2573, pruned_loss=0.03331, over 12115.00 frames. ], tot_loss[loss=0.1736, simple_loss=0.2627, pruned_loss=0.04232, over 2240727.59 frames. ], batch size: 39, lr: 4.09e-03, grad_scale: 16.0 2023-05-16 13:53:58,860 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.786e+02 2.661e+02 2.995e+02 3.654e+02 5.423e+02, threshold=5.990e+02, percent-clipped=0.0 2023-05-16 13:54:05,433 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=210666.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 13:54:17,285 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=210682.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 13:54:25,077 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=210693.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 13:54:31,526 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.7571, 3.5206, 5.1944, 2.7527, 2.7242, 3.8584, 3.1815, 4.0199], device='cuda:0'), covar=tensor([0.0466, 0.1032, 0.0305, 0.1192, 0.2043, 0.1443, 0.1419, 0.1066], device='cuda:0'), in_proj_covar=tensor([0.0229, 0.0229, 0.0235, 0.0179, 0.0233, 0.0284, 0.0218, 0.0252], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:0') 2023-05-16 13:54:33,356 INFO [finetune.py:992] (0/2) Epoch 10, batch 600, loss[loss=0.1613, simple_loss=0.2526, pruned_loss=0.03499, over 12152.00 frames. ], tot_loss[loss=0.1732, simple_loss=0.262, pruned_loss=0.04219, over 2274109.09 frames. ], batch size: 34, lr: 4.09e-03, grad_scale: 16.0 2023-05-16 13:54:39,291 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=210713.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 13:54:49,249 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=210727.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 13:55:00,607 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=210743.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 13:55:09,062 INFO [finetune.py:992] (0/2) Epoch 10, batch 650, loss[loss=0.1794, simple_loss=0.2686, pruned_loss=0.04507, over 12329.00 frames. ], tot_loss[loss=0.1726, simple_loss=0.2611, pruned_loss=0.042, over 2295514.84 frames. ], batch size: 36, lr: 4.08e-03, grad_scale: 16.0 2023-05-16 13:55:10,411 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.114e+02 2.764e+02 3.125e+02 3.796e+02 1.069e+03, threshold=6.251e+02, percent-clipped=3.0 2023-05-16 13:55:13,963 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=210761.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 13:55:16,184 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=210764.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 13:55:45,235 INFO [finetune.py:992] (0/2) Epoch 10, batch 700, loss[loss=0.1366, simple_loss=0.2185, pruned_loss=0.02737, over 12017.00 frames. ], tot_loss[loss=0.1716, simple_loss=0.2602, pruned_loss=0.04154, over 2314079.82 frames. ], batch size: 28, lr: 4.08e-03, grad_scale: 16.0 2023-05-16 13:56:00,272 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=210825.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 13:56:21,291 INFO [finetune.py:992] (0/2) Epoch 10, batch 750, loss[loss=0.1656, simple_loss=0.26, pruned_loss=0.03557, over 12349.00 frames. ], tot_loss[loss=0.1725, simple_loss=0.2609, pruned_loss=0.04206, over 2321337.86 frames. ], batch size: 35, lr: 4.08e-03, grad_scale: 16.0 2023-05-16 13:56:22,748 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.752e+02 2.676e+02 3.272e+02 3.735e+02 7.216e+02, threshold=6.544e+02, percent-clipped=2.0 2023-05-16 13:56:29,319 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.4198, 2.7536, 3.5269, 4.3815, 3.8103, 4.3107, 3.8814, 3.0617], device='cuda:0'), covar=tensor([0.0032, 0.0318, 0.0157, 0.0039, 0.0104, 0.0067, 0.0122, 0.0346], device='cuda:0'), in_proj_covar=tensor([0.0083, 0.0122, 0.0102, 0.0073, 0.0100, 0.0113, 0.0092, 0.0135], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 13:56:47,759 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.3628, 4.7141, 4.1057, 5.0507, 4.5166, 3.0039, 4.3058, 3.0490], device='cuda:0'), covar=tensor([0.0808, 0.0819, 0.1382, 0.0471, 0.1177, 0.1590, 0.1088, 0.3371], device='cuda:0'), in_proj_covar=tensor([0.0296, 0.0371, 0.0345, 0.0285, 0.0358, 0.0263, 0.0335, 0.0356], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 13:56:48,416 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.1569, 5.0827, 4.9409, 4.9458, 4.6075, 5.1058, 5.0704, 5.2762], device='cuda:0'), covar=tensor([0.0172, 0.0128, 0.0179, 0.0322, 0.0788, 0.0238, 0.0158, 0.0163], device='cuda:0'), in_proj_covar=tensor([0.0182, 0.0181, 0.0177, 0.0228, 0.0224, 0.0200, 0.0164, 0.0210], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:0') 2023-05-16 13:56:57,493 INFO [finetune.py:992] (0/2) Epoch 10, batch 800, loss[loss=0.1379, simple_loss=0.2197, pruned_loss=0.02809, over 11380.00 frames. ], tot_loss[loss=0.1719, simple_loss=0.2602, pruned_loss=0.04173, over 2329038.95 frames. ], batch size: 25, lr: 4.08e-03, grad_scale: 16.0 2023-05-16 13:57:09,727 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.3029, 4.6769, 2.7379, 2.6256, 3.9924, 2.6167, 3.9585, 3.0940], device='cuda:0'), covar=tensor([0.0653, 0.0649, 0.1289, 0.1701, 0.0329, 0.1378, 0.0575, 0.0963], device='cuda:0'), in_proj_covar=tensor([0.0183, 0.0245, 0.0173, 0.0195, 0.0136, 0.0177, 0.0190, 0.0173], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 13:57:32,610 INFO [finetune.py:992] (0/2) Epoch 10, batch 850, loss[loss=0.1842, simple_loss=0.2729, pruned_loss=0.04778, over 12257.00 frames. ], tot_loss[loss=0.1726, simple_loss=0.2606, pruned_loss=0.04234, over 2328855.62 frames. ], batch size: 32, lr: 4.08e-03, grad_scale: 16.0 2023-05-16 13:57:34,560 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.020e+02 2.920e+02 3.334e+02 3.958e+02 9.193e+02, threshold=6.668e+02, percent-clipped=2.0 2023-05-16 13:57:46,098 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.3082, 4.7027, 3.9263, 4.9410, 4.4656, 2.8887, 4.2093, 2.9603], device='cuda:0'), covar=tensor([0.0836, 0.0720, 0.1591, 0.0495, 0.1198, 0.1690, 0.1089, 0.3551], device='cuda:0'), in_proj_covar=tensor([0.0295, 0.0370, 0.0344, 0.0284, 0.0357, 0.0263, 0.0333, 0.0356], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 13:57:52,263 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.2634, 5.1161, 4.9908, 5.0506, 4.7137, 5.1857, 5.1655, 5.3174], device='cuda:0'), covar=tensor([0.0179, 0.0120, 0.0169, 0.0312, 0.0689, 0.0231, 0.0153, 0.0145], device='cuda:0'), in_proj_covar=tensor([0.0182, 0.0182, 0.0179, 0.0229, 0.0226, 0.0202, 0.0165, 0.0211], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:0') 2023-05-16 13:57:53,730 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=210984.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 13:57:56,481 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=210988.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 13:58:08,255 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.66 vs. limit=2.0 2023-05-16 13:58:08,592 INFO [finetune.py:992] (0/2) Epoch 10, batch 900, loss[loss=0.1751, simple_loss=0.265, pruned_loss=0.04257, over 12289.00 frames. ], tot_loss[loss=0.1719, simple_loss=0.2599, pruned_loss=0.04196, over 2345057.10 frames. ], batch size: 33, lr: 4.08e-03, grad_scale: 16.0 2023-05-16 13:58:14,813 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.74 vs. limit=2.0 2023-05-16 13:58:15,952 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-05-16 13:58:20,742 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=211022.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 13:58:27,331 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.0901, 4.5764, 2.6651, 2.4378, 3.9738, 2.1278, 3.9504, 3.0222], device='cuda:0'), covar=tensor([0.0811, 0.0560, 0.1344, 0.1824, 0.0296, 0.1665, 0.0513, 0.0926], device='cuda:0'), in_proj_covar=tensor([0.0183, 0.0245, 0.0173, 0.0195, 0.0136, 0.0176, 0.0190, 0.0173], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 13:58:32,101 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=211038.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 13:58:37,900 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=211045.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 13:58:44,833 INFO [finetune.py:992] (0/2) Epoch 10, batch 950, loss[loss=0.1497, simple_loss=0.2311, pruned_loss=0.03414, over 11795.00 frames. ], tot_loss[loss=0.1712, simple_loss=0.2592, pruned_loss=0.04154, over 2351945.57 frames. ], batch size: 26, lr: 4.08e-03, grad_scale: 16.0 2023-05-16 13:58:46,244 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.673e+02 2.872e+02 3.267e+02 3.845e+02 7.278e+02, threshold=6.535e+02, percent-clipped=1.0 2023-05-16 13:58:54,338 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.2410, 4.0096, 4.0102, 4.3585, 3.0605, 4.0872, 2.6731, 4.1286], device='cuda:0'), covar=tensor([0.1701, 0.0758, 0.1037, 0.0689, 0.1082, 0.0556, 0.1781, 0.1508], device='cuda:0'), in_proj_covar=tensor([0.0224, 0.0263, 0.0292, 0.0347, 0.0234, 0.0236, 0.0258, 0.0373], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 13:59:20,409 INFO [finetune.py:992] (0/2) Epoch 10, batch 1000, loss[loss=0.1308, simple_loss=0.2155, pruned_loss=0.02312, over 12034.00 frames. ], tot_loss[loss=0.1697, simple_loss=0.2582, pruned_loss=0.04062, over 2369289.03 frames. ], batch size: 28, lr: 4.08e-03, grad_scale: 16.0 2023-05-16 13:59:28,954 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.9951, 6.0204, 5.7130, 5.2332, 5.1050, 5.8868, 5.4552, 5.2560], device='cuda:0'), covar=tensor([0.0763, 0.0720, 0.0700, 0.1595, 0.0691, 0.0788, 0.1516, 0.1081], device='cuda:0'), in_proj_covar=tensor([0.0594, 0.0534, 0.0497, 0.0607, 0.0404, 0.0685, 0.0753, 0.0549], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-05-16 13:59:31,777 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=211120.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 13:59:56,493 INFO [finetune.py:992] (0/2) Epoch 10, batch 1050, loss[loss=0.1494, simple_loss=0.2342, pruned_loss=0.03233, over 12088.00 frames. ], tot_loss[loss=0.1695, simple_loss=0.258, pruned_loss=0.04046, over 2372941.50 frames. ], batch size: 32, lr: 4.08e-03, grad_scale: 16.0 2023-05-16 13:59:57,914 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.833e+02 2.801e+02 3.196e+02 3.803e+02 6.712e+02, threshold=6.392e+02, percent-clipped=1.0 2023-05-16 14:00:32,881 INFO [finetune.py:992] (0/2) Epoch 10, batch 1100, loss[loss=0.1822, simple_loss=0.2725, pruned_loss=0.04595, over 12192.00 frames. ], tot_loss[loss=0.17, simple_loss=0.2588, pruned_loss=0.04063, over 2376145.47 frames. ], batch size: 35, lr: 4.08e-03, grad_scale: 16.0 2023-05-16 14:01:08,087 INFO [finetune.py:992] (0/2) Epoch 10, batch 1150, loss[loss=0.1661, simple_loss=0.2604, pruned_loss=0.03591, over 12339.00 frames. ], tot_loss[loss=0.1697, simple_loss=0.2584, pruned_loss=0.04053, over 2380742.56 frames. ], batch size: 36, lr: 4.08e-03, grad_scale: 16.0 2023-05-16 14:01:09,500 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.896e+02 2.597e+02 3.087e+02 3.883e+02 5.758e+02, threshold=6.174e+02, percent-clipped=0.0 2023-05-16 14:01:22,615 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.9999, 2.4206, 3.5007, 3.0502, 3.2789, 3.1360, 2.4054, 3.3581], device='cuda:0'), covar=tensor([0.0140, 0.0356, 0.0131, 0.0211, 0.0157, 0.0161, 0.0345, 0.0149], device='cuda:0'), in_proj_covar=tensor([0.0168, 0.0199, 0.0180, 0.0178, 0.0207, 0.0155, 0.0192, 0.0175], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 14:01:32,444 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=211288.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 14:01:43,864 INFO [finetune.py:992] (0/2) Epoch 10, batch 1200, loss[loss=0.2021, simple_loss=0.2846, pruned_loss=0.05981, over 12113.00 frames. ], tot_loss[loss=0.1691, simple_loss=0.2575, pruned_loss=0.04033, over 2383387.33 frames. ], batch size: 39, lr: 4.08e-03, grad_scale: 16.0 2023-05-16 14:01:44,989 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.41 vs. limit=2.0 2023-05-16 14:01:56,210 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=211322.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 14:02:06,144 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=211336.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 14:02:07,631 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=211338.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 14:02:08,934 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=211340.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 14:02:14,809 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.7403, 4.5529, 4.5308, 4.5024, 4.5444, 4.7686, 4.6610, 2.4012], device='cuda:0'), covar=tensor([0.0182, 0.0083, 0.0133, 0.0112, 0.0064, 0.0121, 0.0109, 0.0986], device='cuda:0'), in_proj_covar=tensor([0.0067, 0.0077, 0.0080, 0.0072, 0.0059, 0.0090, 0.0079, 0.0097], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-16 14:02:20,154 INFO [finetune.py:992] (0/2) Epoch 10, batch 1250, loss[loss=0.1814, simple_loss=0.2723, pruned_loss=0.04524, over 11763.00 frames. ], tot_loss[loss=0.1687, simple_loss=0.2572, pruned_loss=0.04014, over 2381250.58 frames. ], batch size: 44, lr: 4.08e-03, grad_scale: 16.0 2023-05-16 14:02:21,502 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.930e+02 2.891e+02 3.376e+02 3.934e+02 8.670e+02, threshold=6.752e+02, percent-clipped=2.0 2023-05-16 14:02:30,888 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=211370.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 14:02:42,015 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=211386.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 14:02:42,138 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=211386.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 14:02:55,709 INFO [finetune.py:992] (0/2) Epoch 10, batch 1300, loss[loss=0.155, simple_loss=0.2535, pruned_loss=0.02825, over 12194.00 frames. ], tot_loss[loss=0.1684, simple_loss=0.2569, pruned_loss=0.03996, over 2385661.34 frames. ], batch size: 35, lr: 4.08e-03, grad_scale: 8.0 2023-05-16 14:03:00,799 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=211412.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 14:03:07,004 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=211420.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 14:03:26,574 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=211447.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 14:03:31,999 INFO [finetune.py:992] (0/2) Epoch 10, batch 1350, loss[loss=0.1369, simple_loss=0.2239, pruned_loss=0.02489, over 12019.00 frames. ], tot_loss[loss=0.1691, simple_loss=0.2577, pruned_loss=0.04026, over 2382562.66 frames. ], batch size: 31, lr: 4.08e-03, grad_scale: 8.0 2023-05-16 14:03:34,131 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.951e+02 2.733e+02 3.111e+02 3.708e+02 5.238e+02, threshold=6.222e+02, percent-clipped=0.0 2023-05-16 14:03:34,381 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.6316, 3.7462, 3.2674, 3.2952, 2.9607, 2.8817, 3.7604, 2.3995], device='cuda:0'), covar=tensor([0.0349, 0.0136, 0.0192, 0.0196, 0.0369, 0.0350, 0.0107, 0.0463], device='cuda:0'), in_proj_covar=tensor([0.0190, 0.0159, 0.0157, 0.0184, 0.0200, 0.0199, 0.0166, 0.0199], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 14:03:41,372 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=211468.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 14:03:45,034 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=211473.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 14:04:07,654 INFO [finetune.py:992] (0/2) Epoch 10, batch 1400, loss[loss=0.203, simple_loss=0.2831, pruned_loss=0.06138, over 7781.00 frames. ], tot_loss[loss=0.1691, simple_loss=0.2575, pruned_loss=0.04032, over 2382019.59 frames. ], batch size: 98, lr: 4.08e-03, grad_scale: 8.0 2023-05-16 14:04:39,787 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.40 vs. limit=2.0 2023-05-16 14:04:43,931 INFO [finetune.py:992] (0/2) Epoch 10, batch 1450, loss[loss=0.2088, simple_loss=0.2927, pruned_loss=0.06246, over 12077.00 frames. ], tot_loss[loss=0.1692, simple_loss=0.258, pruned_loss=0.04025, over 2383161.36 frames. ], batch size: 42, lr: 4.08e-03, grad_scale: 8.0 2023-05-16 14:04:46,038 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.925e+02 2.756e+02 3.139e+02 3.742e+02 1.134e+03, threshold=6.278e+02, percent-clipped=4.0 2023-05-16 14:04:52,449 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([6.1435, 6.0513, 5.8503, 5.3171, 5.1856, 6.0048, 5.6603, 5.4405], device='cuda:0'), covar=tensor([0.0652, 0.1058, 0.0674, 0.1594, 0.0714, 0.0761, 0.1458, 0.1038], device='cuda:0'), in_proj_covar=tensor([0.0607, 0.0544, 0.0504, 0.0619, 0.0410, 0.0697, 0.0766, 0.0557], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0004, 0.0004, 0.0003], device='cuda:0') 2023-05-16 14:05:20,304 INFO [finetune.py:992] (0/2) Epoch 10, batch 1500, loss[loss=0.1802, simple_loss=0.2731, pruned_loss=0.04367, over 12281.00 frames. ], tot_loss[loss=0.1697, simple_loss=0.2586, pruned_loss=0.04038, over 2378835.82 frames. ], batch size: 37, lr: 4.08e-03, grad_scale: 8.0 2023-05-16 14:05:45,310 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=211640.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 14:05:55,972 INFO [finetune.py:992] (0/2) Epoch 10, batch 1550, loss[loss=0.1592, simple_loss=0.2367, pruned_loss=0.04084, over 11785.00 frames. ], tot_loss[loss=0.1703, simple_loss=0.2593, pruned_loss=0.04066, over 2377789.77 frames. ], batch size: 26, lr: 4.08e-03, grad_scale: 8.0 2023-05-16 14:05:58,007 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.015e+02 2.636e+02 3.063e+02 3.605e+02 8.321e+02, threshold=6.125e+02, percent-clipped=1.0 2023-05-16 14:06:20,236 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=211688.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 14:06:31,970 INFO [finetune.py:992] (0/2) Epoch 10, batch 1600, loss[loss=0.1972, simple_loss=0.2849, pruned_loss=0.05471, over 11817.00 frames. ], tot_loss[loss=0.1708, simple_loss=0.2598, pruned_loss=0.04087, over 2381152.20 frames. ], batch size: 44, lr: 4.08e-03, grad_scale: 8.0 2023-05-16 14:06:41,462 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.3733, 4.2460, 4.0618, 4.5198, 2.9320, 4.0935, 2.8953, 4.1835], device='cuda:0'), covar=tensor([0.1512, 0.0643, 0.0899, 0.0658, 0.1111, 0.0561, 0.1567, 0.1356], device='cuda:0'), in_proj_covar=tensor([0.0222, 0.0261, 0.0290, 0.0348, 0.0232, 0.0235, 0.0255, 0.0371], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 14:06:58,872 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=211742.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 14:07:03,947 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.2273, 2.5197, 3.5959, 4.1626, 3.7136, 4.1455, 3.8603, 2.7873], device='cuda:0'), covar=tensor([0.0051, 0.0436, 0.0147, 0.0042, 0.0143, 0.0082, 0.0118, 0.0450], device='cuda:0'), in_proj_covar=tensor([0.0085, 0.0124, 0.0104, 0.0075, 0.0101, 0.0114, 0.0093, 0.0138], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 14:07:07,997 INFO [finetune.py:992] (0/2) Epoch 10, batch 1650, loss[loss=0.1822, simple_loss=0.2691, pruned_loss=0.04765, over 11980.00 frames. ], tot_loss[loss=0.1705, simple_loss=0.259, pruned_loss=0.04099, over 2383287.17 frames. ], batch size: 42, lr: 4.07e-03, grad_scale: 8.0 2023-05-16 14:07:10,074 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.745e+02 2.665e+02 3.073e+02 3.607e+02 6.323e+02, threshold=6.146e+02, percent-clipped=2.0 2023-05-16 14:07:17,121 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=211768.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 14:07:38,193 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=3.33 vs. limit=5.0 2023-05-16 14:07:41,097 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.7028, 2.8164, 4.5852, 4.7168, 2.8842, 2.5682, 2.9244, 2.1237], device='cuda:0'), covar=tensor([0.1557, 0.2983, 0.0416, 0.0420, 0.1262, 0.2414, 0.2722, 0.4140], device='cuda:0'), in_proj_covar=tensor([0.0284, 0.0379, 0.0268, 0.0292, 0.0263, 0.0298, 0.0370, 0.0369], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-16 14:07:43,697 INFO [finetune.py:992] (0/2) Epoch 10, batch 1700, loss[loss=0.2012, simple_loss=0.2972, pruned_loss=0.05256, over 12048.00 frames. ], tot_loss[loss=0.1706, simple_loss=0.2594, pruned_loss=0.04088, over 2382397.86 frames. ], batch size: 37, lr: 4.07e-03, grad_scale: 8.0 2023-05-16 14:07:54,983 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=4.84 vs. limit=5.0 2023-05-16 14:08:01,026 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.7067, 3.0175, 3.8284, 4.6045, 3.9542, 4.5631, 4.0910, 3.3660], device='cuda:0'), covar=tensor([0.0026, 0.0303, 0.0132, 0.0038, 0.0123, 0.0059, 0.0091, 0.0312], device='cuda:0'), in_proj_covar=tensor([0.0084, 0.0123, 0.0104, 0.0075, 0.0101, 0.0114, 0.0092, 0.0137], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 14:08:20,032 INFO [finetune.py:992] (0/2) Epoch 10, batch 1750, loss[loss=0.1616, simple_loss=0.2494, pruned_loss=0.03695, over 12256.00 frames. ], tot_loss[loss=0.1694, simple_loss=0.2583, pruned_loss=0.04025, over 2382097.11 frames. ], batch size: 32, lr: 4.07e-03, grad_scale: 8.0 2023-05-16 14:08:22,202 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.733e+02 2.699e+02 3.136e+02 3.576e+02 9.797e+02, threshold=6.272e+02, percent-clipped=1.0 2023-05-16 14:08:50,788 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.5039, 3.5160, 3.2153, 3.1365, 2.8498, 2.6772, 3.5266, 2.1501], device='cuda:0'), covar=tensor([0.0386, 0.0154, 0.0185, 0.0209, 0.0396, 0.0371, 0.0122, 0.0527], device='cuda:0'), in_proj_covar=tensor([0.0190, 0.0160, 0.0158, 0.0185, 0.0201, 0.0199, 0.0167, 0.0199], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 14:08:56,249 INFO [finetune.py:992] (0/2) Epoch 10, batch 1800, loss[loss=0.1805, simple_loss=0.2727, pruned_loss=0.04417, over 12193.00 frames. ], tot_loss[loss=0.1691, simple_loss=0.2578, pruned_loss=0.04018, over 2384034.51 frames. ], batch size: 35, lr: 4.07e-03, grad_scale: 8.0 2023-05-16 14:09:32,014 INFO [finetune.py:992] (0/2) Epoch 10, batch 1850, loss[loss=0.1828, simple_loss=0.2806, pruned_loss=0.04255, over 12215.00 frames. ], tot_loss[loss=0.1688, simple_loss=0.2576, pruned_loss=0.03995, over 2385055.56 frames. ], batch size: 35, lr: 4.07e-03, grad_scale: 8.0 2023-05-16 14:09:34,861 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.930e+02 2.726e+02 3.285e+02 4.068e+02 5.793e+02, threshold=6.571e+02, percent-clipped=0.0 2023-05-16 14:09:49,419 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([6.2348, 6.0971, 6.0148, 5.4903, 5.2492, 6.0949, 5.6859, 5.4905], device='cuda:0'), covar=tensor([0.0568, 0.1070, 0.0565, 0.1581, 0.0640, 0.0715, 0.1561, 0.0940], device='cuda:0'), in_proj_covar=tensor([0.0606, 0.0546, 0.0503, 0.0620, 0.0411, 0.0700, 0.0770, 0.0560], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0004, 0.0004, 0.0003], device='cuda:0') 2023-05-16 14:10:00,069 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.8811, 4.8175, 4.7024, 4.7610, 4.3443, 4.8736, 4.8543, 5.0355], device='cuda:0'), covar=tensor([0.0210, 0.0142, 0.0178, 0.0293, 0.0804, 0.0261, 0.0158, 0.0183], device='cuda:0'), in_proj_covar=tensor([0.0193, 0.0193, 0.0189, 0.0241, 0.0239, 0.0214, 0.0173, 0.0222], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0005, 0.0004, 0.0004, 0.0003, 0.0005], device='cuda:0') 2023-05-16 14:10:05,030 INFO [checkpoint.py:75] (0/2) Saving checkpoint to pruned_transducer_stateless7/exp_giga_finetune/checkpoint-112000.pt 2023-05-16 14:10:11,303 INFO [finetune.py:992] (0/2) Epoch 10, batch 1900, loss[loss=0.1812, simple_loss=0.2732, pruned_loss=0.04462, over 11558.00 frames. ], tot_loss[loss=0.1691, simple_loss=0.258, pruned_loss=0.0401, over 2385528.43 frames. ], batch size: 48, lr: 4.07e-03, grad_scale: 8.0 2023-05-16 14:10:27,645 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.6340, 3.1502, 4.9713, 2.6440, 2.6072, 3.7705, 3.0936, 3.7122], device='cuda:0'), covar=tensor([0.0397, 0.1204, 0.0363, 0.1131, 0.2029, 0.1342, 0.1401, 0.1189], device='cuda:0'), in_proj_covar=tensor([0.0237, 0.0236, 0.0244, 0.0182, 0.0240, 0.0292, 0.0224, 0.0260], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-16 14:10:38,008 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=212042.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 14:10:47,251 INFO [finetune.py:992] (0/2) Epoch 10, batch 1950, loss[loss=0.1814, simple_loss=0.2741, pruned_loss=0.04432, over 11414.00 frames. ], tot_loss[loss=0.17, simple_loss=0.2591, pruned_loss=0.04041, over 2385371.32 frames. ], batch size: 55, lr: 4.07e-03, grad_scale: 8.0 2023-05-16 14:10:49,204 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.965e+02 2.785e+02 3.345e+02 3.996e+02 7.616e+02, threshold=6.690e+02, percent-clipped=2.0 2023-05-16 14:10:56,398 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=212068.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 14:11:12,129 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=212090.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 14:11:22,832 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=212104.0, num_to_drop=1, layers_to_drop={0} 2023-05-16 14:11:23,360 INFO [finetune.py:992] (0/2) Epoch 10, batch 2000, loss[loss=0.1497, simple_loss=0.2248, pruned_loss=0.03734, over 11860.00 frames. ], tot_loss[loss=0.1704, simple_loss=0.2594, pruned_loss=0.04075, over 2384747.19 frames. ], batch size: 26, lr: 4.07e-03, grad_scale: 8.0 2023-05-16 14:11:31,260 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=212116.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 14:11:59,550 INFO [finetune.py:992] (0/2) Epoch 10, batch 2050, loss[loss=0.1717, simple_loss=0.2597, pruned_loss=0.04187, over 12351.00 frames. ], tot_loss[loss=0.1698, simple_loss=0.2587, pruned_loss=0.04049, over 2376801.76 frames. ], batch size: 35, lr: 4.07e-03, grad_scale: 8.0 2023-05-16 14:12:02,300 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.742e+02 2.686e+02 3.253e+02 3.860e+02 1.028e+03, threshold=6.506e+02, percent-clipped=1.0 2023-05-16 14:12:07,638 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=212165.0, num_to_drop=1, layers_to_drop={2} 2023-05-16 14:12:09,721 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.1582, 2.5593, 3.6348, 3.0599, 3.4533, 3.2130, 2.5722, 3.5120], device='cuda:0'), covar=tensor([0.0128, 0.0329, 0.0128, 0.0234, 0.0161, 0.0164, 0.0360, 0.0122], device='cuda:0'), in_proj_covar=tensor([0.0170, 0.0200, 0.0183, 0.0180, 0.0209, 0.0157, 0.0193, 0.0177], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 14:12:36,293 INFO [finetune.py:992] (0/2) Epoch 10, batch 2100, loss[loss=0.1933, simple_loss=0.2846, pruned_loss=0.05105, over 12137.00 frames. ], tot_loss[loss=0.1708, simple_loss=0.2598, pruned_loss=0.04093, over 2380329.09 frames. ], batch size: 39, lr: 4.07e-03, grad_scale: 8.0 2023-05-16 14:13:11,172 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=212253.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 14:13:12,385 INFO [finetune.py:992] (0/2) Epoch 10, batch 2150, loss[loss=0.1483, simple_loss=0.2351, pruned_loss=0.03076, over 12244.00 frames. ], tot_loss[loss=0.171, simple_loss=0.2604, pruned_loss=0.04076, over 2388062.18 frames. ], batch size: 32, lr: 4.07e-03, grad_scale: 8.0 2023-05-16 14:13:14,585 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.677e+02 2.827e+02 3.250e+02 3.793e+02 5.770e+02, threshold=6.500e+02, percent-clipped=0.0 2023-05-16 14:13:18,946 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.2853, 4.9059, 5.2687, 4.5623, 4.9170, 4.7163, 5.3327, 4.9224], device='cuda:0'), covar=tensor([0.0269, 0.0370, 0.0284, 0.0256, 0.0380, 0.0300, 0.0187, 0.0372], device='cuda:0'), in_proj_covar=tensor([0.0254, 0.0258, 0.0283, 0.0252, 0.0255, 0.0253, 0.0229, 0.0212], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 14:13:43,807 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.5030, 2.7469, 3.3631, 4.4441, 2.5462, 4.4944, 4.5275, 4.6097], device='cuda:0'), covar=tensor([0.0133, 0.1169, 0.0455, 0.0157, 0.1304, 0.0205, 0.0139, 0.0102], device='cuda:0'), in_proj_covar=tensor([0.0121, 0.0203, 0.0184, 0.0116, 0.0190, 0.0177, 0.0172, 0.0117], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 14:13:48,623 INFO [finetune.py:992] (0/2) Epoch 10, batch 2200, loss[loss=0.1697, simple_loss=0.262, pruned_loss=0.03869, over 12272.00 frames. ], tot_loss[loss=0.1706, simple_loss=0.2597, pruned_loss=0.04071, over 2392023.18 frames. ], batch size: 37, lr: 4.07e-03, grad_scale: 8.0 2023-05-16 14:13:55,256 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=212314.0, num_to_drop=1, layers_to_drop={1} 2023-05-16 14:14:01,559 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.8834, 4.6318, 4.6824, 4.7432, 4.5389, 4.8357, 4.7322, 2.6705], device='cuda:0'), covar=tensor([0.0102, 0.0070, 0.0099, 0.0066, 0.0053, 0.0085, 0.0107, 0.0797], device='cuda:0'), in_proj_covar=tensor([0.0066, 0.0076, 0.0080, 0.0072, 0.0059, 0.0089, 0.0078, 0.0095], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-16 14:14:02,511 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.42 vs. limit=2.0 2023-05-16 14:14:24,150 INFO [finetune.py:992] (0/2) Epoch 10, batch 2250, loss[loss=0.1536, simple_loss=0.2414, pruned_loss=0.03292, over 12166.00 frames. ], tot_loss[loss=0.1698, simple_loss=0.2586, pruned_loss=0.04045, over 2396857.54 frames. ], batch size: 29, lr: 4.07e-03, grad_scale: 8.0 2023-05-16 14:14:26,367 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.151e+02 2.714e+02 3.113e+02 3.776e+02 8.100e+02, threshold=6.226e+02, percent-clipped=2.0 2023-05-16 14:15:00,475 INFO [finetune.py:992] (0/2) Epoch 10, batch 2300, loss[loss=0.1902, simple_loss=0.2788, pruned_loss=0.05082, over 12044.00 frames. ], tot_loss[loss=0.1698, simple_loss=0.2585, pruned_loss=0.04058, over 2389130.21 frames. ], batch size: 40, lr: 4.07e-03, grad_scale: 8.0 2023-05-16 14:15:24,605 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=212439.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 14:15:36,445 INFO [finetune.py:992] (0/2) Epoch 10, batch 2350, loss[loss=0.1788, simple_loss=0.2671, pruned_loss=0.04523, over 12133.00 frames. ], tot_loss[loss=0.1706, simple_loss=0.2596, pruned_loss=0.04079, over 2392908.09 frames. ], batch size: 30, lr: 4.07e-03, grad_scale: 8.0 2023-05-16 14:15:38,543 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.030e+02 2.649e+02 3.231e+02 3.820e+02 8.982e+02, threshold=6.461e+02, percent-clipped=3.0 2023-05-16 14:15:40,084 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=212460.0, num_to_drop=1, layers_to_drop={2} 2023-05-16 14:16:08,727 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=212500.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 14:16:11,978 INFO [finetune.py:992] (0/2) Epoch 10, batch 2400, loss[loss=0.1579, simple_loss=0.2427, pruned_loss=0.03652, over 12242.00 frames. ], tot_loss[loss=0.1693, simple_loss=0.2582, pruned_loss=0.04022, over 2397318.95 frames. ], batch size: 32, lr: 4.07e-03, grad_scale: 8.0 2023-05-16 14:16:24,652 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.98 vs. limit=2.0 2023-05-16 14:16:48,158 INFO [finetune.py:992] (0/2) Epoch 10, batch 2450, loss[loss=0.1695, simple_loss=0.2576, pruned_loss=0.04073, over 12033.00 frames. ], tot_loss[loss=0.1696, simple_loss=0.2586, pruned_loss=0.04032, over 2394087.00 frames. ], batch size: 31, lr: 4.07e-03, grad_scale: 8.0 2023-05-16 14:16:50,195 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.790e+02 2.756e+02 3.071e+02 3.589e+02 6.075e+02, threshold=6.142e+02, percent-clipped=0.0 2023-05-16 14:17:24,323 INFO [finetune.py:992] (0/2) Epoch 10, batch 2500, loss[loss=0.1706, simple_loss=0.2592, pruned_loss=0.04105, over 12146.00 frames. ], tot_loss[loss=0.1701, simple_loss=0.2592, pruned_loss=0.04053, over 2389308.47 frames. ], batch size: 34, lr: 4.07e-03, grad_scale: 8.0 2023-05-16 14:17:27,185 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=212609.0, num_to_drop=1, layers_to_drop={0} 2023-05-16 14:17:33,063 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.1387, 4.5533, 3.9989, 4.8203, 4.3377, 2.7607, 4.1952, 2.8892], device='cuda:0'), covar=tensor([0.0859, 0.0731, 0.1488, 0.0554, 0.1197, 0.1704, 0.0949, 0.3493], device='cuda:0'), in_proj_covar=tensor([0.0303, 0.0379, 0.0355, 0.0295, 0.0368, 0.0268, 0.0342, 0.0366], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 14:17:59,951 INFO [finetune.py:992] (0/2) Epoch 10, batch 2550, loss[loss=0.1711, simple_loss=0.2561, pruned_loss=0.04302, over 12093.00 frames. ], tot_loss[loss=0.1692, simple_loss=0.2581, pruned_loss=0.04018, over 2391846.10 frames. ], batch size: 32, lr: 4.07e-03, grad_scale: 8.0 2023-05-16 14:18:02,079 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.871e+02 2.656e+02 3.136e+02 3.615e+02 8.146e+02, threshold=6.272e+02, percent-clipped=2.0 2023-05-16 14:18:17,396 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=4.88 vs. limit=5.0 2023-05-16 14:18:19,233 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.59 vs. limit=2.0 2023-05-16 14:18:30,448 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=212698.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 14:18:35,894 INFO [finetune.py:992] (0/2) Epoch 10, batch 2600, loss[loss=0.1501, simple_loss=0.2347, pruned_loss=0.03271, over 12115.00 frames. ], tot_loss[loss=0.1685, simple_loss=0.2572, pruned_loss=0.03991, over 2389266.92 frames. ], batch size: 30, lr: 4.07e-03, grad_scale: 8.0 2023-05-16 14:18:40,301 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=212711.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 14:18:51,039 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.2904, 4.6250, 4.1397, 4.8862, 4.6790, 2.6540, 4.2672, 3.0111], device='cuda:0'), covar=tensor([0.0793, 0.0744, 0.1244, 0.0486, 0.0941, 0.1769, 0.0842, 0.3339], device='cuda:0'), in_proj_covar=tensor([0.0300, 0.0375, 0.0351, 0.0292, 0.0363, 0.0264, 0.0338, 0.0362], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 14:19:11,460 INFO [finetune.py:992] (0/2) Epoch 10, batch 2650, loss[loss=0.1465, simple_loss=0.2267, pruned_loss=0.03319, over 12178.00 frames. ], tot_loss[loss=0.1683, simple_loss=0.257, pruned_loss=0.0398, over 2391562.24 frames. ], batch size: 29, lr: 4.06e-03, grad_scale: 8.0 2023-05-16 14:19:14,296 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.848e+02 2.632e+02 3.258e+02 3.720e+02 9.493e+02, threshold=6.515e+02, percent-clipped=1.0 2023-05-16 14:19:15,175 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=212759.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 14:19:15,761 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=212760.0, num_to_drop=1, layers_to_drop={2} 2023-05-16 14:19:24,533 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=212772.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 14:19:25,884 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=212774.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 14:19:40,459 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=212795.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 14:19:47,871 INFO [finetune.py:992] (0/2) Epoch 10, batch 2700, loss[loss=0.1516, simple_loss=0.2476, pruned_loss=0.02782, over 12158.00 frames. ], tot_loss[loss=0.1685, simple_loss=0.2571, pruned_loss=0.03998, over 2388587.14 frames. ], batch size: 34, lr: 4.06e-03, grad_scale: 8.0 2023-05-16 14:19:50,000 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=212808.0, num_to_drop=1, layers_to_drop={0} 2023-05-16 14:20:09,296 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=212835.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 14:20:24,044 INFO [finetune.py:992] (0/2) Epoch 10, batch 2750, loss[loss=0.1797, simple_loss=0.2752, pruned_loss=0.04212, over 12041.00 frames. ], tot_loss[loss=0.1684, simple_loss=0.2567, pruned_loss=0.03999, over 2385152.73 frames. ], batch size: 42, lr: 4.06e-03, grad_scale: 8.0 2023-05-16 14:20:26,118 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.015e+02 2.715e+02 3.169e+02 3.755e+02 8.631e+02, threshold=6.338e+02, percent-clipped=1.0 2023-05-16 14:20:41,778 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.6229, 5.2609, 5.6029, 5.0377, 5.2445, 5.0840, 5.6108, 5.1626], device='cuda:0'), covar=tensor([0.0240, 0.0281, 0.0228, 0.0202, 0.0290, 0.0290, 0.0220, 0.0234], device='cuda:0'), in_proj_covar=tensor([0.0252, 0.0256, 0.0282, 0.0250, 0.0253, 0.0254, 0.0227, 0.0211], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 14:20:59,762 INFO [finetune.py:992] (0/2) Epoch 10, batch 2800, loss[loss=0.168, simple_loss=0.2568, pruned_loss=0.03957, over 12333.00 frames. ], tot_loss[loss=0.169, simple_loss=0.2573, pruned_loss=0.04035, over 2379505.77 frames. ], batch size: 36, lr: 4.06e-03, grad_scale: 8.0 2023-05-16 14:21:02,657 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=212909.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 14:21:35,067 INFO [finetune.py:992] (0/2) Epoch 10, batch 2850, loss[loss=0.1817, simple_loss=0.2676, pruned_loss=0.04794, over 12126.00 frames. ], tot_loss[loss=0.1686, simple_loss=0.2569, pruned_loss=0.04022, over 2388101.17 frames. ], batch size: 39, lr: 4.06e-03, grad_scale: 4.0 2023-05-16 14:21:36,568 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=212957.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 14:21:37,844 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.811e+02 2.812e+02 3.269e+02 3.790e+02 1.926e+03, threshold=6.539e+02, percent-clipped=3.0 2023-05-16 14:22:11,612 INFO [finetune.py:992] (0/2) Epoch 10, batch 2900, loss[loss=0.1823, simple_loss=0.2693, pruned_loss=0.04768, over 11152.00 frames. ], tot_loss[loss=0.1692, simple_loss=0.2576, pruned_loss=0.04037, over 2376411.27 frames. ], batch size: 55, lr: 4.06e-03, grad_scale: 4.0 2023-05-16 14:22:46,870 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=213054.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 14:22:47,481 INFO [finetune.py:992] (0/2) Epoch 10, batch 2950, loss[loss=0.1784, simple_loss=0.2666, pruned_loss=0.04511, over 11593.00 frames. ], tot_loss[loss=0.1694, simple_loss=0.2581, pruned_loss=0.04032, over 2378361.31 frames. ], batch size: 48, lr: 4.06e-03, grad_scale: 4.0 2023-05-16 14:22:50,251 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.007e+02 2.807e+02 3.371e+02 4.240e+02 8.214e+02, threshold=6.743e+02, percent-clipped=3.0 2023-05-16 14:22:55,977 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=213067.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 14:23:15,506 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=213095.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 14:23:22,360 INFO [finetune.py:992] (0/2) Epoch 10, batch 3000, loss[loss=0.1411, simple_loss=0.2319, pruned_loss=0.0251, over 12032.00 frames. ], tot_loss[loss=0.17, simple_loss=0.2588, pruned_loss=0.04064, over 2371556.87 frames. ], batch size: 31, lr: 4.06e-03, grad_scale: 4.0 2023-05-16 14:23:22,361 INFO [finetune.py:1017] (0/2) Computing validation loss 2023-05-16 14:23:36,521 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.5367, 2.6588, 4.1172, 4.2157, 2.8137, 2.6068, 2.7661, 2.0532], device='cuda:0'), covar=tensor([0.1637, 0.3056, 0.0559, 0.0497, 0.1324, 0.2281, 0.2809, 0.4594], device='cuda:0'), in_proj_covar=tensor([0.0287, 0.0382, 0.0271, 0.0294, 0.0265, 0.0299, 0.0372, 0.0372], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-16 14:23:37,626 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.9493, 2.6727, 3.3338, 2.0652, 2.5083, 2.8617, 2.7282, 3.0862], device='cuda:0'), covar=tensor([0.0493, 0.0962, 0.0390, 0.1490, 0.1618, 0.1421, 0.1024, 0.0875], device='cuda:0'), in_proj_covar=tensor([0.0237, 0.0236, 0.0247, 0.0182, 0.0239, 0.0292, 0.0225, 0.0261], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-16 14:23:40,833 INFO [finetune.py:1026] (0/2) Epoch 10, validation: loss=0.3191, simple_loss=0.3958, pruned_loss=0.1211, over 1020973.00 frames. 2023-05-16 14:23:40,833 INFO [finetune.py:1027] (0/2) Maximum memory allocated so far is 12827MB 2023-05-16 14:23:58,647 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=213130.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 14:24:07,762 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=213143.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 14:24:16,865 INFO [finetune.py:992] (0/2) Epoch 10, batch 3050, loss[loss=0.1406, simple_loss=0.2303, pruned_loss=0.0255, over 12090.00 frames. ], tot_loss[loss=0.1697, simple_loss=0.2585, pruned_loss=0.04047, over 2375399.37 frames. ], batch size: 32, lr: 4.06e-03, grad_scale: 4.0 2023-05-16 14:24:19,692 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.181e+02 2.841e+02 3.499e+02 4.340e+02 7.854e+02, threshold=6.998e+02, percent-clipped=1.0 2023-05-16 14:24:38,168 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.0091, 5.7813, 5.3752, 5.2320, 5.9323, 5.2197, 5.4024, 5.3157], device='cuda:0'), covar=tensor([0.1436, 0.1026, 0.1027, 0.2256, 0.0895, 0.2336, 0.1905, 0.1151], device='cuda:0'), in_proj_covar=tensor([0.0358, 0.0489, 0.0386, 0.0437, 0.0460, 0.0439, 0.0393, 0.0382], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 14:24:52,533 INFO [finetune.py:992] (0/2) Epoch 10, batch 3100, loss[loss=0.1937, simple_loss=0.2796, pruned_loss=0.05391, over 8029.00 frames. ], tot_loss[loss=0.1702, simple_loss=0.2591, pruned_loss=0.04063, over 2368627.33 frames. ], batch size: 98, lr: 4.06e-03, grad_scale: 4.0 2023-05-16 14:25:24,748 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.3790, 3.4976, 3.1688, 3.1275, 2.8621, 2.6874, 3.5199, 2.1949], device='cuda:0'), covar=tensor([0.0403, 0.0154, 0.0207, 0.0197, 0.0376, 0.0342, 0.0118, 0.0483], device='cuda:0'), in_proj_covar=tensor([0.0192, 0.0162, 0.0158, 0.0185, 0.0202, 0.0199, 0.0167, 0.0198], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 14:25:28,655 INFO [finetune.py:992] (0/2) Epoch 10, batch 3150, loss[loss=0.1555, simple_loss=0.2365, pruned_loss=0.03728, over 12341.00 frames. ], tot_loss[loss=0.1702, simple_loss=0.2587, pruned_loss=0.04082, over 2360507.21 frames. ], batch size: 35, lr: 4.06e-03, grad_scale: 4.0 2023-05-16 14:25:31,431 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.895e+02 2.677e+02 3.305e+02 3.988e+02 1.363e+03, threshold=6.611e+02, percent-clipped=2.0 2023-05-16 14:26:04,526 INFO [finetune.py:992] (0/2) Epoch 10, batch 3200, loss[loss=0.1598, simple_loss=0.2474, pruned_loss=0.03609, over 12173.00 frames. ], tot_loss[loss=0.1705, simple_loss=0.2591, pruned_loss=0.04095, over 2356232.93 frames. ], batch size: 31, lr: 4.06e-03, grad_scale: 8.0 2023-05-16 14:26:37,162 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.9255, 4.5545, 4.5505, 4.6625, 4.6197, 4.8197, 4.6881, 2.5770], device='cuda:0'), covar=tensor([0.0107, 0.0073, 0.0121, 0.0079, 0.0061, 0.0107, 0.0097, 0.0833], device='cuda:0'), in_proj_covar=tensor([0.0066, 0.0076, 0.0080, 0.0071, 0.0059, 0.0089, 0.0078, 0.0095], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-16 14:26:39,293 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=213354.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 14:26:39,876 INFO [finetune.py:992] (0/2) Epoch 10, batch 3250, loss[loss=0.1524, simple_loss=0.2456, pruned_loss=0.02956, over 12024.00 frames. ], tot_loss[loss=0.1703, simple_loss=0.2591, pruned_loss=0.04078, over 2369031.64 frames. ], batch size: 31, lr: 4.06e-03, grad_scale: 8.0 2023-05-16 14:26:42,680 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.116e+02 2.529e+02 3.027e+02 3.616e+02 6.084e+02, threshold=6.053e+02, percent-clipped=0.0 2023-05-16 14:26:48,383 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=213367.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 14:26:55,462 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.9450, 5.7941, 5.7769, 5.1077, 5.0580, 5.9442, 5.1603, 5.2245], device='cuda:0'), covar=tensor([0.1291, 0.2083, 0.1336, 0.2806, 0.1170, 0.1356, 0.3529, 0.2175], device='cuda:0'), in_proj_covar=tensor([0.0613, 0.0548, 0.0510, 0.0627, 0.0416, 0.0710, 0.0771, 0.0565], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0004, 0.0004, 0.0003], device='cuda:0') 2023-05-16 14:27:14,016 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=213402.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 14:27:16,037 INFO [finetune.py:992] (0/2) Epoch 10, batch 3300, loss[loss=0.1925, simple_loss=0.284, pruned_loss=0.05049, over 12081.00 frames. ], tot_loss[loss=0.1698, simple_loss=0.2585, pruned_loss=0.0406, over 2366977.77 frames. ], batch size: 42, lr: 4.06e-03, grad_scale: 8.0 2023-05-16 14:27:23,274 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=213415.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 14:27:25,548 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-05-16 14:27:33,948 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=213430.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 14:27:52,235 INFO [finetune.py:992] (0/2) Epoch 10, batch 3350, loss[loss=0.1849, simple_loss=0.2744, pruned_loss=0.04767, over 12145.00 frames. ], tot_loss[loss=0.1696, simple_loss=0.2581, pruned_loss=0.04054, over 2369729.68 frames. ], batch size: 36, lr: 4.06e-03, grad_scale: 8.0 2023-05-16 14:27:55,116 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.906e+02 2.748e+02 3.058e+02 3.642e+02 5.641e+02, threshold=6.115e+02, percent-clipped=0.0 2023-05-16 14:28:08,656 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=213478.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 14:28:27,429 INFO [finetune.py:992] (0/2) Epoch 10, batch 3400, loss[loss=0.1605, simple_loss=0.2498, pruned_loss=0.03559, over 12335.00 frames. ], tot_loss[loss=0.1707, simple_loss=0.2592, pruned_loss=0.04107, over 2367349.55 frames. ], batch size: 31, lr: 4.06e-03, grad_scale: 8.0 2023-05-16 14:28:40,873 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.39 vs. limit=2.0 2023-05-16 14:28:45,464 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.3325, 3.4645, 3.1383, 3.0949, 2.8184, 2.6071, 3.4586, 2.2154], device='cuda:0'), covar=tensor([0.0420, 0.0150, 0.0178, 0.0195, 0.0345, 0.0322, 0.0133, 0.0427], device='cuda:0'), in_proj_covar=tensor([0.0191, 0.0161, 0.0158, 0.0185, 0.0201, 0.0197, 0.0167, 0.0197], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 14:29:03,098 INFO [finetune.py:992] (0/2) Epoch 10, batch 3450, loss[loss=0.1574, simple_loss=0.2536, pruned_loss=0.03062, over 12098.00 frames. ], tot_loss[loss=0.1692, simple_loss=0.2577, pruned_loss=0.04039, over 2374504.67 frames. ], batch size: 32, lr: 4.06e-03, grad_scale: 8.0 2023-05-16 14:29:05,912 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.000e+02 2.712e+02 3.310e+02 3.896e+02 1.850e+03, threshold=6.620e+02, percent-clipped=6.0 2023-05-16 14:29:10,993 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=213566.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 14:29:16,243 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2023-05-16 14:29:26,635 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.3259, 2.4704, 3.0548, 4.2625, 2.1987, 4.3040, 4.2872, 4.4461], device='cuda:0'), covar=tensor([0.0117, 0.1204, 0.0551, 0.0140, 0.1291, 0.0196, 0.0143, 0.0069], device='cuda:0'), in_proj_covar=tensor([0.0122, 0.0203, 0.0185, 0.0116, 0.0189, 0.0176, 0.0174, 0.0117], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 14:29:36,223 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.91 vs. limit=2.0 2023-05-16 14:29:39,366 INFO [finetune.py:992] (0/2) Epoch 10, batch 3500, loss[loss=0.1805, simple_loss=0.2681, pruned_loss=0.04646, over 12057.00 frames. ], tot_loss[loss=0.1692, simple_loss=0.2574, pruned_loss=0.0405, over 2368916.45 frames. ], batch size: 45, lr: 4.06e-03, grad_scale: 8.0 2023-05-16 14:29:55,183 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=213627.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 14:30:07,848 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.5535, 5.2639, 5.3724, 5.4477, 5.0397, 5.1171, 4.9055, 5.3543], device='cuda:0'), covar=tensor([0.0591, 0.0668, 0.0915, 0.0632, 0.1992, 0.1369, 0.0594, 0.1182], device='cuda:0'), in_proj_covar=tensor([0.0529, 0.0694, 0.0604, 0.0614, 0.0834, 0.0739, 0.0546, 0.0474], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0005, 0.0004, 0.0004, 0.0005, 0.0005, 0.0004, 0.0003], device='cuda:0') 2023-05-16 14:30:14,510 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.69 vs. limit=2.0 2023-05-16 14:30:14,853 INFO [finetune.py:992] (0/2) Epoch 10, batch 3550, loss[loss=0.1479, simple_loss=0.2391, pruned_loss=0.02837, over 11992.00 frames. ], tot_loss[loss=0.1693, simple_loss=0.2574, pruned_loss=0.04063, over 2361215.70 frames. ], batch size: 28, lr: 4.06e-03, grad_scale: 8.0 2023-05-16 14:30:17,581 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.796e+02 2.610e+02 3.178e+02 3.849e+02 9.500e+02, threshold=6.356e+02, percent-clipped=2.0 2023-05-16 14:30:29,731 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.7831, 4.5045, 4.6329, 4.6228, 4.4943, 4.7349, 4.5705, 2.6253], device='cuda:0'), covar=tensor([0.0236, 0.0092, 0.0126, 0.0089, 0.0066, 0.0128, 0.0136, 0.0975], device='cuda:0'), in_proj_covar=tensor([0.0067, 0.0077, 0.0081, 0.0072, 0.0060, 0.0090, 0.0079, 0.0097], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-16 14:30:51,047 INFO [finetune.py:992] (0/2) Epoch 10, batch 3600, loss[loss=0.1699, simple_loss=0.2616, pruned_loss=0.0391, over 12289.00 frames. ], tot_loss[loss=0.1693, simple_loss=0.2573, pruned_loss=0.04062, over 2360127.08 frames. ], batch size: 33, lr: 4.06e-03, grad_scale: 8.0 2023-05-16 14:31:04,763 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-05-16 14:31:26,949 INFO [finetune.py:992] (0/2) Epoch 10, batch 3650, loss[loss=0.1777, simple_loss=0.2721, pruned_loss=0.04172, over 12188.00 frames. ], tot_loss[loss=0.1697, simple_loss=0.258, pruned_loss=0.04069, over 2364415.42 frames. ], batch size: 35, lr: 4.05e-03, grad_scale: 8.0 2023-05-16 14:31:29,779 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.922e+02 2.658e+02 3.103e+02 3.735e+02 8.103e+02, threshold=6.207e+02, percent-clipped=1.0 2023-05-16 14:31:55,349 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.4794, 4.5693, 4.0837, 5.0595, 4.5130, 3.1528, 4.2291, 2.9390], device='cuda:0'), covar=tensor([0.0734, 0.1046, 0.1501, 0.0571, 0.1138, 0.1435, 0.1083, 0.3457], device='cuda:0'), in_proj_covar=tensor([0.0301, 0.0378, 0.0353, 0.0294, 0.0365, 0.0267, 0.0340, 0.0365], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 14:32:02,502 INFO [finetune.py:992] (0/2) Epoch 10, batch 3700, loss[loss=0.1744, simple_loss=0.2676, pruned_loss=0.04059, over 11294.00 frames. ], tot_loss[loss=0.1694, simple_loss=0.2575, pruned_loss=0.04069, over 2368445.33 frames. ], batch size: 55, lr: 4.05e-03, grad_scale: 8.0 2023-05-16 14:32:08,068 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.0470, 2.8892, 2.7224, 2.7169, 2.5024, 2.3754, 2.8185, 1.8655], device='cuda:0'), covar=tensor([0.0447, 0.0185, 0.0226, 0.0217, 0.0396, 0.0320, 0.0154, 0.0495], device='cuda:0'), in_proj_covar=tensor([0.0190, 0.0160, 0.0157, 0.0183, 0.0201, 0.0197, 0.0167, 0.0197], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 14:32:23,188 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.9916, 2.4903, 3.6285, 2.9546, 3.3883, 3.1828, 2.3779, 3.5039], device='cuda:0'), covar=tensor([0.0141, 0.0352, 0.0140, 0.0251, 0.0161, 0.0177, 0.0357, 0.0132], device='cuda:0'), in_proj_covar=tensor([0.0172, 0.0202, 0.0184, 0.0182, 0.0212, 0.0158, 0.0194, 0.0179], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 14:32:38,532 INFO [finetune.py:992] (0/2) Epoch 10, batch 3750, loss[loss=0.1583, simple_loss=0.249, pruned_loss=0.03376, over 12348.00 frames. ], tot_loss[loss=0.1698, simple_loss=0.2582, pruned_loss=0.04071, over 2363020.39 frames. ], batch size: 31, lr: 4.05e-03, grad_scale: 8.0 2023-05-16 14:32:40,490 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=5.03 vs. limit=5.0 2023-05-16 14:32:41,258 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.981e+02 2.726e+02 3.266e+02 3.847e+02 1.047e+03, threshold=6.533e+02, percent-clipped=2.0 2023-05-16 14:32:41,616 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=2.21 vs. limit=5.0 2023-05-16 14:32:50,772 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.3644, 4.8970, 5.3482, 4.6797, 4.9544, 4.7504, 5.3631, 4.9488], device='cuda:0'), covar=tensor([0.0240, 0.0322, 0.0225, 0.0253, 0.0364, 0.0283, 0.0209, 0.0292], device='cuda:0'), in_proj_covar=tensor([0.0254, 0.0259, 0.0283, 0.0254, 0.0256, 0.0254, 0.0230, 0.0213], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 14:33:03,010 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.5682, 5.3485, 5.4486, 5.5155, 5.1090, 5.1856, 4.9652, 5.4413], device='cuda:0'), covar=tensor([0.0590, 0.0577, 0.0668, 0.0561, 0.1814, 0.1163, 0.0505, 0.1077], device='cuda:0'), in_proj_covar=tensor([0.0529, 0.0696, 0.0604, 0.0616, 0.0836, 0.0741, 0.0547, 0.0476], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0005, 0.0004, 0.0004, 0.0005, 0.0005, 0.0004, 0.0003], device='cuda:0') 2023-05-16 14:33:04,579 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.2417, 3.9473, 4.0125, 4.4067, 2.8920, 3.9577, 2.5822, 4.1131], device='cuda:0'), covar=tensor([0.1641, 0.0742, 0.0931, 0.0774, 0.1139, 0.0636, 0.1849, 0.1424], device='cuda:0'), in_proj_covar=tensor([0.0225, 0.0265, 0.0292, 0.0355, 0.0234, 0.0240, 0.0258, 0.0376], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 14:33:08,286 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.15 vs. limit=2.0 2023-05-16 14:33:14,906 INFO [finetune.py:992] (0/2) Epoch 10, batch 3800, loss[loss=0.1803, simple_loss=0.272, pruned_loss=0.04427, over 11353.00 frames. ], tot_loss[loss=0.1699, simple_loss=0.2588, pruned_loss=0.04044, over 2373542.67 frames. ], batch size: 55, lr: 4.05e-03, grad_scale: 8.0 2023-05-16 14:33:27,163 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=213922.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 14:33:50,777 INFO [finetune.py:992] (0/2) Epoch 10, batch 3850, loss[loss=0.1835, simple_loss=0.2685, pruned_loss=0.04928, over 12177.00 frames. ], tot_loss[loss=0.1688, simple_loss=0.2576, pruned_loss=0.03998, over 2371729.65 frames. ], batch size: 31, lr: 4.05e-03, grad_scale: 8.0 2023-05-16 14:33:53,666 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.986e+02 2.760e+02 3.187e+02 3.831e+02 8.275e+02, threshold=6.375e+02, percent-clipped=2.0 2023-05-16 14:34:21,940 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.7583, 2.8648, 4.6997, 4.9250, 3.0125, 2.7225, 2.9883, 2.2463], device='cuda:0'), covar=tensor([0.1425, 0.2814, 0.0406, 0.0312, 0.1168, 0.2136, 0.2588, 0.3848], device='cuda:0'), in_proj_covar=tensor([0.0286, 0.0381, 0.0269, 0.0291, 0.0264, 0.0298, 0.0370, 0.0369], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-16 14:34:24,041 INFO [checkpoint.py:75] (0/2) Saving checkpoint to pruned_transducer_stateless7/exp_giga_finetune/checkpoint-114000.pt 2023-05-16 14:34:30,252 INFO [finetune.py:992] (0/2) Epoch 10, batch 3900, loss[loss=0.1664, simple_loss=0.2411, pruned_loss=0.04584, over 11765.00 frames. ], tot_loss[loss=0.1671, simple_loss=0.2556, pruned_loss=0.03933, over 2379237.61 frames. ], batch size: 26, lr: 4.05e-03, grad_scale: 8.0 2023-05-16 14:35:06,396 INFO [finetune.py:992] (0/2) Epoch 10, batch 3950, loss[loss=0.1766, simple_loss=0.2508, pruned_loss=0.05121, over 12115.00 frames. ], tot_loss[loss=0.1678, simple_loss=0.2566, pruned_loss=0.03953, over 2383243.19 frames. ], batch size: 30, lr: 4.05e-03, grad_scale: 8.0 2023-05-16 14:35:09,281 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.776e+02 2.613e+02 3.033e+02 3.734e+02 5.708e+02, threshold=6.065e+02, percent-clipped=0.0 2023-05-16 14:35:25,547 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=4.75 vs. limit=5.0 2023-05-16 14:35:42,209 INFO [finetune.py:992] (0/2) Epoch 10, batch 4000, loss[loss=0.1478, simple_loss=0.2329, pruned_loss=0.03136, over 11999.00 frames. ], tot_loss[loss=0.1677, simple_loss=0.2563, pruned_loss=0.03952, over 2375015.16 frames. ], batch size: 28, lr: 4.05e-03, grad_scale: 8.0 2023-05-16 14:35:59,894 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=214129.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 14:36:03,562 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.39 vs. limit=2.0 2023-05-16 14:36:18,110 INFO [finetune.py:992] (0/2) Epoch 10, batch 4050, loss[loss=0.1407, simple_loss=0.2227, pruned_loss=0.02936, over 12027.00 frames. ], tot_loss[loss=0.1691, simple_loss=0.2581, pruned_loss=0.04002, over 2376671.42 frames. ], batch size: 28, lr: 4.05e-03, grad_scale: 8.0 2023-05-16 14:36:20,873 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.889e+02 2.768e+02 3.378e+02 3.956e+02 1.343e+03, threshold=6.756e+02, percent-clipped=3.0 2023-05-16 14:36:43,639 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=214190.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 14:36:54,746 INFO [finetune.py:992] (0/2) Epoch 10, batch 4100, loss[loss=0.1735, simple_loss=0.2634, pruned_loss=0.04182, over 12041.00 frames. ], tot_loss[loss=0.1679, simple_loss=0.2568, pruned_loss=0.03951, over 2384129.60 frames. ], batch size: 37, lr: 4.05e-03, grad_scale: 8.0 2023-05-16 14:37:04,158 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=214218.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 14:37:05,921 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=4.48 vs. limit=5.0 2023-05-16 14:37:07,061 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=214222.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 14:37:18,772 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.39 vs. limit=2.0 2023-05-16 14:37:30,852 INFO [finetune.py:992] (0/2) Epoch 10, batch 4150, loss[loss=0.1767, simple_loss=0.26, pruned_loss=0.04669, over 12410.00 frames. ], tot_loss[loss=0.1678, simple_loss=0.2565, pruned_loss=0.03956, over 2383152.30 frames. ], batch size: 32, lr: 4.05e-03, grad_scale: 8.0 2023-05-16 14:37:33,640 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.043e+02 2.613e+02 3.086e+02 3.878e+02 6.591e+02, threshold=6.171e+02, percent-clipped=0.0 2023-05-16 14:37:41,699 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=214270.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 14:37:48,354 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=214279.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 14:38:02,976 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.5034, 5.1902, 5.4694, 5.4256, 4.6839, 4.7239, 4.9025, 5.1596], device='cuda:0'), covar=tensor([0.0974, 0.1191, 0.0799, 0.0818, 0.3268, 0.2069, 0.0741, 0.1947], device='cuda:0'), in_proj_covar=tensor([0.0525, 0.0691, 0.0600, 0.0617, 0.0830, 0.0735, 0.0542, 0.0472], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0005, 0.0004, 0.0004, 0.0005, 0.0005, 0.0003, 0.0003], device='cuda:0') 2023-05-16 14:38:06,250 INFO [finetune.py:992] (0/2) Epoch 10, batch 4200, loss[loss=0.1364, simple_loss=0.2232, pruned_loss=0.02477, over 12346.00 frames. ], tot_loss[loss=0.168, simple_loss=0.2569, pruned_loss=0.03951, over 2385284.58 frames. ], batch size: 31, lr: 4.05e-03, grad_scale: 8.0 2023-05-16 14:38:07,836 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.3791, 5.1486, 5.2411, 5.3058, 4.9482, 4.9178, 4.7276, 5.1758], device='cuda:0'), covar=tensor([0.0573, 0.0596, 0.0872, 0.0523, 0.1816, 0.1265, 0.0554, 0.1236], device='cuda:0'), in_proj_covar=tensor([0.0525, 0.0691, 0.0600, 0.0617, 0.0831, 0.0736, 0.0542, 0.0473], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0005, 0.0004, 0.0004, 0.0005, 0.0005, 0.0003, 0.0003], device='cuda:0') 2023-05-16 14:38:42,423 INFO [finetune.py:992] (0/2) Epoch 10, batch 4250, loss[loss=0.1761, simple_loss=0.2601, pruned_loss=0.0461, over 12274.00 frames. ], tot_loss[loss=0.168, simple_loss=0.2569, pruned_loss=0.0395, over 2390850.64 frames. ], batch size: 37, lr: 4.05e-03, grad_scale: 8.0 2023-05-16 14:38:45,168 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.964e+02 2.824e+02 3.268e+02 4.058e+02 5.640e+02, threshold=6.536e+02, percent-clipped=0.0 2023-05-16 14:38:50,358 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.9175, 4.5993, 4.7483, 4.7432, 4.6156, 4.8128, 4.7306, 2.8226], device='cuda:0'), covar=tensor([0.0106, 0.0066, 0.0080, 0.0067, 0.0052, 0.0100, 0.0079, 0.0716], device='cuda:0'), in_proj_covar=tensor([0.0068, 0.0078, 0.0081, 0.0073, 0.0060, 0.0091, 0.0081, 0.0097], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 14:39:06,151 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2023-05-16 14:39:15,449 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([6.1043, 6.0307, 5.8605, 5.3175, 5.2906, 5.9841, 5.5645, 5.3630], device='cuda:0'), covar=tensor([0.0743, 0.1004, 0.0765, 0.1680, 0.0676, 0.0867, 0.1822, 0.1074], device='cuda:0'), in_proj_covar=tensor([0.0621, 0.0553, 0.0515, 0.0636, 0.0420, 0.0717, 0.0786, 0.0572], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0004, 0.0004, 0.0003], device='cuda:0') 2023-05-16 14:39:18,809 INFO [finetune.py:992] (0/2) Epoch 10, batch 4300, loss[loss=0.1557, simple_loss=0.2362, pruned_loss=0.03766, over 12003.00 frames. ], tot_loss[loss=0.168, simple_loss=0.2567, pruned_loss=0.03965, over 2387414.37 frames. ], batch size: 28, lr: 4.05e-03, grad_scale: 8.0 2023-05-16 14:39:19,200 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-05-16 14:39:52,866 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.8988, 2.3320, 3.5861, 3.0484, 3.4629, 3.1035, 2.3627, 3.5093], device='cuda:0'), covar=tensor([0.0152, 0.0381, 0.0137, 0.0221, 0.0126, 0.0205, 0.0379, 0.0117], device='cuda:0'), in_proj_covar=tensor([0.0173, 0.0202, 0.0187, 0.0183, 0.0212, 0.0159, 0.0196, 0.0180], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 14:39:53,921 INFO [finetune.py:992] (0/2) Epoch 10, batch 4350, loss[loss=0.1742, simple_loss=0.2704, pruned_loss=0.03895, over 12165.00 frames. ], tot_loss[loss=0.1693, simple_loss=0.2581, pruned_loss=0.04022, over 2382180.93 frames. ], batch size: 34, lr: 4.05e-03, grad_scale: 8.0 2023-05-16 14:39:56,689 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.895e+02 2.611e+02 2.964e+02 3.774e+02 9.007e+02, threshold=5.928e+02, percent-clipped=5.0 2023-05-16 14:40:16,121 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=214485.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 14:40:30,146 INFO [finetune.py:992] (0/2) Epoch 10, batch 4400, loss[loss=0.144, simple_loss=0.2245, pruned_loss=0.03177, over 12005.00 frames. ], tot_loss[loss=0.1691, simple_loss=0.2579, pruned_loss=0.04013, over 2385230.24 frames. ], batch size: 28, lr: 4.05e-03, grad_scale: 8.0 2023-05-16 14:40:39,649 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.3967, 5.2671, 5.2068, 4.5909, 4.8446, 5.3846, 4.5868, 4.8069], device='cuda:0'), covar=tensor([0.1702, 0.2079, 0.1453, 0.2877, 0.1437, 0.1605, 0.3959, 0.2214], device='cuda:0'), in_proj_covar=tensor([0.0617, 0.0549, 0.0513, 0.0632, 0.0417, 0.0714, 0.0780, 0.0567], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0004, 0.0004, 0.0003], device='cuda:0') 2023-05-16 14:40:43,398 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.6880, 2.7821, 4.3661, 4.5998, 2.7830, 2.5921, 2.7561, 2.1382], device='cuda:0'), covar=tensor([0.1518, 0.2982, 0.0519, 0.0403, 0.1251, 0.2285, 0.2894, 0.3994], device='cuda:0'), in_proj_covar=tensor([0.0286, 0.0380, 0.0270, 0.0293, 0.0265, 0.0299, 0.0372, 0.0369], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-16 14:40:51,348 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=4.72 vs. limit=5.0 2023-05-16 14:41:06,375 INFO [finetune.py:992] (0/2) Epoch 10, batch 4450, loss[loss=0.1621, simple_loss=0.2616, pruned_loss=0.03132, over 12043.00 frames. ], tot_loss[loss=0.1689, simple_loss=0.2577, pruned_loss=0.04003, over 2381214.68 frames. ], batch size: 40, lr: 4.05e-03, grad_scale: 8.0 2023-05-16 14:41:06,592 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.4475, 2.4670, 3.1086, 4.4002, 2.1411, 4.4165, 4.4810, 4.6381], device='cuda:0'), covar=tensor([0.0147, 0.1273, 0.0572, 0.0127, 0.1477, 0.0226, 0.0151, 0.0073], device='cuda:0'), in_proj_covar=tensor([0.0122, 0.0201, 0.0185, 0.0115, 0.0188, 0.0177, 0.0174, 0.0118], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 14:41:09,101 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.830e+02 2.774e+02 3.267e+02 3.826e+02 7.033e+02, threshold=6.534e+02, percent-clipped=2.0 2023-05-16 14:41:15,864 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.1946, 4.8832, 5.0910, 5.0481, 4.8334, 5.1510, 5.0037, 2.7782], device='cuda:0'), covar=tensor([0.0115, 0.0063, 0.0064, 0.0054, 0.0052, 0.0093, 0.0077, 0.0653], device='cuda:0'), in_proj_covar=tensor([0.0068, 0.0078, 0.0082, 0.0074, 0.0061, 0.0092, 0.0081, 0.0098], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 14:41:20,130 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=214574.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 14:41:41,939 INFO [finetune.py:992] (0/2) Epoch 10, batch 4500, loss[loss=0.2313, simple_loss=0.2999, pruned_loss=0.08129, over 7793.00 frames. ], tot_loss[loss=0.1691, simple_loss=0.2578, pruned_loss=0.04024, over 2377294.44 frames. ], batch size: 98, lr: 4.05e-03, grad_scale: 8.0 2023-05-16 14:41:58,149 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.93 vs. limit=2.0 2023-05-16 14:42:18,442 INFO [finetune.py:992] (0/2) Epoch 10, batch 4550, loss[loss=0.2509, simple_loss=0.3229, pruned_loss=0.08947, over 7701.00 frames. ], tot_loss[loss=0.1689, simple_loss=0.2575, pruned_loss=0.0401, over 2373230.39 frames. ], batch size: 97, lr: 4.05e-03, grad_scale: 8.0 2023-05-16 14:42:21,187 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.995e+02 2.657e+02 3.127e+02 3.785e+02 6.147e+02, threshold=6.255e+02, percent-clipped=0.0 2023-05-16 14:42:51,806 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 2023-05-16 14:42:54,232 INFO [finetune.py:992] (0/2) Epoch 10, batch 4600, loss[loss=0.1668, simple_loss=0.2594, pruned_loss=0.03709, over 12117.00 frames. ], tot_loss[loss=0.1692, simple_loss=0.2584, pruned_loss=0.03999, over 2373902.44 frames. ], batch size: 33, lr: 4.05e-03, grad_scale: 8.0 2023-05-16 14:43:10,320 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=3.42 vs. limit=5.0 2023-05-16 14:43:29,586 INFO [finetune.py:992] (0/2) Epoch 10, batch 4650, loss[loss=0.1484, simple_loss=0.2381, pruned_loss=0.02934, over 12409.00 frames. ], tot_loss[loss=0.1699, simple_loss=0.259, pruned_loss=0.04037, over 2362329.84 frames. ], batch size: 32, lr: 4.04e-03, grad_scale: 8.0 2023-05-16 14:43:32,371 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.902e+02 2.776e+02 3.306e+02 3.949e+02 5.894e+02, threshold=6.612e+02, percent-clipped=0.0 2023-05-16 14:43:34,763 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.4729, 4.8675, 3.0992, 2.7285, 4.1590, 2.7075, 4.2217, 3.4905], device='cuda:0'), covar=tensor([0.0696, 0.0394, 0.1081, 0.1542, 0.0337, 0.1312, 0.0417, 0.0710], device='cuda:0'), in_proj_covar=tensor([0.0189, 0.0255, 0.0178, 0.0197, 0.0141, 0.0180, 0.0196, 0.0175], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 14:43:40,362 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.5085, 3.7400, 3.3152, 3.0617, 2.9168, 2.7157, 3.6714, 2.2911], device='cuda:0'), covar=tensor([0.0408, 0.0102, 0.0187, 0.0231, 0.0367, 0.0392, 0.0115, 0.0492], device='cuda:0'), in_proj_covar=tensor([0.0190, 0.0161, 0.0158, 0.0183, 0.0200, 0.0196, 0.0167, 0.0197], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 14:43:51,466 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=214785.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 14:44:05,841 INFO [finetune.py:992] (0/2) Epoch 10, batch 4700, loss[loss=0.1957, simple_loss=0.2933, pruned_loss=0.04907, over 12026.00 frames. ], tot_loss[loss=0.1705, simple_loss=0.2594, pruned_loss=0.04085, over 2366240.39 frames. ], batch size: 42, lr: 4.04e-03, grad_scale: 8.0 2023-05-16 14:44:19,744 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.8721, 4.0330, 3.5957, 4.2109, 3.8425, 2.6569, 3.7545, 2.8837], device='cuda:0'), covar=tensor([0.0733, 0.0796, 0.1386, 0.0563, 0.1180, 0.1634, 0.1024, 0.3062], device='cuda:0'), in_proj_covar=tensor([0.0300, 0.0376, 0.0352, 0.0292, 0.0361, 0.0265, 0.0339, 0.0360], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 14:44:26,363 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=214833.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 14:44:42,008 INFO [finetune.py:992] (0/2) Epoch 10, batch 4750, loss[loss=0.1676, simple_loss=0.2583, pruned_loss=0.03844, over 12311.00 frames. ], tot_loss[loss=0.1704, simple_loss=0.259, pruned_loss=0.04095, over 2358392.86 frames. ], batch size: 34, lr: 4.04e-03, grad_scale: 8.0 2023-05-16 14:44:44,881 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.848e+02 2.892e+02 3.369e+02 3.981e+02 7.528e+02, threshold=6.739e+02, percent-clipped=2.0 2023-05-16 14:44:55,962 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=214874.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 14:45:18,172 INFO [finetune.py:992] (0/2) Epoch 10, batch 4800, loss[loss=0.1574, simple_loss=0.254, pruned_loss=0.03034, over 12097.00 frames. ], tot_loss[loss=0.171, simple_loss=0.2596, pruned_loss=0.04126, over 2354813.53 frames. ], batch size: 33, lr: 4.04e-03, grad_scale: 8.0 2023-05-16 14:45:30,405 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=214922.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 14:45:51,562 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 2023-05-16 14:45:53,951 INFO [finetune.py:992] (0/2) Epoch 10, batch 4850, loss[loss=0.1737, simple_loss=0.2652, pruned_loss=0.0411, over 11910.00 frames. ], tot_loss[loss=0.1711, simple_loss=0.2595, pruned_loss=0.04131, over 2357423.64 frames. ], batch size: 44, lr: 4.04e-03, grad_scale: 16.0 2023-05-16 14:45:56,631 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.810e+02 2.716e+02 3.299e+02 3.755e+02 7.934e+02, threshold=6.598e+02, percent-clipped=1.0 2023-05-16 14:46:08,961 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.8683, 3.5063, 3.7299, 3.8029, 3.7461, 3.8408, 3.6527, 2.5643], device='cuda:0'), covar=tensor([0.0098, 0.0115, 0.0125, 0.0082, 0.0066, 0.0120, 0.0116, 0.0764], device='cuda:0'), in_proj_covar=tensor([0.0067, 0.0078, 0.0081, 0.0073, 0.0060, 0.0091, 0.0080, 0.0097], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-16 14:46:23,309 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.1708, 3.8084, 4.0507, 4.2922, 3.0612, 3.8254, 2.5960, 3.9884], device='cuda:0'), covar=tensor([0.1777, 0.0877, 0.0972, 0.0696, 0.1146, 0.0747, 0.1979, 0.1566], device='cuda:0'), in_proj_covar=tensor([0.0225, 0.0264, 0.0295, 0.0353, 0.0234, 0.0239, 0.0259, 0.0374], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 14:46:30,481 INFO [finetune.py:992] (0/2) Epoch 10, batch 4900, loss[loss=0.1744, simple_loss=0.2716, pruned_loss=0.03862, over 12099.00 frames. ], tot_loss[loss=0.171, simple_loss=0.2598, pruned_loss=0.04105, over 2366830.06 frames. ], batch size: 38, lr: 4.04e-03, grad_scale: 16.0 2023-05-16 14:46:38,357 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.6838, 3.7561, 3.3212, 3.1772, 3.0089, 2.8932, 3.6770, 2.3598], device='cuda:0'), covar=tensor([0.0342, 0.0115, 0.0202, 0.0197, 0.0351, 0.0307, 0.0137, 0.0422], device='cuda:0'), in_proj_covar=tensor([0.0191, 0.0161, 0.0158, 0.0183, 0.0202, 0.0197, 0.0169, 0.0197], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 14:47:06,474 INFO [finetune.py:992] (0/2) Epoch 10, batch 4950, loss[loss=0.1694, simple_loss=0.2536, pruned_loss=0.04259, over 12188.00 frames. ], tot_loss[loss=0.1705, simple_loss=0.2591, pruned_loss=0.04095, over 2358185.05 frames. ], batch size: 31, lr: 4.04e-03, grad_scale: 8.0 2023-05-16 14:47:09,868 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.873e+02 2.771e+02 3.187e+02 3.647e+02 6.173e+02, threshold=6.375e+02, percent-clipped=0.0 2023-05-16 14:47:41,988 INFO [finetune.py:992] (0/2) Epoch 10, batch 5000, loss[loss=0.1513, simple_loss=0.2276, pruned_loss=0.03752, over 12296.00 frames. ], tot_loss[loss=0.1703, simple_loss=0.2587, pruned_loss=0.04095, over 2357516.11 frames. ], batch size: 28, lr: 4.04e-03, grad_scale: 8.0 2023-05-16 14:48:17,131 INFO [finetune.py:992] (0/2) Epoch 10, batch 5050, loss[loss=0.1703, simple_loss=0.2615, pruned_loss=0.03953, over 12296.00 frames. ], tot_loss[loss=0.1701, simple_loss=0.2586, pruned_loss=0.04074, over 2361376.18 frames. ], batch size: 34, lr: 4.04e-03, grad_scale: 8.0 2023-05-16 14:48:20,564 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.997e+02 2.832e+02 3.427e+02 4.150e+02 6.907e+02, threshold=6.854e+02, percent-clipped=3.0 2023-05-16 14:48:53,160 INFO [finetune.py:992] (0/2) Epoch 10, batch 5100, loss[loss=0.1571, simple_loss=0.2501, pruned_loss=0.03205, over 12044.00 frames. ], tot_loss[loss=0.1697, simple_loss=0.2581, pruned_loss=0.04063, over 2370497.12 frames. ], batch size: 40, lr: 4.04e-03, grad_scale: 8.0 2023-05-16 14:49:11,561 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=215230.0, num_to_drop=1, layers_to_drop={1} 2023-05-16 14:49:29,925 INFO [finetune.py:992] (0/2) Epoch 10, batch 5150, loss[loss=0.1736, simple_loss=0.2691, pruned_loss=0.03902, over 12110.00 frames. ], tot_loss[loss=0.1696, simple_loss=0.2581, pruned_loss=0.04053, over 2365819.29 frames. ], batch size: 33, lr: 4.04e-03, grad_scale: 8.0 2023-05-16 14:49:33,488 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.088e+02 2.718e+02 3.092e+02 3.626e+02 6.998e+02, threshold=6.184e+02, percent-clipped=1.0 2023-05-16 14:49:55,733 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=215291.0, num_to_drop=1, layers_to_drop={0} 2023-05-16 14:50:05,032 INFO [finetune.py:992] (0/2) Epoch 10, batch 5200, loss[loss=0.1808, simple_loss=0.2726, pruned_loss=0.04446, over 12197.00 frames. ], tot_loss[loss=0.1694, simple_loss=0.2578, pruned_loss=0.04045, over 2370127.16 frames. ], batch size: 35, lr: 4.04e-03, grad_scale: 8.0 2023-05-16 14:50:19,353 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=215325.0, num_to_drop=1, layers_to_drop={1} 2023-05-16 14:50:24,962 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=215333.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 14:50:41,085 INFO [finetune.py:992] (0/2) Epoch 10, batch 5250, loss[loss=0.162, simple_loss=0.2466, pruned_loss=0.0387, over 12345.00 frames. ], tot_loss[loss=0.1691, simple_loss=0.2577, pruned_loss=0.04028, over 2377304.75 frames. ], batch size: 31, lr: 4.04e-03, grad_scale: 8.0 2023-05-16 14:50:44,487 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.857e+02 2.743e+02 3.102e+02 3.650e+02 7.283e+02, threshold=6.204e+02, percent-clipped=2.0 2023-05-16 14:50:56,183 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.87 vs. limit=2.0 2023-05-16 14:51:03,120 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=215386.0, num_to_drop=1, layers_to_drop={3} 2023-05-16 14:51:09,343 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=215394.0, num_to_drop=1, layers_to_drop={1} 2023-05-16 14:51:17,317 INFO [finetune.py:992] (0/2) Epoch 10, batch 5300, loss[loss=0.1919, simple_loss=0.2823, pruned_loss=0.05078, over 12109.00 frames. ], tot_loss[loss=0.1697, simple_loss=0.2582, pruned_loss=0.0406, over 2381361.16 frames. ], batch size: 39, lr: 4.04e-03, grad_scale: 8.0 2023-05-16 14:51:43,774 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.0039, 5.8079, 5.4076, 5.2746, 5.9460, 5.1644, 5.3916, 5.3586], device='cuda:0'), covar=tensor([0.1277, 0.0755, 0.0880, 0.1686, 0.0761, 0.1856, 0.1507, 0.1020], device='cuda:0'), in_proj_covar=tensor([0.0352, 0.0484, 0.0386, 0.0434, 0.0455, 0.0437, 0.0391, 0.0375], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 14:51:52,987 INFO [finetune.py:992] (0/2) Epoch 10, batch 5350, loss[loss=0.1472, simple_loss=0.2245, pruned_loss=0.03497, over 12113.00 frames. ], tot_loss[loss=0.1688, simple_loss=0.2572, pruned_loss=0.0402, over 2377778.07 frames. ], batch size: 30, lr: 4.04e-03, grad_scale: 8.0 2023-05-16 14:51:56,653 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.043e+02 2.929e+02 3.222e+02 3.897e+02 6.738e+02, threshold=6.444e+02, percent-clipped=2.0 2023-05-16 14:52:28,084 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=215503.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 14:52:29,290 INFO [finetune.py:992] (0/2) Epoch 10, batch 5400, loss[loss=0.1543, simple_loss=0.2345, pruned_loss=0.037, over 11859.00 frames. ], tot_loss[loss=0.1686, simple_loss=0.257, pruned_loss=0.04012, over 2373533.95 frames. ], batch size: 26, lr: 4.04e-03, grad_scale: 8.0 2023-05-16 14:53:00,869 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=215548.0, num_to_drop=1, layers_to_drop={1} 2023-05-16 14:53:05,543 INFO [finetune.py:992] (0/2) Epoch 10, batch 5450, loss[loss=0.1686, simple_loss=0.2551, pruned_loss=0.04101, over 12308.00 frames. ], tot_loss[loss=0.1691, simple_loss=0.2577, pruned_loss=0.0403, over 2367518.78 frames. ], batch size: 34, lr: 4.04e-03, grad_scale: 8.0 2023-05-16 14:53:09,062 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.038e+02 2.735e+02 3.191e+02 3.702e+02 7.831e+02, threshold=6.383e+02, percent-clipped=1.0 2023-05-16 14:53:12,037 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=215564.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 14:53:12,780 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=215565.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 14:53:27,736 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=215586.0, num_to_drop=1, layers_to_drop={3} 2023-05-16 14:53:41,303 INFO [finetune.py:992] (0/2) Epoch 10, batch 5500, loss[loss=0.1764, simple_loss=0.2732, pruned_loss=0.03981, over 12311.00 frames. ], tot_loss[loss=0.1699, simple_loss=0.2587, pruned_loss=0.04054, over 2362464.95 frames. ], batch size: 34, lr: 4.04e-03, grad_scale: 8.0 2023-05-16 14:53:44,317 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=215609.0, num_to_drop=1, layers_to_drop={2} 2023-05-16 14:53:56,988 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=215626.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 14:54:17,697 INFO [finetune.py:992] (0/2) Epoch 10, batch 5550, loss[loss=0.1797, simple_loss=0.28, pruned_loss=0.0397, over 12187.00 frames. ], tot_loss[loss=0.1698, simple_loss=0.2586, pruned_loss=0.04045, over 2360270.75 frames. ], batch size: 35, lr: 4.04e-03, grad_scale: 8.0 2023-05-16 14:54:21,194 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.854e+02 2.670e+02 3.182e+02 3.606e+02 3.832e+03, threshold=6.365e+02, percent-clipped=3.0 2023-05-16 14:54:26,534 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.3973, 4.6613, 4.1353, 4.9508, 4.5656, 2.9834, 4.3593, 3.0554], device='cuda:0'), covar=tensor([0.0746, 0.0753, 0.1422, 0.0427, 0.1009, 0.1564, 0.0921, 0.3175], device='cuda:0'), in_proj_covar=tensor([0.0301, 0.0379, 0.0354, 0.0294, 0.0365, 0.0267, 0.0343, 0.0363], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 14:54:36,853 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=215681.0, num_to_drop=1, layers_to_drop={2} 2023-05-16 14:54:42,511 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=215689.0, num_to_drop=1, layers_to_drop={0} 2023-05-16 14:54:53,694 INFO [finetune.py:992] (0/2) Epoch 10, batch 5600, loss[loss=0.1607, simple_loss=0.2382, pruned_loss=0.04163, over 11353.00 frames. ], tot_loss[loss=0.1703, simple_loss=0.2589, pruned_loss=0.04079, over 2356329.99 frames. ], batch size: 25, lr: 4.04e-03, grad_scale: 8.0 2023-05-16 14:55:06,582 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.6180, 5.4167, 5.5170, 5.5944, 5.1915, 5.2028, 5.0102, 5.5371], device='cuda:0'), covar=tensor([0.0608, 0.0554, 0.0726, 0.0475, 0.1638, 0.1215, 0.0463, 0.0874], device='cuda:0'), in_proj_covar=tensor([0.0527, 0.0696, 0.0597, 0.0616, 0.0836, 0.0731, 0.0540, 0.0470], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0005, 0.0004, 0.0004, 0.0005, 0.0005, 0.0003, 0.0003], device='cuda:0') 2023-05-16 14:55:26,850 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.0373, 4.9725, 4.8847, 4.8539, 4.5171, 5.0344, 5.0688, 5.1696], device='cuda:0'), covar=tensor([0.0229, 0.0139, 0.0180, 0.0326, 0.0824, 0.0305, 0.0143, 0.0174], device='cuda:0'), in_proj_covar=tensor([0.0190, 0.0191, 0.0184, 0.0241, 0.0235, 0.0212, 0.0170, 0.0220], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0005, 0.0004, 0.0004, 0.0003, 0.0005], device='cuda:0') 2023-05-16 14:55:29,449 INFO [finetune.py:992] (0/2) Epoch 10, batch 5650, loss[loss=0.1396, simple_loss=0.2243, pruned_loss=0.02749, over 12272.00 frames. ], tot_loss[loss=0.1701, simple_loss=0.2585, pruned_loss=0.04089, over 2350805.43 frames. ], batch size: 28, lr: 4.03e-03, grad_scale: 8.0 2023-05-16 14:55:33,056 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.023e+02 2.706e+02 3.144e+02 3.811e+02 8.174e+02, threshold=6.288e+02, percent-clipped=6.0 2023-05-16 14:56:04,448 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.42 vs. limit=2.0 2023-05-16 14:56:05,988 INFO [finetune.py:992] (0/2) Epoch 10, batch 5700, loss[loss=0.1617, simple_loss=0.2517, pruned_loss=0.03584, over 12028.00 frames. ], tot_loss[loss=0.1695, simple_loss=0.2577, pruned_loss=0.04066, over 2356228.00 frames. ], batch size: 40, lr: 4.03e-03, grad_scale: 8.0 2023-05-16 14:56:23,846 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=215829.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 14:56:40,188 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.4425, 4.7441, 4.2852, 4.9462, 4.5504, 3.0039, 4.3109, 3.1458], device='cuda:0'), covar=tensor([0.0761, 0.0722, 0.1265, 0.0549, 0.1085, 0.1610, 0.0939, 0.3123], device='cuda:0'), in_proj_covar=tensor([0.0304, 0.0383, 0.0356, 0.0297, 0.0367, 0.0269, 0.0346, 0.0365], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 14:56:41,854 INFO [finetune.py:992] (0/2) Epoch 10, batch 5750, loss[loss=0.3673, simple_loss=0.3981, pruned_loss=0.1682, over 8297.00 frames. ], tot_loss[loss=0.1702, simple_loss=0.2585, pruned_loss=0.04095, over 2361063.85 frames. ], batch size: 100, lr: 4.03e-03, grad_scale: 8.0 2023-05-16 14:56:44,823 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=215859.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 14:56:45,449 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.598e+02 2.768e+02 3.361e+02 4.210e+02 8.925e+02, threshold=6.721e+02, percent-clipped=2.0 2023-05-16 14:57:04,036 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=215886.0, num_to_drop=1, layers_to_drop={2} 2023-05-16 14:57:06,769 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=215890.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 14:57:16,499 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=215904.0, num_to_drop=1, layers_to_drop={1} 2023-05-16 14:57:17,101 INFO [finetune.py:992] (0/2) Epoch 10, batch 5800, loss[loss=0.2177, simple_loss=0.3, pruned_loss=0.06772, over 12367.00 frames. ], tot_loss[loss=0.171, simple_loss=0.2594, pruned_loss=0.04128, over 2356238.55 frames. ], batch size: 38, lr: 4.03e-03, grad_scale: 8.0 2023-05-16 14:57:29,065 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=215921.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 14:57:38,270 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=215934.0, num_to_drop=1, layers_to_drop={0} 2023-05-16 14:57:50,338 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.5476, 5.3185, 5.4223, 5.5149, 5.0127, 5.0455, 4.9218, 5.3820], device='cuda:0'), covar=tensor([0.0805, 0.0870, 0.1015, 0.0837, 0.2663, 0.1831, 0.0768, 0.1353], device='cuda:0'), in_proj_covar=tensor([0.0530, 0.0701, 0.0599, 0.0619, 0.0838, 0.0734, 0.0543, 0.0476], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0005, 0.0004, 0.0004, 0.0005, 0.0005, 0.0003, 0.0003], device='cuda:0') 2023-05-16 14:57:52,887 INFO [finetune.py:992] (0/2) Epoch 10, batch 5850, loss[loss=0.1845, simple_loss=0.2585, pruned_loss=0.05521, over 12405.00 frames. ], tot_loss[loss=0.1697, simple_loss=0.2577, pruned_loss=0.04089, over 2365391.75 frames. ], batch size: 32, lr: 4.03e-03, grad_scale: 8.0 2023-05-16 14:57:56,865 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.067e+02 2.721e+02 3.210e+02 4.010e+02 7.242e+02, threshold=6.420e+02, percent-clipped=2.0 2023-05-16 14:58:11,891 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=215981.0, num_to_drop=1, layers_to_drop={2} 2023-05-16 14:58:17,477 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=215989.0, num_to_drop=1, layers_to_drop={0} 2023-05-16 14:58:25,393 INFO [checkpoint.py:75] (0/2) Saving checkpoint to pruned_transducer_stateless7/exp_giga_finetune/checkpoint-116000.pt 2023-05-16 14:58:31,769 INFO [finetune.py:992] (0/2) Epoch 10, batch 5900, loss[loss=0.1512, simple_loss=0.2363, pruned_loss=0.03303, over 11529.00 frames. ], tot_loss[loss=0.1706, simple_loss=0.2585, pruned_loss=0.04135, over 2361052.89 frames. ], batch size: 25, lr: 4.03e-03, grad_scale: 8.0 2023-05-16 14:58:48,664 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=216028.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 14:58:49,234 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=216029.0, num_to_drop=1, layers_to_drop={0} 2023-05-16 14:58:54,916 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=216037.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 14:59:08,234 INFO [finetune.py:992] (0/2) Epoch 10, batch 5950, loss[loss=0.1868, simple_loss=0.2713, pruned_loss=0.05115, over 12048.00 frames. ], tot_loss[loss=0.1712, simple_loss=0.2593, pruned_loss=0.04154, over 2359125.12 frames. ], batch size: 37, lr: 4.03e-03, grad_scale: 8.0 2023-05-16 14:59:11,722 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.189e+02 2.666e+02 3.056e+02 3.587e+02 8.191e+02, threshold=6.112e+02, percent-clipped=1.0 2023-05-16 14:59:27,680 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.1320, 2.5327, 3.6052, 3.0176, 3.4009, 3.1757, 2.4968, 3.5123], device='cuda:0'), covar=tensor([0.0142, 0.0353, 0.0194, 0.0244, 0.0165, 0.0207, 0.0381, 0.0154], device='cuda:0'), in_proj_covar=tensor([0.0173, 0.0204, 0.0187, 0.0183, 0.0212, 0.0159, 0.0195, 0.0181], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 14:59:32,583 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=216089.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 14:59:36,086 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.0939, 2.5581, 3.6921, 2.9854, 3.4543, 3.1853, 2.4626, 3.5139], device='cuda:0'), covar=tensor([0.0118, 0.0365, 0.0106, 0.0242, 0.0136, 0.0184, 0.0337, 0.0131], device='cuda:0'), in_proj_covar=tensor([0.0173, 0.0204, 0.0186, 0.0183, 0.0212, 0.0159, 0.0195, 0.0181], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 14:59:43,223 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.4000, 5.2209, 5.3155, 5.4067, 4.9866, 5.0844, 4.8086, 5.3540], device='cuda:0'), covar=tensor([0.0722, 0.0655, 0.0787, 0.0557, 0.1909, 0.1299, 0.0531, 0.0918], device='cuda:0'), in_proj_covar=tensor([0.0529, 0.0696, 0.0597, 0.0615, 0.0837, 0.0732, 0.0541, 0.0472], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0005, 0.0004, 0.0004, 0.0005, 0.0005, 0.0003, 0.0003], device='cuda:0') 2023-05-16 14:59:44,526 INFO [finetune.py:992] (0/2) Epoch 10, batch 6000, loss[loss=0.1661, simple_loss=0.2561, pruned_loss=0.03803, over 12187.00 frames. ], tot_loss[loss=0.1701, simple_loss=0.2586, pruned_loss=0.04076, over 2368163.75 frames. ], batch size: 31, lr: 4.03e-03, grad_scale: 8.0 2023-05-16 14:59:44,527 INFO [finetune.py:1017] (0/2) Computing validation loss 2023-05-16 15:00:03,003 INFO [finetune.py:1026] (0/2) Epoch 10, validation: loss=0.3144, simple_loss=0.3924, pruned_loss=0.1182, over 1020973.00 frames. 2023-05-16 15:00:03,004 INFO [finetune.py:1027] (0/2) Maximum memory allocated so far is 12827MB 2023-05-16 15:00:38,508 INFO [finetune.py:992] (0/2) Epoch 10, batch 6050, loss[loss=0.1503, simple_loss=0.2388, pruned_loss=0.03095, over 11992.00 frames. ], tot_loss[loss=0.1702, simple_loss=0.2589, pruned_loss=0.04078, over 2370773.70 frames. ], batch size: 28, lr: 4.03e-03, grad_scale: 8.0 2023-05-16 15:00:42,191 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=216159.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 15:00:42,704 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.927e+02 2.596e+02 3.153e+02 3.700e+02 5.951e+02, threshold=6.306e+02, percent-clipped=0.0 2023-05-16 15:00:53,017 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.0306, 3.9765, 3.9760, 4.0891, 3.7812, 3.8465, 3.7308, 4.0016], device='cuda:0'), covar=tensor([0.1079, 0.0711, 0.1274, 0.0688, 0.1830, 0.1360, 0.0623, 0.0997], device='cuda:0'), in_proj_covar=tensor([0.0530, 0.0694, 0.0596, 0.0613, 0.0835, 0.0730, 0.0539, 0.0472], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0005, 0.0004, 0.0004, 0.0005, 0.0005, 0.0003, 0.0003], device='cuda:0') 2023-05-16 15:01:00,918 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=216185.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 15:01:15,302 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=216204.0, num_to_drop=1, layers_to_drop={0} 2023-05-16 15:01:15,862 INFO [finetune.py:992] (0/2) Epoch 10, batch 6100, loss[loss=0.173, simple_loss=0.2608, pruned_loss=0.04262, over 12052.00 frames. ], tot_loss[loss=0.1703, simple_loss=0.2588, pruned_loss=0.04091, over 2368661.92 frames. ], batch size: 42, lr: 4.03e-03, grad_scale: 8.0 2023-05-16 15:01:17,333 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=216207.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 15:01:27,277 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=216221.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 15:01:30,708 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.39 vs. limit=2.0 2023-05-16 15:01:49,238 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=216252.0, num_to_drop=1, layers_to_drop={0} 2023-05-16 15:01:51,213 INFO [finetune.py:992] (0/2) Epoch 10, batch 6150, loss[loss=0.1513, simple_loss=0.2276, pruned_loss=0.03752, over 12010.00 frames. ], tot_loss[loss=0.1704, simple_loss=0.2588, pruned_loss=0.04105, over 2371190.94 frames. ], batch size: 28, lr: 4.03e-03, grad_scale: 8.0 2023-05-16 15:01:54,546 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.108e+02 2.856e+02 3.329e+02 3.835e+02 6.242e+02, threshold=6.658e+02, percent-clipped=0.0 2023-05-16 15:02:01,040 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=216269.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 15:02:26,044 INFO [finetune.py:992] (0/2) Epoch 10, batch 6200, loss[loss=0.1936, simple_loss=0.2564, pruned_loss=0.06539, over 11815.00 frames. ], tot_loss[loss=0.1706, simple_loss=0.2589, pruned_loss=0.04121, over 2382754.84 frames. ], batch size: 26, lr: 4.03e-03, grad_scale: 8.0 2023-05-16 15:02:33,496 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.54 vs. limit=2.0 2023-05-16 15:02:44,743 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.3778, 4.6327, 4.0273, 4.9747, 4.6118, 2.9969, 4.1618, 3.0584], device='cuda:0'), covar=tensor([0.0772, 0.0781, 0.1608, 0.0446, 0.1132, 0.1611, 0.1159, 0.3248], device='cuda:0'), in_proj_covar=tensor([0.0305, 0.0382, 0.0359, 0.0298, 0.0367, 0.0269, 0.0346, 0.0365], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 15:03:02,656 INFO [finetune.py:992] (0/2) Epoch 10, batch 6250, loss[loss=0.1672, simple_loss=0.2499, pruned_loss=0.04225, over 11768.00 frames. ], tot_loss[loss=0.1703, simple_loss=0.2585, pruned_loss=0.04103, over 2385658.36 frames. ], batch size: 44, lr: 4.03e-03, grad_scale: 8.0 2023-05-16 15:03:06,154 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.818e+02 2.634e+02 3.207e+02 4.091e+02 8.473e+02, threshold=6.414e+02, percent-clipped=5.0 2023-05-16 15:03:23,143 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=216384.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 15:03:31,116 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.4304, 4.7947, 4.1409, 5.0652, 4.6980, 2.9487, 4.1715, 3.1945], device='cuda:0'), covar=tensor([0.0740, 0.0672, 0.1390, 0.0431, 0.0922, 0.1645, 0.1115, 0.2911], device='cuda:0'), in_proj_covar=tensor([0.0306, 0.0382, 0.0359, 0.0298, 0.0367, 0.0269, 0.0346, 0.0365], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 15:03:38,230 INFO [finetune.py:992] (0/2) Epoch 10, batch 6300, loss[loss=0.1842, simple_loss=0.2755, pruned_loss=0.04645, over 12047.00 frames. ], tot_loss[loss=0.1712, simple_loss=0.2597, pruned_loss=0.04137, over 2377866.78 frames. ], batch size: 40, lr: 4.03e-03, grad_scale: 8.0 2023-05-16 15:04:03,301 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.9471, 4.7891, 4.7433, 4.8121, 4.3973, 4.9469, 4.9359, 5.1234], device='cuda:0'), covar=tensor([0.0246, 0.0168, 0.0208, 0.0354, 0.0892, 0.0281, 0.0163, 0.0173], device='cuda:0'), in_proj_covar=tensor([0.0193, 0.0196, 0.0188, 0.0246, 0.0242, 0.0216, 0.0173, 0.0224], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0005, 0.0004, 0.0005, 0.0003, 0.0005], device='cuda:0') 2023-05-16 15:04:13,770 INFO [finetune.py:992] (0/2) Epoch 10, batch 6350, loss[loss=0.2554, simple_loss=0.3274, pruned_loss=0.09171, over 8102.00 frames. ], tot_loss[loss=0.1705, simple_loss=0.2589, pruned_loss=0.04106, over 2377847.30 frames. ], batch size: 98, lr: 4.03e-03, grad_scale: 8.0 2023-05-16 15:04:16,886 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.4042, 2.4726, 3.6566, 4.3371, 3.8082, 4.2285, 3.8719, 2.7842], device='cuda:0'), covar=tensor([0.0038, 0.0423, 0.0165, 0.0038, 0.0118, 0.0094, 0.0119, 0.0419], device='cuda:0'), in_proj_covar=tensor([0.0086, 0.0126, 0.0107, 0.0077, 0.0104, 0.0117, 0.0096, 0.0141], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 15:04:17,378 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.032e+02 2.790e+02 3.027e+02 3.610e+02 6.704e+02, threshold=6.054e+02, percent-clipped=1.0 2023-05-16 15:04:35,957 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=216485.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 15:04:42,422 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=2.58 vs. limit=5.0 2023-05-16 15:04:50,461 INFO [finetune.py:992] (0/2) Epoch 10, batch 6400, loss[loss=0.1788, simple_loss=0.2677, pruned_loss=0.04488, over 12045.00 frames. ], tot_loss[loss=0.17, simple_loss=0.2584, pruned_loss=0.04079, over 2384259.96 frames. ], batch size: 40, lr: 4.03e-03, grad_scale: 8.0 2023-05-16 15:05:10,428 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=216533.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 15:05:26,262 INFO [finetune.py:992] (0/2) Epoch 10, batch 6450, loss[loss=0.2487, simple_loss=0.314, pruned_loss=0.09174, over 8236.00 frames. ], tot_loss[loss=0.1704, simple_loss=0.2587, pruned_loss=0.04098, over 2375496.38 frames. ], batch size: 97, lr: 4.03e-03, grad_scale: 8.0 2023-05-16 15:05:29,778 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.061e+02 2.703e+02 3.234e+02 4.071e+02 9.218e+02, threshold=6.469e+02, percent-clipped=5.0 2023-05-16 15:05:51,645 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=216591.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 15:06:01,439 INFO [finetune.py:992] (0/2) Epoch 10, batch 6500, loss[loss=0.1786, simple_loss=0.2644, pruned_loss=0.04638, over 12170.00 frames. ], tot_loss[loss=0.1702, simple_loss=0.2586, pruned_loss=0.04089, over 2377479.78 frames. ], batch size: 29, lr: 4.03e-03, grad_scale: 8.0 2023-05-16 15:06:35,890 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=216652.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 15:06:37,802 INFO [finetune.py:992] (0/2) Epoch 10, batch 6550, loss[loss=0.1642, simple_loss=0.257, pruned_loss=0.03571, over 11599.00 frames. ], tot_loss[loss=0.1706, simple_loss=0.2593, pruned_loss=0.04091, over 2378190.45 frames. ], batch size: 48, lr: 4.03e-03, grad_scale: 8.0 2023-05-16 15:06:37,990 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=216655.0, num_to_drop=1, layers_to_drop={1} 2023-05-16 15:06:41,340 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.881e+02 2.848e+02 3.255e+02 3.871e+02 6.183e+02, threshold=6.511e+02, percent-clipped=0.0 2023-05-16 15:06:58,185 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=216684.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 15:07:04,031 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.0393, 4.5659, 4.7774, 4.8901, 4.6413, 4.9505, 4.8382, 2.4022], device='cuda:0'), covar=tensor([0.0135, 0.0082, 0.0106, 0.0068, 0.0064, 0.0113, 0.0083, 0.0931], device='cuda:0'), in_proj_covar=tensor([0.0067, 0.0078, 0.0080, 0.0072, 0.0060, 0.0091, 0.0080, 0.0096], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 15:07:13,052 INFO [finetune.py:992] (0/2) Epoch 10, batch 6600, loss[loss=0.1636, simple_loss=0.2578, pruned_loss=0.03474, over 11136.00 frames. ], tot_loss[loss=0.1715, simple_loss=0.2604, pruned_loss=0.04134, over 2375671.38 frames. ], batch size: 55, lr: 4.03e-03, grad_scale: 8.0 2023-05-16 15:07:13,261 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=216705.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 15:07:21,066 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=216716.0, num_to_drop=1, layers_to_drop={1} 2023-05-16 15:07:32,393 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=216732.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 15:07:36,087 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.0296, 2.4895, 3.4880, 3.9300, 3.6253, 3.8944, 3.6027, 2.6302], device='cuda:0'), covar=tensor([0.0042, 0.0366, 0.0146, 0.0047, 0.0125, 0.0095, 0.0111, 0.0398], device='cuda:0'), in_proj_covar=tensor([0.0085, 0.0124, 0.0106, 0.0076, 0.0103, 0.0116, 0.0096, 0.0139], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 15:07:49,386 INFO [finetune.py:992] (0/2) Epoch 10, batch 6650, loss[loss=0.1364, simple_loss=0.2154, pruned_loss=0.02869, over 11827.00 frames. ], tot_loss[loss=0.1718, simple_loss=0.2606, pruned_loss=0.04149, over 2372021.18 frames. ], batch size: 26, lr: 4.02e-03, grad_scale: 8.0 2023-05-16 15:07:53,063 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.862e+02 2.856e+02 3.449e+02 3.977e+02 7.392e+02, threshold=6.898e+02, percent-clipped=2.0 2023-05-16 15:07:57,587 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=216766.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 15:08:26,183 INFO [finetune.py:992] (0/2) Epoch 10, batch 6700, loss[loss=0.2055, simple_loss=0.2955, pruned_loss=0.0578, over 12151.00 frames. ], tot_loss[loss=0.1714, simple_loss=0.2598, pruned_loss=0.04146, over 2360194.92 frames. ], batch size: 36, lr: 4.02e-03, grad_scale: 8.0 2023-05-16 15:08:57,852 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.7918, 4.6854, 4.7310, 4.7604, 4.2816, 4.8539, 4.7897, 5.0281], device='cuda:0'), covar=tensor([0.0340, 0.0187, 0.0210, 0.0360, 0.0878, 0.0387, 0.0208, 0.0205], device='cuda:0'), in_proj_covar=tensor([0.0194, 0.0196, 0.0188, 0.0246, 0.0241, 0.0216, 0.0173, 0.0225], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0005, 0.0004, 0.0005, 0.0003, 0.0005], device='cuda:0') 2023-05-16 15:09:01,825 INFO [finetune.py:992] (0/2) Epoch 10, batch 6750, loss[loss=0.1579, simple_loss=0.2458, pruned_loss=0.03504, over 12173.00 frames. ], tot_loss[loss=0.1707, simple_loss=0.2591, pruned_loss=0.04113, over 2373689.23 frames. ], batch size: 31, lr: 4.02e-03, grad_scale: 8.0 2023-05-16 15:09:05,468 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.957e+02 2.677e+02 3.140e+02 3.668e+02 5.488e+02, threshold=6.280e+02, percent-clipped=0.0 2023-05-16 15:09:28,271 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.5183, 5.3132, 5.4072, 5.4788, 5.0940, 5.1026, 4.8253, 5.3761], device='cuda:0'), covar=tensor([0.0636, 0.0557, 0.0777, 0.0569, 0.1747, 0.1252, 0.0550, 0.1092], device='cuda:0'), in_proj_covar=tensor([0.0542, 0.0704, 0.0606, 0.0627, 0.0847, 0.0742, 0.0550, 0.0480], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0005, 0.0004, 0.0004, 0.0005, 0.0005, 0.0004, 0.0003], device='cuda:0') 2023-05-16 15:09:38,007 INFO [finetune.py:992] (0/2) Epoch 10, batch 6800, loss[loss=0.1553, simple_loss=0.2448, pruned_loss=0.03294, over 12096.00 frames. ], tot_loss[loss=0.1711, simple_loss=0.2597, pruned_loss=0.04127, over 2381725.34 frames. ], batch size: 33, lr: 4.02e-03, grad_scale: 8.0 2023-05-16 15:09:43,142 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.0859, 4.9555, 4.9932, 5.0122, 4.5913, 5.1250, 5.0925, 5.3130], device='cuda:0'), covar=tensor([0.0176, 0.0156, 0.0166, 0.0298, 0.0744, 0.0224, 0.0135, 0.0162], device='cuda:0'), in_proj_covar=tensor([0.0193, 0.0195, 0.0187, 0.0245, 0.0240, 0.0215, 0.0172, 0.0224], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0005, 0.0004, 0.0004, 0.0003, 0.0005], device='cuda:0') 2023-05-16 15:09:59,714 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.6378, 4.3493, 4.3387, 4.7175, 3.5507, 4.2164, 2.9800, 4.4537], device='cuda:0'), covar=tensor([0.1382, 0.0590, 0.0779, 0.0587, 0.0905, 0.0519, 0.1446, 0.1112], device='cuda:0'), in_proj_covar=tensor([0.0225, 0.0266, 0.0295, 0.0355, 0.0236, 0.0240, 0.0259, 0.0378], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 15:10:08,071 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=216947.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 15:10:13,668 INFO [finetune.py:992] (0/2) Epoch 10, batch 6850, loss[loss=0.1778, simple_loss=0.2732, pruned_loss=0.04124, over 12017.00 frames. ], tot_loss[loss=0.1712, simple_loss=0.2599, pruned_loss=0.04123, over 2387475.09 frames. ], batch size: 40, lr: 4.02e-03, grad_scale: 8.0 2023-05-16 15:10:17,030 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.576e+02 2.810e+02 3.435e+02 3.923e+02 9.714e+02, threshold=6.870e+02, percent-clipped=3.0 2023-05-16 15:10:20,681 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=216965.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 15:10:21,502 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.5890, 2.8542, 3.3323, 4.4790, 2.5303, 4.5150, 4.4944, 4.6412], device='cuda:0'), covar=tensor([0.0110, 0.1139, 0.0482, 0.0161, 0.1341, 0.0251, 0.0181, 0.0094], device='cuda:0'), in_proj_covar=tensor([0.0123, 0.0203, 0.0186, 0.0116, 0.0189, 0.0179, 0.0176, 0.0119], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 15:10:44,075 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.4365, 2.5053, 3.1633, 4.3568, 2.3017, 4.4980, 4.4704, 4.4516], device='cuda:0'), covar=tensor([0.0139, 0.1369, 0.0545, 0.0201, 0.1460, 0.0197, 0.0182, 0.0123], device='cuda:0'), in_proj_covar=tensor([0.0122, 0.0202, 0.0185, 0.0116, 0.0189, 0.0178, 0.0176, 0.0119], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 15:10:48,869 INFO [finetune.py:992] (0/2) Epoch 10, batch 6900, loss[loss=0.175, simple_loss=0.259, pruned_loss=0.04555, over 12108.00 frames. ], tot_loss[loss=0.1708, simple_loss=0.259, pruned_loss=0.04131, over 2382486.28 frames. ], batch size: 33, lr: 4.02e-03, grad_scale: 8.0 2023-05-16 15:10:53,246 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=217011.0, num_to_drop=1, layers_to_drop={2} 2023-05-16 15:11:03,867 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=217026.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 15:11:15,779 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.85 vs. limit=2.0 2023-05-16 15:11:24,439 INFO [finetune.py:992] (0/2) Epoch 10, batch 6950, loss[loss=0.1484, simple_loss=0.2428, pruned_loss=0.02701, over 12076.00 frames. ], tot_loss[loss=0.171, simple_loss=0.2596, pruned_loss=0.04118, over 2376655.84 frames. ], batch size: 32, lr: 4.02e-03, grad_scale: 16.0 2023-05-16 15:11:28,592 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.871e+02 2.811e+02 3.227e+02 3.716e+02 6.951e+02, threshold=6.455e+02, percent-clipped=1.0 2023-05-16 15:11:29,370 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=217061.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 15:11:32,584 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.41 vs. limit=2.0 2023-05-16 15:11:34,538 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.7567, 2.8555, 3.3118, 4.6105, 2.7226, 4.6621, 4.6651, 4.7918], device='cuda:0'), covar=tensor([0.0088, 0.1117, 0.0500, 0.0163, 0.1159, 0.0202, 0.0136, 0.0072], device='cuda:0'), in_proj_covar=tensor([0.0122, 0.0202, 0.0184, 0.0116, 0.0188, 0.0178, 0.0175, 0.0119], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 15:11:40,868 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.2504, 6.0554, 5.6268, 5.5995, 6.1359, 5.5113, 5.6398, 5.5964], device='cuda:0'), covar=tensor([0.1227, 0.0812, 0.0877, 0.1711, 0.0856, 0.1942, 0.1705, 0.1057], device='cuda:0'), in_proj_covar=tensor([0.0349, 0.0482, 0.0383, 0.0431, 0.0453, 0.0433, 0.0387, 0.0371], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 15:12:01,371 INFO [finetune.py:992] (0/2) Epoch 10, batch 7000, loss[loss=0.1847, simple_loss=0.274, pruned_loss=0.04765, over 11634.00 frames. ], tot_loss[loss=0.1698, simple_loss=0.2587, pruned_loss=0.04043, over 2380118.71 frames. ], batch size: 48, lr: 4.02e-03, grad_scale: 16.0 2023-05-16 15:12:08,812 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.59 vs. limit=2.0 2023-05-16 15:12:36,438 INFO [finetune.py:992] (0/2) Epoch 10, batch 7050, loss[loss=0.1815, simple_loss=0.2746, pruned_loss=0.04417, over 10612.00 frames. ], tot_loss[loss=0.1706, simple_loss=0.2595, pruned_loss=0.04085, over 2368790.13 frames. ], batch size: 68, lr: 4.02e-03, grad_scale: 16.0 2023-05-16 15:12:39,933 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.916e+02 2.825e+02 3.318e+02 4.081e+02 8.200e+02, threshold=6.636e+02, percent-clipped=0.0 2023-05-16 15:13:06,882 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.1731, 2.4071, 3.5905, 4.0463, 3.7427, 4.1268, 3.6947, 2.7828], device='cuda:0'), covar=tensor([0.0043, 0.0410, 0.0141, 0.0044, 0.0118, 0.0074, 0.0119, 0.0414], device='cuda:0'), in_proj_covar=tensor([0.0085, 0.0124, 0.0106, 0.0076, 0.0102, 0.0115, 0.0095, 0.0139], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 15:13:12,711 INFO [finetune.py:992] (0/2) Epoch 10, batch 7100, loss[loss=0.1833, simple_loss=0.2797, pruned_loss=0.04348, over 11311.00 frames. ], tot_loss[loss=0.1703, simple_loss=0.2591, pruned_loss=0.04077, over 2365865.41 frames. ], batch size: 55, lr: 4.02e-03, grad_scale: 16.0 2023-05-16 15:13:43,132 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=217247.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 15:13:48,453 INFO [finetune.py:992] (0/2) Epoch 10, batch 7150, loss[loss=0.1704, simple_loss=0.2474, pruned_loss=0.04667, over 12023.00 frames. ], tot_loss[loss=0.1706, simple_loss=0.2593, pruned_loss=0.04095, over 2361197.51 frames. ], batch size: 28, lr: 4.02e-03, grad_scale: 16.0 2023-05-16 15:13:51,964 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.139e+02 2.573e+02 3.098e+02 4.027e+02 8.429e+02, threshold=6.196e+02, percent-clipped=2.0 2023-05-16 15:14:16,735 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=217295.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 15:14:23,737 INFO [finetune.py:992] (0/2) Epoch 10, batch 7200, loss[loss=0.1635, simple_loss=0.2533, pruned_loss=0.03686, over 11842.00 frames. ], tot_loss[loss=0.171, simple_loss=0.2592, pruned_loss=0.04144, over 2357968.08 frames. ], batch size: 44, lr: 4.02e-03, grad_scale: 16.0 2023-05-16 15:14:28,036 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=217311.0, num_to_drop=1, layers_to_drop={2} 2023-05-16 15:14:34,901 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=217321.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 15:14:59,445 INFO [finetune.py:992] (0/2) Epoch 10, batch 7250, loss[loss=0.1876, simple_loss=0.2784, pruned_loss=0.04836, over 11740.00 frames. ], tot_loss[loss=0.1714, simple_loss=0.2599, pruned_loss=0.04146, over 2364748.87 frames. ], batch size: 49, lr: 4.02e-03, grad_scale: 16.0 2023-05-16 15:15:02,434 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=217359.0, num_to_drop=1, layers_to_drop={0} 2023-05-16 15:15:02,999 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.837e+02 2.680e+02 3.204e+02 3.936e+02 1.205e+03, threshold=6.408e+02, percent-clipped=3.0 2023-05-16 15:15:03,936 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=217361.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 15:15:11,403 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.37 vs. limit=2.0 2023-05-16 15:15:23,850 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=217388.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 15:15:35,814 INFO [finetune.py:992] (0/2) Epoch 10, batch 7300, loss[loss=0.1669, simple_loss=0.254, pruned_loss=0.03992, over 12124.00 frames. ], tot_loss[loss=0.1711, simple_loss=0.2594, pruned_loss=0.04136, over 2365508.15 frames. ], batch size: 39, lr: 4.02e-03, grad_scale: 16.0 2023-05-16 15:15:38,697 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=217409.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 15:15:47,390 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.5120, 2.5773, 3.7586, 4.3566, 3.9034, 4.4228, 3.8822, 3.2732], device='cuda:0'), covar=tensor([0.0034, 0.0398, 0.0136, 0.0043, 0.0123, 0.0075, 0.0117, 0.0325], device='cuda:0'), in_proj_covar=tensor([0.0085, 0.0125, 0.0107, 0.0076, 0.0103, 0.0117, 0.0097, 0.0140], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 15:15:49,569 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.1525, 2.7153, 3.7675, 3.1257, 3.6028, 3.2920, 2.6454, 3.6672], device='cuda:0'), covar=tensor([0.0130, 0.0310, 0.0124, 0.0231, 0.0124, 0.0156, 0.0326, 0.0103], device='cuda:0'), in_proj_covar=tensor([0.0174, 0.0206, 0.0188, 0.0185, 0.0215, 0.0161, 0.0196, 0.0183], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 15:16:07,302 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=217449.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 15:16:10,133 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.6560, 3.0172, 3.8197, 4.5966, 4.0025, 4.5885, 3.9939, 3.4900], device='cuda:0'), covar=tensor([0.0028, 0.0294, 0.0136, 0.0028, 0.0112, 0.0060, 0.0123, 0.0274], device='cuda:0'), in_proj_covar=tensor([0.0086, 0.0126, 0.0107, 0.0076, 0.0104, 0.0117, 0.0097, 0.0141], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 15:16:11,445 INFO [finetune.py:992] (0/2) Epoch 10, batch 7350, loss[loss=0.1602, simple_loss=0.2536, pruned_loss=0.03345, over 12178.00 frames. ], tot_loss[loss=0.1701, simple_loss=0.2585, pruned_loss=0.04081, over 2376529.34 frames. ], batch size: 35, lr: 4.02e-03, grad_scale: 16.0 2023-05-16 15:16:14,996 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.128e+02 2.746e+02 3.216e+02 3.952e+02 7.752e+02, threshold=6.432e+02, percent-clipped=2.0 2023-05-16 15:16:15,407 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=2.02 vs. limit=2.0 2023-05-16 15:16:37,912 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=217492.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 15:16:47,577 INFO [finetune.py:992] (0/2) Epoch 10, batch 7400, loss[loss=0.1638, simple_loss=0.2615, pruned_loss=0.03307, over 12147.00 frames. ], tot_loss[loss=0.1699, simple_loss=0.2586, pruned_loss=0.04063, over 2372843.14 frames. ], batch size: 34, lr: 4.02e-03, grad_scale: 16.0 2023-05-16 15:17:22,067 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=217553.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 15:17:23,268 INFO [finetune.py:992] (0/2) Epoch 10, batch 7450, loss[loss=0.1551, simple_loss=0.2467, pruned_loss=0.03179, over 12238.00 frames. ], tot_loss[loss=0.1715, simple_loss=0.2602, pruned_loss=0.04145, over 2366064.02 frames. ], batch size: 32, lr: 4.02e-03, grad_scale: 8.0 2023-05-16 15:17:27,441 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.888e+02 2.751e+02 3.168e+02 3.931e+02 8.735e+02, threshold=6.336e+02, percent-clipped=4.0 2023-05-16 15:17:58,944 INFO [finetune.py:992] (0/2) Epoch 10, batch 7500, loss[loss=0.1833, simple_loss=0.2696, pruned_loss=0.04847, over 11790.00 frames. ], tot_loss[loss=0.1714, simple_loss=0.26, pruned_loss=0.04135, over 2368508.38 frames. ], batch size: 44, lr: 4.02e-03, grad_scale: 8.0 2023-05-16 15:18:10,357 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=217621.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 15:18:13,796 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.3721, 5.1539, 5.2646, 5.3327, 4.9716, 5.0132, 4.7597, 5.2678], device='cuda:0'), covar=tensor([0.0602, 0.0574, 0.0833, 0.0588, 0.1791, 0.1237, 0.0593, 0.0846], device='cuda:0'), in_proj_covar=tensor([0.0524, 0.0685, 0.0592, 0.0618, 0.0829, 0.0726, 0.0539, 0.0468], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0005, 0.0004, 0.0004, 0.0005, 0.0005, 0.0003, 0.0003], device='cuda:0') 2023-05-16 15:18:34,886 INFO [finetune.py:992] (0/2) Epoch 10, batch 7550, loss[loss=0.1637, simple_loss=0.2626, pruned_loss=0.03242, over 12166.00 frames. ], tot_loss[loss=0.1714, simple_loss=0.2601, pruned_loss=0.04135, over 2371899.43 frames. ], batch size: 36, lr: 4.02e-03, grad_scale: 8.0 2023-05-16 15:18:39,782 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.896e+02 2.719e+02 3.185e+02 3.698e+02 8.240e+02, threshold=6.370e+02, percent-clipped=1.0 2023-05-16 15:18:42,092 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=217664.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 15:18:45,380 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=217669.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 15:19:09,573 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.2239, 4.5289, 4.0939, 4.7921, 4.5609, 2.8338, 4.1715, 3.1065], device='cuda:0'), covar=tensor([0.0839, 0.0869, 0.1416, 0.0464, 0.1002, 0.1678, 0.1056, 0.3207], device='cuda:0'), in_proj_covar=tensor([0.0306, 0.0381, 0.0357, 0.0298, 0.0367, 0.0267, 0.0346, 0.0363], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 15:19:10,690 INFO [finetune.py:992] (0/2) Epoch 10, batch 7600, loss[loss=0.1486, simple_loss=0.232, pruned_loss=0.03255, over 12079.00 frames. ], tot_loss[loss=0.1715, simple_loss=0.2604, pruned_loss=0.04129, over 2374804.44 frames. ], batch size: 32, lr: 4.02e-03, grad_scale: 8.0 2023-05-16 15:19:13,712 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.9133, 5.8754, 5.6335, 5.1670, 5.0907, 5.7644, 5.3747, 5.1625], device='cuda:0'), covar=tensor([0.0629, 0.0878, 0.0673, 0.1503, 0.0676, 0.0705, 0.1615, 0.1160], device='cuda:0'), in_proj_covar=tensor([0.0609, 0.0552, 0.0514, 0.0633, 0.0409, 0.0714, 0.0775, 0.0570], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0004, 0.0004, 0.0003], device='cuda:0') 2023-05-16 15:19:14,004 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.12 vs. limit=2.0 2023-05-16 15:19:25,194 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=217725.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 15:19:38,431 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=217744.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 15:19:45,713 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.6170, 3.7440, 3.3223, 3.2671, 3.0630, 2.8925, 3.6691, 2.4544], device='cuda:0'), covar=tensor([0.0360, 0.0107, 0.0176, 0.0178, 0.0317, 0.0329, 0.0118, 0.0445], device='cuda:0'), in_proj_covar=tensor([0.0189, 0.0159, 0.0156, 0.0182, 0.0199, 0.0194, 0.0167, 0.0196], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 15:19:46,220 INFO [finetune.py:992] (0/2) Epoch 10, batch 7650, loss[loss=0.1487, simple_loss=0.2405, pruned_loss=0.0285, over 12286.00 frames. ], tot_loss[loss=0.1719, simple_loss=0.2608, pruned_loss=0.0415, over 2371168.19 frames. ], batch size: 33, lr: 4.01e-03, grad_scale: 8.0 2023-05-16 15:19:50,510 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.694e+02 2.731e+02 3.164e+02 3.954e+02 1.132e+03, threshold=6.329e+02, percent-clipped=2.0 2023-05-16 15:20:05,776 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.5536, 2.8176, 3.9472, 2.3361, 2.5642, 3.1724, 2.8669, 3.3102], device='cuda:0'), covar=tensor([0.0666, 0.1260, 0.0384, 0.1302, 0.1813, 0.1493, 0.1370, 0.1081], device='cuda:0'), in_proj_covar=tensor([0.0239, 0.0235, 0.0247, 0.0180, 0.0236, 0.0293, 0.0222, 0.0262], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-16 15:20:23,910 INFO [finetune.py:992] (0/2) Epoch 10, batch 7700, loss[loss=0.2136, simple_loss=0.3011, pruned_loss=0.06302, over 10487.00 frames. ], tot_loss[loss=0.1724, simple_loss=0.2614, pruned_loss=0.0417, over 2365025.53 frames. ], batch size: 68, lr: 4.01e-03, grad_scale: 8.0 2023-05-16 15:20:25,480 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.7751, 2.1651, 3.5898, 2.8593, 3.4140, 3.0553, 2.2070, 3.4116], device='cuda:0'), covar=tensor([0.0207, 0.0500, 0.0181, 0.0288, 0.0173, 0.0217, 0.0473, 0.0152], device='cuda:0'), in_proj_covar=tensor([0.0170, 0.0201, 0.0184, 0.0182, 0.0210, 0.0157, 0.0192, 0.0180], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 15:20:54,623 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=217848.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 15:20:59,402 INFO [finetune.py:992] (0/2) Epoch 10, batch 7750, loss[loss=0.1947, simple_loss=0.2926, pruned_loss=0.04843, over 12347.00 frames. ], tot_loss[loss=0.1724, simple_loss=0.2615, pruned_loss=0.04169, over 2366195.23 frames. ], batch size: 36, lr: 4.01e-03, grad_scale: 8.0 2023-05-16 15:21:03,597 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.707e+02 2.759e+02 3.257e+02 3.915e+02 7.146e+02, threshold=6.515e+02, percent-clipped=1.0 2023-05-16 15:21:09,528 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.1791, 6.0084, 5.6105, 5.6490, 6.1262, 5.4035, 5.7594, 5.5211], device='cuda:0'), covar=tensor([0.1443, 0.0862, 0.1028, 0.1826, 0.0860, 0.1993, 0.1509, 0.1218], device='cuda:0'), in_proj_covar=tensor([0.0356, 0.0491, 0.0390, 0.0437, 0.0463, 0.0442, 0.0393, 0.0379], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 15:21:30,258 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=217898.0, num_to_drop=1, layers_to_drop={1} 2023-05-16 15:21:35,069 INFO [finetune.py:992] (0/2) Epoch 10, batch 7800, loss[loss=0.2324, simple_loss=0.3015, pruned_loss=0.08165, over 7926.00 frames. ], tot_loss[loss=0.1723, simple_loss=0.2609, pruned_loss=0.04185, over 2364429.53 frames. ], batch size: 98, lr: 4.01e-03, grad_scale: 8.0 2023-05-16 15:22:11,812 INFO [finetune.py:992] (0/2) Epoch 10, batch 7850, loss[loss=0.1736, simple_loss=0.2505, pruned_loss=0.04831, over 12123.00 frames. ], tot_loss[loss=0.1735, simple_loss=0.2616, pruned_loss=0.04273, over 2349605.05 frames. ], batch size: 30, lr: 4.01e-03, grad_scale: 8.0 2023-05-16 15:22:14,726 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=217959.0, num_to_drop=1, layers_to_drop={2} 2023-05-16 15:22:15,901 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.956e+02 2.816e+02 3.296e+02 4.410e+02 8.166e+02, threshold=6.592e+02, percent-clipped=3.0 2023-05-16 15:22:31,055 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.2978, 3.9960, 4.1953, 4.3985, 3.0525, 3.8854, 2.6479, 4.1187], device='cuda:0'), covar=tensor([0.1568, 0.0727, 0.0791, 0.0678, 0.1178, 0.0631, 0.1844, 0.1298], device='cuda:0'), in_proj_covar=tensor([0.0227, 0.0266, 0.0298, 0.0357, 0.0236, 0.0240, 0.0261, 0.0381], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 15:22:41,748 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.4721, 2.4323, 3.6381, 4.2461, 3.7926, 4.2229, 3.8439, 3.0183], device='cuda:0'), covar=tensor([0.0030, 0.0429, 0.0153, 0.0046, 0.0147, 0.0099, 0.0133, 0.0389], device='cuda:0'), in_proj_covar=tensor([0.0085, 0.0125, 0.0107, 0.0076, 0.0103, 0.0117, 0.0096, 0.0139], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 15:22:43,982 INFO [checkpoint.py:75] (0/2) Saving checkpoint to pruned_transducer_stateless7/exp_giga_finetune/checkpoint-118000.pt 2023-05-16 15:22:50,568 INFO [finetune.py:992] (0/2) Epoch 10, batch 7900, loss[loss=0.1893, simple_loss=0.2848, pruned_loss=0.04692, over 10704.00 frames. ], tot_loss[loss=0.1749, simple_loss=0.2628, pruned_loss=0.0435, over 2338988.90 frames. ], batch size: 68, lr: 4.01e-03, grad_scale: 8.0 2023-05-16 15:22:57,240 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.7741, 2.7049, 3.8138, 4.5923, 4.0642, 4.6327, 4.0924, 3.1460], device='cuda:0'), covar=tensor([0.0033, 0.0413, 0.0146, 0.0036, 0.0114, 0.0074, 0.0097, 0.0412], device='cuda:0'), in_proj_covar=tensor([0.0085, 0.0124, 0.0106, 0.0076, 0.0103, 0.0117, 0.0096, 0.0139], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 15:23:01,342 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=218020.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 15:23:18,369 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=218044.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 15:23:26,577 INFO [finetune.py:992] (0/2) Epoch 10, batch 7950, loss[loss=0.1712, simple_loss=0.2721, pruned_loss=0.03515, over 12049.00 frames. ], tot_loss[loss=0.1736, simple_loss=0.2619, pruned_loss=0.04259, over 2352608.19 frames. ], batch size: 40, lr: 4.01e-03, grad_scale: 8.0 2023-05-16 15:23:30,789 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.897e+02 2.590e+02 3.175e+02 4.044e+02 9.490e+02, threshold=6.351e+02, percent-clipped=3.0 2023-05-16 15:23:53,232 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=218092.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 15:23:57,059 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.2844, 4.0468, 4.1384, 4.4251, 3.0656, 3.9610, 2.8113, 4.0829], device='cuda:0'), covar=tensor([0.1650, 0.0726, 0.0783, 0.0666, 0.1141, 0.0611, 0.1700, 0.1306], device='cuda:0'), in_proj_covar=tensor([0.0227, 0.0267, 0.0298, 0.0356, 0.0236, 0.0240, 0.0260, 0.0382], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 15:23:59,182 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=218100.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 15:24:02,446 INFO [finetune.py:992] (0/2) Epoch 10, batch 8000, loss[loss=0.1657, simple_loss=0.2573, pruned_loss=0.03708, over 12117.00 frames. ], tot_loss[loss=0.1742, simple_loss=0.2621, pruned_loss=0.04316, over 2350034.09 frames. ], batch size: 33, lr: 4.01e-03, grad_scale: 8.0 2023-05-16 15:24:12,516 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.1109, 4.7579, 5.0626, 4.9836, 4.8950, 5.0726, 4.9141, 2.9620], device='cuda:0'), covar=tensor([0.0085, 0.0074, 0.0066, 0.0052, 0.0045, 0.0082, 0.0102, 0.0641], device='cuda:0'), in_proj_covar=tensor([0.0067, 0.0077, 0.0080, 0.0073, 0.0060, 0.0090, 0.0079, 0.0096], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-16 15:24:24,484 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.7491, 2.8317, 3.8867, 4.6029, 4.0348, 4.5846, 4.0892, 3.2246], device='cuda:0'), covar=tensor([0.0027, 0.0395, 0.0124, 0.0036, 0.0122, 0.0072, 0.0111, 0.0354], device='cuda:0'), in_proj_covar=tensor([0.0084, 0.0124, 0.0106, 0.0076, 0.0102, 0.0116, 0.0095, 0.0138], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 15:24:32,928 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=218148.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 15:24:37,729 INFO [finetune.py:992] (0/2) Epoch 10, batch 8050, loss[loss=0.1422, simple_loss=0.2173, pruned_loss=0.03361, over 12005.00 frames. ], tot_loss[loss=0.1734, simple_loss=0.2611, pruned_loss=0.04291, over 2351296.56 frames. ], batch size: 28, lr: 4.01e-03, grad_scale: 8.0 2023-05-16 15:24:38,257 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.31 vs. limit=2.0 2023-05-16 15:24:41,970 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.991e+02 2.803e+02 3.296e+02 3.767e+02 8.425e+02, threshold=6.591e+02, percent-clipped=3.0 2023-05-16 15:24:42,238 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=218161.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 15:24:43,658 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.0491, 4.6935, 4.8932, 4.9027, 4.8355, 5.0618, 4.8046, 2.7899], device='cuda:0'), covar=tensor([0.0139, 0.0106, 0.0088, 0.0066, 0.0046, 0.0081, 0.0094, 0.0766], device='cuda:0'), in_proj_covar=tensor([0.0067, 0.0077, 0.0080, 0.0072, 0.0060, 0.0090, 0.0079, 0.0096], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-16 15:24:57,253 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=218182.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 15:25:07,045 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=218196.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 15:25:14,227 INFO [finetune.py:992] (0/2) Epoch 10, batch 8100, loss[loss=0.1773, simple_loss=0.2757, pruned_loss=0.03943, over 12291.00 frames. ], tot_loss[loss=0.1725, simple_loss=0.2602, pruned_loss=0.0424, over 2353448.87 frames. ], batch size: 37, lr: 4.01e-03, grad_scale: 8.0 2023-05-16 15:25:15,908 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.9737, 3.0402, 4.4284, 2.4019, 2.7431, 3.3414, 2.9547, 3.5452], device='cuda:0'), covar=tensor([0.0615, 0.1195, 0.0430, 0.1254, 0.1802, 0.1381, 0.1412, 0.1120], device='cuda:0'), in_proj_covar=tensor([0.0239, 0.0235, 0.0246, 0.0180, 0.0236, 0.0294, 0.0223, 0.0261], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-16 15:25:18,049 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=218210.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 15:25:42,043 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=218243.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 15:25:49,815 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=218254.0, num_to_drop=1, layers_to_drop={1} 2023-05-16 15:25:50,442 INFO [finetune.py:992] (0/2) Epoch 10, batch 8150, loss[loss=0.1772, simple_loss=0.2612, pruned_loss=0.04663, over 12185.00 frames. ], tot_loss[loss=0.1728, simple_loss=0.2606, pruned_loss=0.04246, over 2353051.15 frames. ], batch size: 35, lr: 4.01e-03, grad_scale: 8.0 2023-05-16 15:25:54,535 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.745e+02 2.787e+02 3.392e+02 4.159e+02 9.116e+02, threshold=6.785e+02, percent-clipped=3.0 2023-05-16 15:26:01,816 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=218271.0, num_to_drop=1, layers_to_drop={1} 2023-05-16 15:26:02,772 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=4.64 vs. limit=5.0 2023-05-16 15:26:25,826 INFO [finetune.py:992] (0/2) Epoch 10, batch 8200, loss[loss=0.1669, simple_loss=0.2629, pruned_loss=0.03545, over 12047.00 frames. ], tot_loss[loss=0.173, simple_loss=0.2606, pruned_loss=0.04271, over 2358459.64 frames. ], batch size: 42, lr: 4.01e-03, grad_scale: 8.0 2023-05-16 15:26:36,383 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=218320.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 15:26:44,128 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.5685, 2.8168, 4.3573, 4.5462, 2.8312, 2.5869, 2.7522, 2.0198], device='cuda:0'), covar=tensor([0.1686, 0.3102, 0.0568, 0.0497, 0.1334, 0.2392, 0.2896, 0.4249], device='cuda:0'), in_proj_covar=tensor([0.0288, 0.0382, 0.0271, 0.0298, 0.0268, 0.0300, 0.0374, 0.0370], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-16 15:27:01,508 INFO [finetune.py:992] (0/2) Epoch 10, batch 8250, loss[loss=0.1757, simple_loss=0.27, pruned_loss=0.04067, over 12193.00 frames. ], tot_loss[loss=0.1737, simple_loss=0.2615, pruned_loss=0.04299, over 2345794.25 frames. ], batch size: 35, lr: 4.01e-03, grad_scale: 8.0 2023-05-16 15:27:05,745 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.999e+02 2.826e+02 3.147e+02 3.997e+02 2.145e+03, threshold=6.293e+02, percent-clipped=2.0 2023-05-16 15:27:11,575 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=218368.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 15:27:38,230 INFO [finetune.py:992] (0/2) Epoch 10, batch 8300, loss[loss=0.2113, simple_loss=0.3027, pruned_loss=0.05993, over 12343.00 frames. ], tot_loss[loss=0.173, simple_loss=0.2613, pruned_loss=0.04238, over 2358556.37 frames. ], batch size: 36, lr: 4.01e-03, grad_scale: 8.0 2023-05-16 15:28:06,544 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.9599, 4.8067, 4.7955, 4.7525, 4.4594, 4.8690, 4.9542, 5.0553], device='cuda:0'), covar=tensor([0.0213, 0.0168, 0.0171, 0.0386, 0.0703, 0.0315, 0.0146, 0.0200], device='cuda:0'), in_proj_covar=tensor([0.0192, 0.0195, 0.0187, 0.0246, 0.0238, 0.0216, 0.0171, 0.0222], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0005, 0.0004, 0.0005, 0.0003, 0.0005], device='cuda:0') 2023-05-16 15:28:13,560 INFO [finetune.py:992] (0/2) Epoch 10, batch 8350, loss[loss=0.201, simple_loss=0.2988, pruned_loss=0.05155, over 11243.00 frames. ], tot_loss[loss=0.1732, simple_loss=0.2617, pruned_loss=0.04233, over 2356125.89 frames. ], batch size: 55, lr: 4.01e-03, grad_scale: 8.0 2023-05-16 15:28:14,345 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=218456.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 15:28:17,978 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.747e+02 2.649e+02 3.148e+02 3.755e+02 8.260e+02, threshold=6.296e+02, percent-clipped=2.0 2023-05-16 15:28:31,601 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.3913, 4.9477, 5.4110, 4.6658, 4.9167, 4.8084, 5.4325, 5.0627], device='cuda:0'), covar=tensor([0.0269, 0.0398, 0.0242, 0.0293, 0.0433, 0.0293, 0.0196, 0.0288], device='cuda:0'), in_proj_covar=tensor([0.0258, 0.0265, 0.0288, 0.0258, 0.0261, 0.0257, 0.0233, 0.0219], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 15:28:46,547 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=3.05 vs. limit=5.0 2023-05-16 15:28:49,592 INFO [finetune.py:992] (0/2) Epoch 10, batch 8400, loss[loss=0.1525, simple_loss=0.2505, pruned_loss=0.02721, over 12212.00 frames. ], tot_loss[loss=0.1726, simple_loss=0.261, pruned_loss=0.04207, over 2360042.61 frames. ], batch size: 35, lr: 4.01e-03, grad_scale: 8.0 2023-05-16 15:28:51,215 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.0990, 4.7737, 4.9370, 4.9683, 4.8078, 5.0353, 4.8459, 2.5556], device='cuda:0'), covar=tensor([0.0089, 0.0051, 0.0068, 0.0055, 0.0044, 0.0070, 0.0058, 0.0796], device='cuda:0'), in_proj_covar=tensor([0.0067, 0.0076, 0.0079, 0.0071, 0.0059, 0.0089, 0.0078, 0.0095], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-16 15:28:53,325 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=218509.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 15:29:14,020 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=218538.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 15:29:20,507 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.0937, 2.6068, 3.7204, 3.1692, 3.5511, 3.2354, 2.5864, 3.6354], device='cuda:0'), covar=tensor([0.0145, 0.0330, 0.0165, 0.0251, 0.0158, 0.0189, 0.0373, 0.0135], device='cuda:0'), in_proj_covar=tensor([0.0174, 0.0205, 0.0186, 0.0184, 0.0216, 0.0160, 0.0195, 0.0183], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 15:29:25,394 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=218554.0, num_to_drop=1, layers_to_drop={1} 2023-05-16 15:29:25,960 INFO [finetune.py:992] (0/2) Epoch 10, batch 8450, loss[loss=0.1851, simple_loss=0.2762, pruned_loss=0.04697, over 12140.00 frames. ], tot_loss[loss=0.1719, simple_loss=0.2608, pruned_loss=0.04153, over 2372233.16 frames. ], batch size: 39, lr: 4.01e-03, grad_scale: 8.0 2023-05-16 15:29:30,168 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.285e+02 2.848e+02 3.305e+02 3.959e+02 6.215e+02, threshold=6.610e+02, percent-clipped=0.0 2023-05-16 15:29:33,805 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=218566.0, num_to_drop=1, layers_to_drop={2} 2023-05-16 15:29:36,834 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=218570.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 15:29:37,735 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2023-05-16 15:29:39,743 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=218574.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 15:29:50,368 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.4204, 2.2938, 3.1501, 4.3086, 1.9372, 4.3299, 4.4530, 4.4498], device='cuda:0'), covar=tensor([0.0127, 0.1213, 0.0468, 0.0155, 0.1508, 0.0207, 0.0117, 0.0096], device='cuda:0'), in_proj_covar=tensor([0.0122, 0.0200, 0.0182, 0.0114, 0.0187, 0.0178, 0.0174, 0.0119], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 15:29:59,485 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.48 vs. limit=2.0 2023-05-16 15:29:59,954 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=218602.0, num_to_drop=1, layers_to_drop={1} 2023-05-16 15:30:01,932 INFO [finetune.py:992] (0/2) Epoch 10, batch 8500, loss[loss=0.1644, simple_loss=0.2371, pruned_loss=0.0458, over 12085.00 frames. ], tot_loss[loss=0.1715, simple_loss=0.2598, pruned_loss=0.04159, over 2365716.57 frames. ], batch size: 32, lr: 4.01e-03, grad_scale: 8.0 2023-05-16 15:30:24,132 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=218635.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 15:30:38,837 INFO [finetune.py:992] (0/2) Epoch 10, batch 8550, loss[loss=0.1393, simple_loss=0.2321, pruned_loss=0.02332, over 12343.00 frames. ], tot_loss[loss=0.1712, simple_loss=0.2595, pruned_loss=0.04142, over 2368842.50 frames. ], batch size: 31, lr: 4.01e-03, grad_scale: 8.0 2023-05-16 15:30:42,994 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.959e+02 2.716e+02 3.290e+02 3.941e+02 1.023e+03, threshold=6.580e+02, percent-clipped=1.0 2023-05-16 15:31:07,315 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.65 vs. limit=2.0 2023-05-16 15:31:14,701 INFO [finetune.py:992] (0/2) Epoch 10, batch 8600, loss[loss=0.1651, simple_loss=0.2492, pruned_loss=0.04054, over 12360.00 frames. ], tot_loss[loss=0.1715, simple_loss=0.2594, pruned_loss=0.04183, over 2361269.86 frames. ], batch size: 30, lr: 4.01e-03, grad_scale: 8.0 2023-05-16 15:31:26,489 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.62 vs. limit=2.0 2023-05-16 15:31:41,195 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2023-05-16 15:31:50,208 INFO [finetune.py:992] (0/2) Epoch 10, batch 8650, loss[loss=0.1661, simple_loss=0.2628, pruned_loss=0.03464, over 12116.00 frames. ], tot_loss[loss=0.1713, simple_loss=0.26, pruned_loss=0.04128, over 2370941.23 frames. ], batch size: 33, lr: 4.00e-03, grad_scale: 8.0 2023-05-16 15:31:51,065 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=218756.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 15:31:54,413 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.573e+02 2.817e+02 3.099e+02 3.578e+02 8.400e+02, threshold=6.197e+02, percent-clipped=5.0 2023-05-16 15:32:14,493 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.9727, 5.9090, 5.7354, 5.3155, 5.2083, 5.8606, 5.4771, 5.3063], device='cuda:0'), covar=tensor([0.0708, 0.0864, 0.0659, 0.1612, 0.0645, 0.0709, 0.1481, 0.0967], device='cuda:0'), in_proj_covar=tensor([0.0604, 0.0545, 0.0508, 0.0621, 0.0405, 0.0707, 0.0760, 0.0560], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0004, 0.0004, 0.0003], device='cuda:0') 2023-05-16 15:32:27,114 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=218804.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 15:32:27,786 INFO [finetune.py:992] (0/2) Epoch 10, batch 8700, loss[loss=0.1899, simple_loss=0.2808, pruned_loss=0.04949, over 11174.00 frames. ], tot_loss[loss=0.1706, simple_loss=0.2591, pruned_loss=0.04106, over 2375946.68 frames. ], batch size: 55, lr: 4.00e-03, grad_scale: 8.0 2023-05-16 15:32:30,234 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.1102, 4.5512, 4.0175, 4.8745, 4.4214, 2.9167, 4.2232, 3.0957], device='cuda:0'), covar=tensor([0.0908, 0.0794, 0.1454, 0.0451, 0.1179, 0.1631, 0.0978, 0.3034], device='cuda:0'), in_proj_covar=tensor([0.0307, 0.0381, 0.0357, 0.0301, 0.0366, 0.0268, 0.0348, 0.0364], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 15:32:51,439 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=218838.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 15:33:03,196 INFO [finetune.py:992] (0/2) Epoch 10, batch 8750, loss[loss=0.171, simple_loss=0.2609, pruned_loss=0.04057, over 10428.00 frames. ], tot_loss[loss=0.1699, simple_loss=0.2585, pruned_loss=0.04067, over 2381542.04 frames. ], batch size: 68, lr: 4.00e-03, grad_scale: 8.0 2023-05-16 15:33:07,470 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.025e+02 2.606e+02 3.159e+02 3.755e+02 6.243e+02, threshold=6.317e+02, percent-clipped=2.0 2023-05-16 15:33:10,407 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=218865.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 15:33:11,157 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=218866.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 15:33:25,167 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=218886.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 15:33:38,622 INFO [finetune.py:992] (0/2) Epoch 10, batch 8800, loss[loss=0.1856, simple_loss=0.2794, pruned_loss=0.04588, over 12082.00 frames. ], tot_loss[loss=0.1689, simple_loss=0.2577, pruned_loss=0.04001, over 2388847.50 frames. ], batch size: 32, lr: 4.00e-03, grad_scale: 8.0 2023-05-16 15:33:45,092 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=218914.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 15:33:56,849 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=218930.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 15:34:15,520 INFO [finetune.py:992] (0/2) Epoch 10, batch 8850, loss[loss=0.168, simple_loss=0.2604, pruned_loss=0.03778, over 12357.00 frames. ], tot_loss[loss=0.1701, simple_loss=0.2588, pruned_loss=0.04067, over 2387240.67 frames. ], batch size: 35, lr: 4.00e-03, grad_scale: 8.0 2023-05-16 15:34:19,768 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.900e+02 2.641e+02 3.077e+02 3.873e+02 7.892e+02, threshold=6.154e+02, percent-clipped=4.0 2023-05-16 15:34:22,922 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=218965.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 15:34:36,365 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=218984.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 15:34:39,250 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.3573, 5.1528, 5.3020, 5.3099, 4.9332, 4.9823, 4.7138, 5.2839], device='cuda:0'), covar=tensor([0.0622, 0.0587, 0.0764, 0.0617, 0.1888, 0.1272, 0.0593, 0.0929], device='cuda:0'), in_proj_covar=tensor([0.0526, 0.0691, 0.0595, 0.0619, 0.0831, 0.0729, 0.0541, 0.0464], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0005, 0.0004, 0.0004, 0.0005, 0.0005, 0.0003, 0.0003], device='cuda:0') 2023-05-16 15:34:51,339 INFO [finetune.py:992] (0/2) Epoch 10, batch 8900, loss[loss=0.1606, simple_loss=0.2361, pruned_loss=0.0425, over 11881.00 frames. ], tot_loss[loss=0.1698, simple_loss=0.2583, pruned_loss=0.04067, over 2388326.93 frames. ], batch size: 26, lr: 4.00e-03, grad_scale: 8.0 2023-05-16 15:35:03,195 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=219021.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 15:35:06,862 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=219026.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 15:35:20,422 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=219045.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 15:35:27,323 INFO [finetune.py:992] (0/2) Epoch 10, batch 8950, loss[loss=0.1618, simple_loss=0.2592, pruned_loss=0.03223, over 12268.00 frames. ], tot_loss[loss=0.1697, simple_loss=0.2583, pruned_loss=0.04059, over 2384737.72 frames. ], batch size: 37, lr: 4.00e-03, grad_scale: 8.0 2023-05-16 15:35:32,437 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.718e+02 2.720e+02 3.240e+02 3.849e+02 1.255e+03, threshold=6.481e+02, percent-clipped=2.0 2023-05-16 15:35:40,513 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=219072.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 15:35:47,694 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=219082.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 15:35:57,438 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.9786, 6.0164, 5.6917, 5.2182, 5.1891, 5.8231, 5.4706, 5.2383], device='cuda:0'), covar=tensor([0.0688, 0.0773, 0.0688, 0.1545, 0.0624, 0.0798, 0.1403, 0.1026], device='cuda:0'), in_proj_covar=tensor([0.0608, 0.0550, 0.0513, 0.0625, 0.0408, 0.0711, 0.0769, 0.0567], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0004, 0.0004, 0.0003], device='cuda:0') 2023-05-16 15:36:04,400 INFO [finetune.py:992] (0/2) Epoch 10, batch 9000, loss[loss=0.1497, simple_loss=0.2408, pruned_loss=0.02935, over 12180.00 frames. ], tot_loss[loss=0.1686, simple_loss=0.2572, pruned_loss=0.04001, over 2383894.26 frames. ], batch size: 31, lr: 4.00e-03, grad_scale: 8.0 2023-05-16 15:36:04,401 INFO [finetune.py:1017] (0/2) Computing validation loss 2023-05-16 15:36:11,316 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.1038, 5.0281, 4.9984, 4.9782, 4.5142, 5.1308, 5.1088, 5.1676], device='cuda:0'), covar=tensor([0.0217, 0.0133, 0.0181, 0.0298, 0.0768, 0.0312, 0.0139, 0.0192], device='cuda:0'), in_proj_covar=tensor([0.0189, 0.0191, 0.0184, 0.0241, 0.0235, 0.0213, 0.0168, 0.0217], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0005, 0.0004, 0.0004, 0.0003, 0.0005], device='cuda:0') 2023-05-16 15:36:23,645 INFO [finetune.py:1026] (0/2) Epoch 10, validation: loss=0.33, simple_loss=0.4019, pruned_loss=0.1291, over 1020973.00 frames. 2023-05-16 15:36:23,645 INFO [finetune.py:1027] (0/2) Maximum memory allocated so far is 12827MB 2023-05-16 15:36:43,971 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=219133.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 15:36:59,336 INFO [finetune.py:992] (0/2) Epoch 10, batch 9050, loss[loss=0.1512, simple_loss=0.2365, pruned_loss=0.03292, over 12112.00 frames. ], tot_loss[loss=0.1682, simple_loss=0.2569, pruned_loss=0.03974, over 2389587.26 frames. ], batch size: 30, lr: 4.00e-03, grad_scale: 8.0 2023-05-16 15:37:03,549 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.639e+02 2.910e+02 3.386e+02 3.882e+02 8.426e+02, threshold=6.771e+02, percent-clipped=1.0 2023-05-16 15:37:06,495 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=219165.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 15:37:20,796 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.97 vs. limit=2.0 2023-05-16 15:37:24,167 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.0090, 2.0604, 2.9156, 3.9820, 1.9329, 4.1324, 4.0650, 4.2191], device='cuda:0'), covar=tensor([0.0172, 0.1430, 0.0583, 0.0158, 0.1531, 0.0184, 0.0169, 0.0108], device='cuda:0'), in_proj_covar=tensor([0.0123, 0.0202, 0.0184, 0.0116, 0.0188, 0.0179, 0.0176, 0.0119], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 15:37:35,205 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=219203.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 15:37:36,352 INFO [finetune.py:992] (0/2) Epoch 10, batch 9100, loss[loss=0.2484, simple_loss=0.3138, pruned_loss=0.09146, over 8303.00 frames. ], tot_loss[loss=0.1674, simple_loss=0.2561, pruned_loss=0.03931, over 2393572.30 frames. ], batch size: 98, lr: 4.00e-03, grad_scale: 8.0 2023-05-16 15:37:41,546 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.8660, 5.6148, 5.1152, 5.2301, 5.6787, 5.0567, 5.2023, 5.2168], device='cuda:0'), covar=tensor([0.1505, 0.0965, 0.1084, 0.1873, 0.0976, 0.2341, 0.1921, 0.1085], device='cuda:0'), in_proj_covar=tensor([0.0366, 0.0496, 0.0400, 0.0448, 0.0472, 0.0450, 0.0401, 0.0387], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 15:37:42,239 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=219213.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 15:37:54,303 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=219230.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 15:38:11,906 INFO [finetune.py:992] (0/2) Epoch 10, batch 9150, loss[loss=0.1378, simple_loss=0.2192, pruned_loss=0.02825, over 12285.00 frames. ], tot_loss[loss=0.1673, simple_loss=0.2563, pruned_loss=0.03915, over 2395117.06 frames. ], batch size: 28, lr: 4.00e-03, grad_scale: 8.0 2023-05-16 15:38:16,099 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.802e+02 2.605e+02 3.021e+02 3.487e+02 6.301e+02, threshold=6.043e+02, percent-clipped=0.0 2023-05-16 15:38:18,408 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=219264.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 15:38:20,004 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 2023-05-16 15:38:28,243 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=219278.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 15:38:47,340 INFO [finetune.py:992] (0/2) Epoch 10, batch 9200, loss[loss=0.1699, simple_loss=0.262, pruned_loss=0.03893, over 11815.00 frames. ], tot_loss[loss=0.1672, simple_loss=0.2561, pruned_loss=0.03916, over 2389522.42 frames. ], batch size: 44, lr: 4.00e-03, grad_scale: 8.0 2023-05-16 15:38:59,227 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=219321.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 15:39:13,308 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=219340.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 15:39:14,882 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=219342.0, num_to_drop=1, layers_to_drop={0} 2023-05-16 15:39:16,691 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.85 vs. limit=2.0 2023-05-16 15:39:17,734 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.3455, 4.6836, 2.8626, 2.6414, 3.9050, 2.4596, 3.8944, 3.1384], device='cuda:0'), covar=tensor([0.0720, 0.0473, 0.1195, 0.1580, 0.0313, 0.1358, 0.0552, 0.0791], device='cuda:0'), in_proj_covar=tensor([0.0186, 0.0253, 0.0176, 0.0196, 0.0139, 0.0176, 0.0195, 0.0173], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 15:39:23,753 INFO [finetune.py:992] (0/2) Epoch 10, batch 9250, loss[loss=0.1916, simple_loss=0.2808, pruned_loss=0.05119, over 11126.00 frames. ], tot_loss[loss=0.1677, simple_loss=0.2568, pruned_loss=0.0393, over 2393634.18 frames. ], batch size: 55, lr: 4.00e-03, grad_scale: 8.0 2023-05-16 15:39:24,011 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.7212, 3.4519, 5.1355, 2.5612, 2.9052, 3.7228, 3.2094, 3.8386], device='cuda:0'), covar=tensor([0.0424, 0.1121, 0.0260, 0.1195, 0.1802, 0.1449, 0.1362, 0.1168], device='cuda:0'), in_proj_covar=tensor([0.0242, 0.0239, 0.0251, 0.0183, 0.0239, 0.0300, 0.0227, 0.0267], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-16 15:39:27,997 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.931e+02 2.725e+02 3.129e+02 3.820e+02 8.515e+02, threshold=6.257e+02, percent-clipped=3.0 2023-05-16 15:39:32,538 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.8795, 2.3945, 3.4305, 2.9060, 3.2681, 3.1022, 2.4084, 3.3768], device='cuda:0'), covar=tensor([0.0149, 0.0391, 0.0173, 0.0279, 0.0166, 0.0179, 0.0389, 0.0129], device='cuda:0'), in_proj_covar=tensor([0.0175, 0.0206, 0.0188, 0.0185, 0.0217, 0.0162, 0.0196, 0.0184], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 15:39:36,201 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.0778, 4.3626, 3.8230, 4.6558, 4.2718, 2.6821, 4.0573, 2.7848], device='cuda:0'), covar=tensor([0.0927, 0.0964, 0.1493, 0.0571, 0.1317, 0.1827, 0.1035, 0.3602], device='cuda:0'), in_proj_covar=tensor([0.0306, 0.0381, 0.0358, 0.0302, 0.0367, 0.0268, 0.0348, 0.0363], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 15:39:38,328 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.7757, 3.4209, 5.2107, 2.5606, 2.9579, 3.8946, 3.1788, 4.0644], device='cuda:0'), covar=tensor([0.0444, 0.1114, 0.0300, 0.1137, 0.1824, 0.1414, 0.1362, 0.0988], device='cuda:0'), in_proj_covar=tensor([0.0242, 0.0238, 0.0251, 0.0183, 0.0239, 0.0299, 0.0227, 0.0267], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-16 15:39:39,555 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=219377.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 15:39:56,918 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.4019, 3.9017, 3.8590, 4.3686, 3.2893, 4.0644, 2.5394, 4.3395], device='cuda:0'), covar=tensor([0.1333, 0.0804, 0.1343, 0.0802, 0.1001, 0.0509, 0.1744, 0.1092], device='cuda:0'), in_proj_covar=tensor([0.0224, 0.0263, 0.0294, 0.0352, 0.0234, 0.0236, 0.0255, 0.0374], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 15:39:58,326 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=219403.0, num_to_drop=1, layers_to_drop={1} 2023-05-16 15:39:59,516 INFO [finetune.py:992] (0/2) Epoch 10, batch 9300, loss[loss=0.152, simple_loss=0.2448, pruned_loss=0.02964, over 12118.00 frames. ], tot_loss[loss=0.1676, simple_loss=0.2567, pruned_loss=0.03925, over 2397552.43 frames. ], batch size: 33, lr: 4.00e-03, grad_scale: 8.0 2023-05-16 15:40:04,104 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.0618, 3.8759, 3.9647, 4.3226, 2.8492, 3.9506, 2.5186, 4.0698], device='cuda:0'), covar=tensor([0.1604, 0.0775, 0.0895, 0.0627, 0.1205, 0.0545, 0.1850, 0.1030], device='cuda:0'), in_proj_covar=tensor([0.0224, 0.0263, 0.0294, 0.0353, 0.0234, 0.0237, 0.0255, 0.0374], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 15:40:16,091 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=219428.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 15:40:34,930 INFO [finetune.py:992] (0/2) Epoch 10, batch 9350, loss[loss=0.1946, simple_loss=0.2895, pruned_loss=0.04988, over 11321.00 frames. ], tot_loss[loss=0.1684, simple_loss=0.2575, pruned_loss=0.0396, over 2398977.83 frames. ], batch size: 55, lr: 4.00e-03, grad_scale: 8.0 2023-05-16 15:40:39,261 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.859e+02 2.892e+02 3.242e+02 3.758e+02 1.096e+03, threshold=6.484e+02, percent-clipped=4.0 2023-05-16 15:40:55,941 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.86 vs. limit=2.0 2023-05-16 15:40:59,431 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2023-05-16 15:41:01,345 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=219490.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 15:41:11,718 INFO [finetune.py:992] (0/2) Epoch 10, batch 9400, loss[loss=0.1684, simple_loss=0.2543, pruned_loss=0.04122, over 12295.00 frames. ], tot_loss[loss=0.1693, simple_loss=0.2582, pruned_loss=0.0402, over 2391325.70 frames. ], batch size: 33, lr: 4.00e-03, grad_scale: 8.0 2023-05-16 15:41:12,739 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=3.07 vs. limit=5.0 2023-05-16 15:41:19,583 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.2191, 2.8781, 2.7949, 2.7676, 2.4922, 2.3378, 2.8444, 1.9622], device='cuda:0'), covar=tensor([0.0391, 0.0189, 0.0214, 0.0216, 0.0407, 0.0349, 0.0170, 0.0428], device='cuda:0'), in_proj_covar=tensor([0.0191, 0.0162, 0.0157, 0.0183, 0.0201, 0.0200, 0.0169, 0.0198], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 15:41:43,242 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.0067, 4.6460, 4.8630, 4.8945, 4.7294, 4.9689, 4.7998, 2.3817], device='cuda:0'), covar=tensor([0.0108, 0.0067, 0.0080, 0.0062, 0.0044, 0.0094, 0.0071, 0.0887], device='cuda:0'), in_proj_covar=tensor([0.0068, 0.0078, 0.0081, 0.0073, 0.0060, 0.0091, 0.0080, 0.0096], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-16 15:41:44,683 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=219551.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 15:41:47,288 INFO [finetune.py:992] (0/2) Epoch 10, batch 9450, loss[loss=0.1775, simple_loss=0.2666, pruned_loss=0.04418, over 10797.00 frames. ], tot_loss[loss=0.1694, simple_loss=0.258, pruned_loss=0.04037, over 2374751.21 frames. ], batch size: 68, lr: 4.00e-03, grad_scale: 16.0 2023-05-16 15:41:50,196 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=219559.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 15:41:51,494 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.827e+02 2.762e+02 3.070e+02 3.810e+02 6.920e+02, threshold=6.141e+02, percent-clipped=1.0 2023-05-16 15:42:23,221 INFO [finetune.py:992] (0/2) Epoch 10, batch 9500, loss[loss=0.1594, simple_loss=0.2473, pruned_loss=0.03577, over 12176.00 frames. ], tot_loss[loss=0.1688, simple_loss=0.2576, pruned_loss=0.04002, over 2380086.25 frames. ], batch size: 31, lr: 4.00e-03, grad_scale: 16.0 2023-05-16 15:42:34,780 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=219621.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 15:42:49,566 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=219640.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 15:43:00,007 INFO [finetune.py:992] (0/2) Epoch 10, batch 9550, loss[loss=0.2973, simple_loss=0.351, pruned_loss=0.1218, over 7550.00 frames. ], tot_loss[loss=0.1693, simple_loss=0.2578, pruned_loss=0.04042, over 2376393.64 frames. ], batch size: 97, lr: 4.00e-03, grad_scale: 16.0 2023-05-16 15:43:04,259 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.810e+02 2.626e+02 3.089e+02 3.885e+02 9.046e+02, threshold=6.179e+02, percent-clipped=3.0 2023-05-16 15:43:10,066 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=219669.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 15:43:15,644 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=219677.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 15:43:23,605 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=219688.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 15:43:30,540 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=219698.0, num_to_drop=1, layers_to_drop={1} 2023-05-16 15:43:35,477 INFO [finetune.py:992] (0/2) Epoch 10, batch 9600, loss[loss=0.2802, simple_loss=0.3386, pruned_loss=0.1109, over 7680.00 frames. ], tot_loss[loss=0.1704, simple_loss=0.2585, pruned_loss=0.04109, over 2366557.00 frames. ], batch size: 97, lr: 4.00e-03, grad_scale: 16.0 2023-05-16 15:43:49,948 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=219725.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 15:43:52,196 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=219728.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 15:43:56,433 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=2.43 vs. limit=5.0 2023-05-16 15:44:10,821 INFO [finetune.py:992] (0/2) Epoch 10, batch 9650, loss[loss=0.1887, simple_loss=0.2856, pruned_loss=0.04588, over 12346.00 frames. ], tot_loss[loss=0.1715, simple_loss=0.2597, pruned_loss=0.0417, over 2361489.28 frames. ], batch size: 35, lr: 3.99e-03, grad_scale: 16.0 2023-05-16 15:44:15,605 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.169e+02 2.820e+02 3.328e+02 3.898e+02 8.382e+02, threshold=6.656e+02, percent-clipped=4.0 2023-05-16 15:44:17,254 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=219763.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 15:44:27,231 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=219776.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 15:44:47,918 INFO [finetune.py:992] (0/2) Epoch 10, batch 9700, loss[loss=0.1754, simple_loss=0.271, pruned_loss=0.03992, over 12333.00 frames. ], tot_loss[loss=0.1708, simple_loss=0.2592, pruned_loss=0.04121, over 2368125.32 frames. ], batch size: 36, lr: 3.99e-03, grad_scale: 16.0 2023-05-16 15:44:50,188 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([6.0742, 6.0019, 5.8182, 5.2950, 5.2227, 5.9585, 5.5839, 5.3490], device='cuda:0'), covar=tensor([0.0746, 0.1010, 0.0653, 0.1462, 0.0611, 0.0777, 0.1535, 0.0971], device='cuda:0'), in_proj_covar=tensor([0.0611, 0.0555, 0.0515, 0.0627, 0.0410, 0.0714, 0.0770, 0.0568], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0004, 0.0004, 0.0003], device='cuda:0') 2023-05-16 15:45:01,497 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=219824.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 15:45:17,173 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=219846.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 15:45:20,784 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.5972, 2.7841, 3.7111, 4.4499, 3.8609, 4.4471, 3.8167, 3.2191], device='cuda:0'), covar=tensor([0.0030, 0.0355, 0.0148, 0.0037, 0.0110, 0.0083, 0.0112, 0.0342], device='cuda:0'), in_proj_covar=tensor([0.0086, 0.0124, 0.0107, 0.0077, 0.0104, 0.0117, 0.0097, 0.0140], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 15:45:23,391 INFO [finetune.py:992] (0/2) Epoch 10, batch 9750, loss[loss=0.1717, simple_loss=0.2586, pruned_loss=0.04241, over 12416.00 frames. ], tot_loss[loss=0.1711, simple_loss=0.2598, pruned_loss=0.04116, over 2374278.37 frames. ], batch size: 32, lr: 3.99e-03, grad_scale: 16.0 2023-05-16 15:45:24,933 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=219857.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 15:45:26,246 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=219859.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 15:45:27,478 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.060e+02 2.682e+02 3.236e+02 3.689e+02 1.131e+03, threshold=6.473e+02, percent-clipped=3.0 2023-05-16 15:45:58,393 INFO [finetune.py:992] (0/2) Epoch 10, batch 9800, loss[loss=0.1758, simple_loss=0.2671, pruned_loss=0.04221, over 12157.00 frames. ], tot_loss[loss=0.1717, simple_loss=0.26, pruned_loss=0.04164, over 2370482.25 frames. ], batch size: 34, lr: 3.99e-03, grad_scale: 16.0 2023-05-16 15:46:00,524 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=219907.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 15:46:09,217 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=219918.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 15:46:30,876 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.66 vs. limit=2.0 2023-05-16 15:46:35,404 INFO [finetune.py:992] (0/2) Epoch 10, batch 9850, loss[loss=0.1653, simple_loss=0.2533, pruned_loss=0.03868, over 12178.00 frames. ], tot_loss[loss=0.1706, simple_loss=0.2588, pruned_loss=0.04116, over 2363482.88 frames. ], batch size: 31, lr: 3.99e-03, grad_scale: 16.0 2023-05-16 15:46:39,494 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.091e+02 2.736e+02 3.238e+02 3.957e+02 6.493e+02, threshold=6.477e+02, percent-clipped=2.0 2023-05-16 15:46:57,488 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.0188, 4.6817, 5.0364, 4.4209, 4.7095, 4.4881, 5.0290, 4.6182], device='cuda:0'), covar=tensor([0.0276, 0.0359, 0.0252, 0.0255, 0.0325, 0.0320, 0.0204, 0.0501], device='cuda:0'), in_proj_covar=tensor([0.0259, 0.0264, 0.0288, 0.0260, 0.0261, 0.0259, 0.0234, 0.0220], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 15:47:02,084 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.91 vs. limit=2.0 2023-05-16 15:47:05,887 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=219998.0, num_to_drop=1, layers_to_drop={2} 2023-05-16 15:47:07,376 INFO [checkpoint.py:75] (0/2) Saving checkpoint to pruned_transducer_stateless7/exp_giga_finetune/checkpoint-120000.pt 2023-05-16 15:47:13,716 INFO [finetune.py:992] (0/2) Epoch 10, batch 9900, loss[loss=0.1729, simple_loss=0.2602, pruned_loss=0.04282, over 12087.00 frames. ], tot_loss[loss=0.171, simple_loss=0.2592, pruned_loss=0.04137, over 2362191.30 frames. ], batch size: 32, lr: 3.99e-03, grad_scale: 16.0 2023-05-16 15:47:25,914 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.8274, 3.5386, 5.3196, 2.7125, 2.9510, 3.9178, 3.3897, 4.0708], device='cuda:0'), covar=tensor([0.0481, 0.1040, 0.0226, 0.1158, 0.1764, 0.1441, 0.1229, 0.0988], device='cuda:0'), in_proj_covar=tensor([0.0242, 0.0237, 0.0249, 0.0183, 0.0238, 0.0297, 0.0226, 0.0265], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-16 15:47:42,713 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=220046.0, num_to_drop=1, layers_to_drop={1} 2023-05-16 15:47:44,206 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.8735, 4.5844, 4.7966, 4.7763, 4.6411, 4.8594, 4.7022, 2.4678], device='cuda:0'), covar=tensor([0.0173, 0.0104, 0.0122, 0.0102, 0.0068, 0.0126, 0.0105, 0.0977], device='cuda:0'), in_proj_covar=tensor([0.0068, 0.0078, 0.0081, 0.0073, 0.0060, 0.0091, 0.0080, 0.0097], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 15:47:49,598 INFO [finetune.py:992] (0/2) Epoch 10, batch 9950, loss[loss=0.1605, simple_loss=0.2577, pruned_loss=0.03164, over 11145.00 frames. ], tot_loss[loss=0.1703, simple_loss=0.2585, pruned_loss=0.04104, over 2377770.73 frames. ], batch size: 55, lr: 3.99e-03, grad_scale: 16.0 2023-05-16 15:47:54,581 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.799e+02 2.719e+02 3.158e+02 3.980e+02 6.647e+02, threshold=6.316e+02, percent-clipped=1.0 2023-05-16 15:48:13,044 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=220087.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 15:48:21,024 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=220098.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 15:48:25,764 INFO [finetune.py:992] (0/2) Epoch 10, batch 10000, loss[loss=0.2002, simple_loss=0.2795, pruned_loss=0.06047, over 8572.00 frames. ], tot_loss[loss=0.1704, simple_loss=0.2586, pruned_loss=0.04114, over 2382507.07 frames. ], batch size: 98, lr: 3.99e-03, grad_scale: 16.0 2023-05-16 15:48:26,579 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.1031, 5.0337, 4.9582, 5.0368, 3.9981, 5.1866, 5.1508, 5.3354], device='cuda:0'), covar=tensor([0.0208, 0.0162, 0.0193, 0.0312, 0.1206, 0.0318, 0.0157, 0.0192], device='cuda:0'), in_proj_covar=tensor([0.0193, 0.0195, 0.0187, 0.0247, 0.0240, 0.0218, 0.0171, 0.0223], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0005, 0.0004, 0.0005, 0.0003, 0.0005], device='cuda:0') 2023-05-16 15:48:29,577 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.7876, 3.1641, 4.7281, 4.8404, 2.9166, 2.8751, 3.0971, 2.1748], device='cuda:0'), covar=tensor([0.1430, 0.2480, 0.0403, 0.0366, 0.1293, 0.2176, 0.2422, 0.3788], device='cuda:0'), in_proj_covar=tensor([0.0288, 0.0380, 0.0271, 0.0297, 0.0266, 0.0299, 0.0373, 0.0367], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-16 15:48:35,831 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=220119.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 15:48:49,686 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.52 vs. limit=2.0 2023-05-16 15:48:55,130 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=220146.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 15:48:56,418 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=220148.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 15:49:01,133 INFO [finetune.py:992] (0/2) Epoch 10, batch 10050, loss[loss=0.1894, simple_loss=0.2799, pruned_loss=0.0494, over 12145.00 frames. ], tot_loss[loss=0.1694, simple_loss=0.2575, pruned_loss=0.04068, over 2385154.93 frames. ], batch size: 39, lr: 3.99e-03, grad_scale: 16.0 2023-05-16 15:49:01,524 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 2023-05-16 15:49:04,080 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=220159.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 15:49:05,272 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.261e+02 2.862e+02 3.244e+02 4.147e+02 1.035e+03, threshold=6.489e+02, percent-clipped=3.0 2023-05-16 15:49:29,534 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=220194.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 15:49:37,554 INFO [finetune.py:992] (0/2) Epoch 10, batch 10100, loss[loss=0.1829, simple_loss=0.2797, pruned_loss=0.043, over 12168.00 frames. ], tot_loss[loss=0.1698, simple_loss=0.2582, pruned_loss=0.04069, over 2381710.65 frames. ], batch size: 39, lr: 3.99e-03, grad_scale: 16.0 2023-05-16 15:49:43,901 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=220213.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 15:49:55,579 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=220229.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 15:50:10,275 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.98 vs. limit=2.0 2023-05-16 15:50:13,921 INFO [finetune.py:992] (0/2) Epoch 10, batch 10150, loss[loss=0.1817, simple_loss=0.272, pruned_loss=0.04573, over 11908.00 frames. ], tot_loss[loss=0.1696, simple_loss=0.2582, pruned_loss=0.04047, over 2379481.10 frames. ], batch size: 44, lr: 3.99e-03, grad_scale: 16.0 2023-05-16 15:50:16,284 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=220258.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 15:50:18,220 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.637e+02 2.689e+02 3.151e+02 3.706e+02 1.036e+03, threshold=6.302e+02, percent-clipped=2.0 2023-05-16 15:50:39,195 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=220290.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 15:50:49,864 INFO [finetune.py:992] (0/2) Epoch 10, batch 10200, loss[loss=0.1749, simple_loss=0.2603, pruned_loss=0.04475, over 12044.00 frames. ], tot_loss[loss=0.1697, simple_loss=0.2581, pruned_loss=0.04059, over 2380686.17 frames. ], batch size: 42, lr: 3.99e-03, grad_scale: 16.0 2023-05-16 15:50:59,978 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=220319.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 15:51:23,692 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2023-05-16 15:51:26,756 INFO [finetune.py:992] (0/2) Epoch 10, batch 10250, loss[loss=0.1634, simple_loss=0.2533, pruned_loss=0.03678, over 12284.00 frames. ], tot_loss[loss=0.1692, simple_loss=0.2579, pruned_loss=0.04028, over 2381693.26 frames. ], batch size: 33, lr: 3.99e-03, grad_scale: 16.0 2023-05-16 15:51:31,013 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.808e+02 2.634e+02 2.920e+02 3.635e+02 1.121e+03, threshold=5.839e+02, percent-clipped=1.0 2023-05-16 15:51:41,283 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=4.52 vs. limit=5.0 2023-05-16 15:51:42,990 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.2151, 5.1495, 5.0666, 5.0152, 4.6885, 5.2447, 5.1285, 5.3693], device='cuda:0'), covar=tensor([0.0261, 0.0142, 0.0201, 0.0364, 0.0810, 0.0361, 0.0160, 0.0184], device='cuda:0'), in_proj_covar=tensor([0.0189, 0.0192, 0.0184, 0.0243, 0.0237, 0.0215, 0.0168, 0.0220], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0005, 0.0004, 0.0004, 0.0003, 0.0005], device='cuda:0') 2023-05-16 15:51:43,051 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.6488, 3.7480, 3.3410, 3.3816, 3.0869, 2.8815, 3.8478, 2.3917], device='cuda:0'), covar=tensor([0.0368, 0.0134, 0.0210, 0.0187, 0.0366, 0.0372, 0.0101, 0.0439], device='cuda:0'), in_proj_covar=tensor([0.0192, 0.0163, 0.0160, 0.0185, 0.0204, 0.0202, 0.0170, 0.0201], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 15:52:02,288 INFO [finetune.py:992] (0/2) Epoch 10, batch 10300, loss[loss=0.1772, simple_loss=0.2719, pruned_loss=0.04128, over 12139.00 frames. ], tot_loss[loss=0.1691, simple_loss=0.2577, pruned_loss=0.04022, over 2367779.96 frames. ], batch size: 36, lr: 3.99e-03, grad_scale: 16.0 2023-05-16 15:52:12,253 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=220419.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 15:52:14,995 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.8621, 4.5592, 4.1437, 4.0467, 4.6504, 3.9253, 4.2557, 3.9935], device='cuda:0'), covar=tensor([0.1690, 0.1161, 0.1458, 0.2261, 0.0994, 0.2422, 0.1528, 0.1577], device='cuda:0'), in_proj_covar=tensor([0.0360, 0.0489, 0.0397, 0.0444, 0.0463, 0.0437, 0.0395, 0.0380], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 15:52:21,596 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.0917, 3.9443, 3.9904, 4.3419, 2.9900, 3.8497, 2.5038, 4.0399], device='cuda:0'), covar=tensor([0.1644, 0.0744, 0.0991, 0.0627, 0.1115, 0.0636, 0.1896, 0.1070], device='cuda:0'), in_proj_covar=tensor([0.0226, 0.0265, 0.0298, 0.0356, 0.0237, 0.0240, 0.0262, 0.0379], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 15:52:26,562 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.4796, 2.4029, 3.5965, 4.3289, 3.8392, 4.2843, 3.6900, 3.2065], device='cuda:0'), covar=tensor([0.0039, 0.0420, 0.0160, 0.0051, 0.0124, 0.0085, 0.0158, 0.0318], device='cuda:0'), in_proj_covar=tensor([0.0085, 0.0123, 0.0107, 0.0077, 0.0102, 0.0116, 0.0096, 0.0138], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 15:52:29,308 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=220443.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 15:52:37,220 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=220454.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 15:52:37,776 INFO [finetune.py:992] (0/2) Epoch 10, batch 10350, loss[loss=0.1692, simple_loss=0.2592, pruned_loss=0.03958, over 12193.00 frames. ], tot_loss[loss=0.1693, simple_loss=0.258, pruned_loss=0.04033, over 2370108.79 frames. ], batch size: 35, lr: 3.99e-03, grad_scale: 16.0 2023-05-16 15:52:41,952 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.058e+02 2.853e+02 3.399e+02 3.948e+02 7.256e+02, threshold=6.798e+02, percent-clipped=6.0 2023-05-16 15:52:46,219 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=220467.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 15:53:14,397 INFO [finetune.py:992] (0/2) Epoch 10, batch 10400, loss[loss=0.1581, simple_loss=0.2543, pruned_loss=0.03092, over 11910.00 frames. ], tot_loss[loss=0.1691, simple_loss=0.2579, pruned_loss=0.04017, over 2371291.28 frames. ], batch size: 44, lr: 3.99e-03, grad_scale: 16.0 2023-05-16 15:53:20,264 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=220513.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 15:53:21,119 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.8447, 2.3596, 3.3806, 2.8916, 3.1732, 3.0263, 2.3849, 3.2647], device='cuda:0'), covar=tensor([0.0134, 0.0357, 0.0179, 0.0226, 0.0153, 0.0183, 0.0329, 0.0139], device='cuda:0'), in_proj_covar=tensor([0.0176, 0.0207, 0.0189, 0.0187, 0.0219, 0.0163, 0.0199, 0.0187], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 15:53:24,641 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.0078, 3.1603, 4.4107, 2.4334, 2.5887, 3.3556, 3.0054, 3.5442], device='cuda:0'), covar=tensor([0.0536, 0.1089, 0.0326, 0.1173, 0.1904, 0.1374, 0.1301, 0.0961], device='cuda:0'), in_proj_covar=tensor([0.0239, 0.0234, 0.0248, 0.0181, 0.0236, 0.0294, 0.0223, 0.0262], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-16 15:53:27,612 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 2023-05-16 15:53:32,666 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.42 vs. limit=2.0 2023-05-16 15:53:35,195 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.5228, 2.4003, 3.6134, 4.3797, 3.8565, 4.2623, 3.7593, 2.8263], device='cuda:0'), covar=tensor([0.0033, 0.0394, 0.0148, 0.0041, 0.0122, 0.0100, 0.0126, 0.0376], device='cuda:0'), in_proj_covar=tensor([0.0086, 0.0124, 0.0107, 0.0077, 0.0103, 0.0116, 0.0096, 0.0139], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 15:53:49,503 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.4835, 4.1844, 4.4146, 4.4092, 4.2726, 4.4521, 4.2443, 2.4305], device='cuda:0'), covar=tensor([0.0228, 0.0150, 0.0160, 0.0136, 0.0101, 0.0199, 0.0242, 0.1127], device='cuda:0'), in_proj_covar=tensor([0.0070, 0.0080, 0.0083, 0.0074, 0.0061, 0.0093, 0.0082, 0.0099], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 15:53:50,007 INFO [finetune.py:992] (0/2) Epoch 10, batch 10450, loss[loss=0.1592, simple_loss=0.2499, pruned_loss=0.03425, over 11798.00 frames. ], tot_loss[loss=0.169, simple_loss=0.2582, pruned_loss=0.03995, over 2372961.87 frames. ], batch size: 44, lr: 3.99e-03, grad_scale: 16.0 2023-05-16 15:53:50,232 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=220555.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 15:53:54,257 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.943e+02 2.943e+02 3.296e+02 3.873e+02 5.866e+02, threshold=6.592e+02, percent-clipped=0.0 2023-05-16 15:53:54,352 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=220561.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 15:54:11,556 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=220585.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 15:54:16,710 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.9998, 4.6225, 4.7978, 4.8416, 4.6694, 4.8963, 4.8154, 2.4314], device='cuda:0'), covar=tensor([0.0108, 0.0070, 0.0086, 0.0065, 0.0053, 0.0092, 0.0079, 0.0946], device='cuda:0'), in_proj_covar=tensor([0.0070, 0.0080, 0.0083, 0.0075, 0.0061, 0.0093, 0.0082, 0.0099], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 15:54:17,508 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.8130, 3.4733, 5.1067, 2.6928, 2.7375, 3.8151, 3.2442, 3.8331], device='cuda:0'), covar=tensor([0.0430, 0.1093, 0.0344, 0.1126, 0.1945, 0.1465, 0.1300, 0.1125], device='cuda:0'), in_proj_covar=tensor([0.0241, 0.0236, 0.0250, 0.0183, 0.0237, 0.0296, 0.0225, 0.0264], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-16 15:54:25,953 INFO [finetune.py:992] (0/2) Epoch 10, batch 10500, loss[loss=0.1988, simple_loss=0.2808, pruned_loss=0.05844, over 12157.00 frames. ], tot_loss[loss=0.1683, simple_loss=0.2572, pruned_loss=0.03971, over 2377237.79 frames. ], batch size: 39, lr: 3.99e-03, grad_scale: 16.0 2023-05-16 15:54:32,432 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=220614.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 15:54:34,008 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=220616.0, num_to_drop=1, layers_to_drop={0} 2023-05-16 15:55:00,108 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=220651.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 15:55:02,704 INFO [finetune.py:992] (0/2) Epoch 10, batch 10550, loss[loss=0.1531, simple_loss=0.2323, pruned_loss=0.03699, over 12298.00 frames. ], tot_loss[loss=0.1682, simple_loss=0.2571, pruned_loss=0.03969, over 2380628.68 frames. ], batch size: 28, lr: 3.99e-03, grad_scale: 16.0 2023-05-16 15:55:04,964 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.1267, 3.9283, 4.0113, 4.4835, 3.3486, 4.0014, 2.6254, 4.1587], device='cuda:0'), covar=tensor([0.1636, 0.0802, 0.1033, 0.0610, 0.1005, 0.0592, 0.1775, 0.1120], device='cuda:0'), in_proj_covar=tensor([0.0226, 0.0264, 0.0298, 0.0355, 0.0236, 0.0239, 0.0261, 0.0378], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 15:55:06,781 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.959e+02 2.732e+02 3.114e+02 3.768e+02 7.868e+02, threshold=6.228e+02, percent-clipped=1.0 2023-05-16 15:55:28,326 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.1342, 4.7273, 4.1864, 4.9407, 4.4929, 2.8145, 4.3521, 3.0610], device='cuda:0'), covar=tensor([0.0881, 0.0683, 0.1229, 0.0416, 0.1077, 0.1607, 0.0917, 0.3072], device='cuda:0'), in_proj_covar=tensor([0.0303, 0.0376, 0.0354, 0.0299, 0.0365, 0.0266, 0.0344, 0.0364], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 15:55:37,819 INFO [finetune.py:992] (0/2) Epoch 10, batch 10600, loss[loss=0.1572, simple_loss=0.2467, pruned_loss=0.03387, over 12179.00 frames. ], tot_loss[loss=0.169, simple_loss=0.2576, pruned_loss=0.04018, over 2372232.69 frames. ], batch size: 31, lr: 3.99e-03, grad_scale: 16.0 2023-05-16 15:55:43,122 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=220712.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 15:55:51,102 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=4.95 vs. limit=5.0 2023-05-16 15:56:04,971 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=220743.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 15:56:12,724 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=220754.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 15:56:13,262 INFO [finetune.py:992] (0/2) Epoch 10, batch 10650, loss[loss=0.1581, simple_loss=0.2481, pruned_loss=0.03409, over 12185.00 frames. ], tot_loss[loss=0.1694, simple_loss=0.2581, pruned_loss=0.04032, over 2373688.98 frames. ], batch size: 31, lr: 3.99e-03, grad_scale: 16.0 2023-05-16 15:56:17,465 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.747e+02 2.734e+02 3.095e+02 3.683e+02 6.434e+02, threshold=6.190e+02, percent-clipped=1.0 2023-05-16 15:56:39,634 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=220791.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 15:56:42,775 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=2.52 vs. limit=5.0 2023-05-16 15:56:48,416 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=220802.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 15:56:50,481 INFO [finetune.py:992] (0/2) Epoch 10, batch 10700, loss[loss=0.163, simple_loss=0.2508, pruned_loss=0.03762, over 12292.00 frames. ], tot_loss[loss=0.1699, simple_loss=0.2584, pruned_loss=0.0407, over 2367848.87 frames. ], batch size: 33, lr: 3.98e-03, grad_scale: 16.0 2023-05-16 15:57:02,791 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=220822.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 15:57:26,106 INFO [finetune.py:992] (0/2) Epoch 10, batch 10750, loss[loss=0.1638, simple_loss=0.2596, pruned_loss=0.03405, over 10633.00 frames. ], tot_loss[loss=0.1703, simple_loss=0.2587, pruned_loss=0.04095, over 2366792.48 frames. ], batch size: 68, lr: 3.98e-03, grad_scale: 16.0 2023-05-16 15:57:27,739 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.5068, 2.3598, 3.6918, 4.4288, 3.9696, 4.3038, 3.7621, 3.1881], device='cuda:0'), covar=tensor([0.0038, 0.0409, 0.0133, 0.0034, 0.0094, 0.0082, 0.0116, 0.0303], device='cuda:0'), in_proj_covar=tensor([0.0086, 0.0124, 0.0108, 0.0077, 0.0102, 0.0116, 0.0096, 0.0138], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 15:57:30,320 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.656e+02 2.819e+02 3.336e+02 4.154e+02 5.728e+02, threshold=6.671e+02, percent-clipped=0.0 2023-05-16 15:57:46,254 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=220883.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 15:57:47,537 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=220885.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 15:58:00,834 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=220904.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 15:58:01,328 INFO [finetune.py:992] (0/2) Epoch 10, batch 10800, loss[loss=0.1728, simple_loss=0.2659, pruned_loss=0.03987, over 11289.00 frames. ], tot_loss[loss=0.1711, simple_loss=0.2595, pruned_loss=0.04131, over 2353221.09 frames. ], batch size: 55, lr: 3.98e-03, grad_scale: 16.0 2023-05-16 15:58:05,598 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=220911.0, num_to_drop=1, layers_to_drop={0} 2023-05-16 15:58:07,557 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=220914.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 15:58:18,373 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.3087, 4.6422, 2.8234, 2.7400, 3.9684, 2.6329, 4.0175, 3.2735], device='cuda:0'), covar=tensor([0.0720, 0.0587, 0.1216, 0.1496, 0.0338, 0.1280, 0.0478, 0.0780], device='cuda:0'), in_proj_covar=tensor([0.0191, 0.0259, 0.0180, 0.0200, 0.0141, 0.0181, 0.0198, 0.0177], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 15:58:21,679 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=220933.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 15:58:33,385 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.1877, 5.0317, 4.9170, 4.9999, 4.6460, 5.0280, 5.0396, 5.2808], device='cuda:0'), covar=tensor([0.0210, 0.0137, 0.0185, 0.0307, 0.0738, 0.0353, 0.0159, 0.0158], device='cuda:0'), in_proj_covar=tensor([0.0190, 0.0193, 0.0187, 0.0244, 0.0239, 0.0215, 0.0169, 0.0222], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0005, 0.0004, 0.0004, 0.0003, 0.0005], device='cuda:0') 2023-05-16 15:58:38,197 INFO [finetune.py:992] (0/2) Epoch 10, batch 10850, loss[loss=0.1783, simple_loss=0.2665, pruned_loss=0.04508, over 12059.00 frames. ], tot_loss[loss=0.1702, simple_loss=0.2586, pruned_loss=0.04094, over 2368148.01 frames. ], batch size: 42, lr: 3.98e-03, grad_scale: 16.0 2023-05-16 15:58:42,546 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.475e+02 2.730e+02 3.365e+02 4.016e+02 7.164e+02, threshold=6.730e+02, percent-clipped=2.0 2023-05-16 15:58:43,404 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=220962.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 15:58:45,637 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=220965.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 15:59:15,179 INFO [finetune.py:992] (0/2) Epoch 10, batch 10900, loss[loss=0.1876, simple_loss=0.2746, pruned_loss=0.05032, over 12053.00 frames. ], tot_loss[loss=0.1691, simple_loss=0.2574, pruned_loss=0.04042, over 2375818.18 frames. ], batch size: 37, lr: 3.98e-03, grad_scale: 16.0 2023-05-16 15:59:16,751 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=221007.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 15:59:43,187 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.4292, 2.3382, 3.1828, 4.3384, 2.1237, 4.4155, 4.5114, 4.4937], device='cuda:0'), covar=tensor([0.0144, 0.1343, 0.0491, 0.0176, 0.1552, 0.0192, 0.0127, 0.0102], device='cuda:0'), in_proj_covar=tensor([0.0122, 0.0201, 0.0183, 0.0117, 0.0187, 0.0179, 0.0174, 0.0119], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 15:59:49,661 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=4.36 vs. limit=5.0 2023-05-16 15:59:50,674 INFO [finetune.py:992] (0/2) Epoch 10, batch 10950, loss[loss=0.2027, simple_loss=0.2927, pruned_loss=0.05636, over 10670.00 frames. ], tot_loss[loss=0.1702, simple_loss=0.2587, pruned_loss=0.04083, over 2370119.62 frames. ], batch size: 68, lr: 3.98e-03, grad_scale: 16.0 2023-05-16 15:59:54,856 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.948e+02 2.682e+02 3.296e+02 4.099e+02 7.687e+02, threshold=6.592e+02, percent-clipped=1.0 2023-05-16 16:00:17,737 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.9223, 3.1567, 4.4264, 2.4768, 2.5603, 3.4777, 3.0125, 3.5907], device='cuda:0'), covar=tensor([0.0603, 0.1132, 0.0258, 0.1231, 0.1950, 0.1278, 0.1428, 0.0956], device='cuda:0'), in_proj_covar=tensor([0.0241, 0.0236, 0.0249, 0.0183, 0.0237, 0.0295, 0.0225, 0.0264], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-16 16:00:27,507 INFO [finetune.py:992] (0/2) Epoch 10, batch 11000, loss[loss=0.1825, simple_loss=0.2696, pruned_loss=0.04775, over 12046.00 frames. ], tot_loss[loss=0.1729, simple_loss=0.2612, pruned_loss=0.04227, over 2350156.59 frames. ], batch size: 37, lr: 3.98e-03, grad_scale: 16.0 2023-05-16 16:00:40,999 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.4535, 5.2896, 5.4337, 5.4384, 5.0463, 5.0795, 4.9556, 5.2479], device='cuda:0'), covar=tensor([0.0633, 0.0593, 0.0741, 0.0548, 0.1941, 0.1427, 0.0516, 0.1410], device='cuda:0'), in_proj_covar=tensor([0.0529, 0.0694, 0.0591, 0.0623, 0.0837, 0.0731, 0.0541, 0.0472], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0005, 0.0004, 0.0004, 0.0005, 0.0005, 0.0003, 0.0003], device='cuda:0') 2023-05-16 16:01:02,730 INFO [finetune.py:992] (0/2) Epoch 10, batch 11050, loss[loss=0.1569, simple_loss=0.2379, pruned_loss=0.038, over 12350.00 frames. ], tot_loss[loss=0.175, simple_loss=0.263, pruned_loss=0.04354, over 2324746.35 frames. ], batch size: 31, lr: 3.98e-03, grad_scale: 16.0 2023-05-16 16:01:07,035 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.894e+02 3.083e+02 3.628e+02 4.356e+02 1.327e+03, threshold=7.256e+02, percent-clipped=10.0 2023-05-16 16:01:19,863 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=221178.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 16:01:39,602 INFO [finetune.py:992] (0/2) Epoch 10, batch 11100, loss[loss=0.1784, simple_loss=0.2747, pruned_loss=0.04104, over 12287.00 frames. ], tot_loss[loss=0.1822, simple_loss=0.2694, pruned_loss=0.04751, over 2267174.87 frames. ], batch size: 33, lr: 3.98e-03, grad_scale: 16.0 2023-05-16 16:01:44,035 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=221211.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 16:01:47,193 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=4.94 vs. limit=5.0 2023-05-16 16:02:14,507 INFO [finetune.py:992] (0/2) Epoch 10, batch 11150, loss[loss=0.2034, simple_loss=0.293, pruned_loss=0.05691, over 12028.00 frames. ], tot_loss[loss=0.1892, simple_loss=0.2758, pruned_loss=0.05131, over 2217195.86 frames. ], batch size: 40, lr: 3.98e-03, grad_scale: 16.0 2023-05-16 16:02:18,159 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=221259.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 16:02:18,921 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=221260.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 16:02:19,491 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.155e+02 3.250e+02 3.791e+02 4.612e+02 9.960e+02, threshold=7.583e+02, percent-clipped=2.0 2023-05-16 16:02:29,234 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.8161, 3.7037, 3.8222, 3.5367, 3.6644, 3.5095, 3.8209, 3.5130], device='cuda:0'), covar=tensor([0.0421, 0.0337, 0.0316, 0.0264, 0.0381, 0.0318, 0.0314, 0.1273], device='cuda:0'), in_proj_covar=tensor([0.0255, 0.0260, 0.0284, 0.0257, 0.0258, 0.0255, 0.0230, 0.0217], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 16:02:50,808 INFO [finetune.py:992] (0/2) Epoch 10, batch 11200, loss[loss=0.2309, simple_loss=0.3161, pruned_loss=0.07284, over 11241.00 frames. ], tot_loss[loss=0.1961, simple_loss=0.2817, pruned_loss=0.05528, over 2153500.60 frames. ], batch size: 55, lr: 3.98e-03, grad_scale: 16.0 2023-05-16 16:02:52,279 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=221307.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 16:03:26,527 INFO [finetune.py:992] (0/2) Epoch 10, batch 11250, loss[loss=0.2681, simple_loss=0.3331, pruned_loss=0.1016, over 7450.00 frames. ], tot_loss[loss=0.2055, simple_loss=0.2899, pruned_loss=0.06054, over 2078191.58 frames. ], batch size: 98, lr: 3.98e-03, grad_scale: 16.0 2023-05-16 16:03:26,605 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=221355.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 16:03:30,562 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.422e+02 3.396e+02 4.025e+02 5.294e+02 9.338e+02, threshold=8.049e+02, percent-clipped=5.0 2023-05-16 16:03:39,821 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.8478, 2.2122, 2.7520, 2.6544, 3.0116, 2.9476, 2.9436, 2.3384], device='cuda:0'), covar=tensor([0.0072, 0.0349, 0.0193, 0.0112, 0.0105, 0.0102, 0.0115, 0.0368], device='cuda:0'), in_proj_covar=tensor([0.0085, 0.0122, 0.0106, 0.0076, 0.0100, 0.0115, 0.0095, 0.0137], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 16:04:00,876 INFO [finetune.py:992] (0/2) Epoch 10, batch 11300, loss[loss=0.1951, simple_loss=0.2847, pruned_loss=0.05278, over 10362.00 frames. ], tot_loss[loss=0.21, simple_loss=0.2942, pruned_loss=0.06293, over 2051698.80 frames. ], batch size: 68, lr: 3.98e-03, grad_scale: 16.0 2023-05-16 16:04:16,201 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.4996, 5.1339, 5.4735, 4.8722, 5.1021, 4.9930, 5.5186, 5.1695], device='cuda:0'), covar=tensor([0.0221, 0.0325, 0.0254, 0.0222, 0.0366, 0.0250, 0.0188, 0.0241], device='cuda:0'), in_proj_covar=tensor([0.0249, 0.0254, 0.0276, 0.0250, 0.0251, 0.0249, 0.0225, 0.0211], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 16:04:36,476 INFO [finetune.py:992] (0/2) Epoch 10, batch 11350, loss[loss=0.2808, simple_loss=0.3402, pruned_loss=0.1107, over 6996.00 frames. ], tot_loss[loss=0.2161, simple_loss=0.2993, pruned_loss=0.06647, over 1985186.86 frames. ], batch size: 98, lr: 3.98e-03, grad_scale: 16.0 2023-05-16 16:04:40,616 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.540e+02 3.534e+02 4.230e+02 4.906e+02 1.074e+03, threshold=8.461e+02, percent-clipped=2.0 2023-05-16 16:04:52,776 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=221478.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 16:05:11,508 INFO [finetune.py:992] (0/2) Epoch 10, batch 11400, loss[loss=0.2489, simple_loss=0.3285, pruned_loss=0.08466, over 12016.00 frames. ], tot_loss[loss=0.2204, simple_loss=0.3031, pruned_loss=0.06887, over 1948772.48 frames. ], batch size: 40, lr: 3.98e-03, grad_scale: 16.0 2023-05-16 16:05:13,031 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=221507.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 16:05:25,932 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=221526.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 16:05:42,141 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.5255, 4.9487, 3.1204, 2.8745, 4.3240, 2.7593, 4.2530, 3.5013], device='cuda:0'), covar=tensor([0.0614, 0.0351, 0.1078, 0.1479, 0.0185, 0.1266, 0.0378, 0.0731], device='cuda:0'), in_proj_covar=tensor([0.0186, 0.0249, 0.0175, 0.0196, 0.0137, 0.0177, 0.0193, 0.0173], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 16:05:46,542 INFO [finetune.py:992] (0/2) Epoch 10, batch 11450, loss[loss=0.2692, simple_loss=0.3308, pruned_loss=0.1038, over 6996.00 frames. ], tot_loss[loss=0.226, simple_loss=0.3071, pruned_loss=0.07238, over 1890527.97 frames. ], batch size: 98, lr: 3.98e-03, grad_scale: 32.0 2023-05-16 16:05:50,210 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=221560.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 16:05:50,662 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.263e+02 3.595e+02 4.162e+02 4.999e+02 1.118e+03, threshold=8.323e+02, percent-clipped=4.0 2023-05-16 16:05:55,859 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=221568.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 16:06:02,581 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.4012, 3.1473, 2.9979, 3.3258, 2.6420, 3.1486, 2.5406, 2.7397], device='cuda:0'), covar=tensor([0.1397, 0.0794, 0.0744, 0.0420, 0.0989, 0.0675, 0.1547, 0.0577], device='cuda:0'), in_proj_covar=tensor([0.0223, 0.0261, 0.0291, 0.0346, 0.0233, 0.0235, 0.0255, 0.0370], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 16:06:11,223 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.7750, 3.0576, 2.3619, 2.1570, 2.7645, 2.2214, 2.9663, 2.5924], device='cuda:0'), covar=tensor([0.0573, 0.0476, 0.0881, 0.1478, 0.0258, 0.1137, 0.0454, 0.0728], device='cuda:0'), in_proj_covar=tensor([0.0187, 0.0249, 0.0175, 0.0197, 0.0137, 0.0178, 0.0193, 0.0174], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 16:06:21,253 INFO [finetune.py:992] (0/2) Epoch 10, batch 11500, loss[loss=0.3213, simple_loss=0.3577, pruned_loss=0.1424, over 6646.00 frames. ], tot_loss[loss=0.2305, simple_loss=0.3104, pruned_loss=0.07529, over 1837551.78 frames. ], batch size: 100, lr: 3.98e-03, grad_scale: 16.0 2023-05-16 16:06:24,003 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=221608.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 16:06:25,500 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.9967, 3.0608, 4.4822, 2.4618, 2.5233, 3.4920, 2.9391, 3.5761], device='cuda:0'), covar=tensor([0.0536, 0.1202, 0.0205, 0.1289, 0.1970, 0.1150, 0.1405, 0.0873], device='cuda:0'), in_proj_covar=tensor([0.0231, 0.0228, 0.0238, 0.0177, 0.0228, 0.0284, 0.0217, 0.0252], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:0') 2023-05-16 16:06:33,702 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.59 vs. limit=2.0 2023-05-16 16:06:36,336 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.91 vs. limit=2.0 2023-05-16 16:06:40,182 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.7681, 3.1066, 2.3767, 2.2643, 2.8480, 2.2573, 3.0180, 2.6079], device='cuda:0'), covar=tensor([0.0651, 0.0461, 0.0929, 0.1428, 0.0284, 0.1168, 0.0525, 0.0787], device='cuda:0'), in_proj_covar=tensor([0.0186, 0.0248, 0.0174, 0.0197, 0.0137, 0.0178, 0.0192, 0.0174], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 16:06:51,282 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.1357, 1.9351, 2.2783, 2.0825, 2.1307, 2.3115, 1.7951, 2.2402], device='cuda:0'), covar=tensor([0.0095, 0.0298, 0.0118, 0.0205, 0.0128, 0.0144, 0.0314, 0.0115], device='cuda:0'), in_proj_covar=tensor([0.0167, 0.0196, 0.0177, 0.0177, 0.0204, 0.0154, 0.0187, 0.0175], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 16:06:55,868 INFO [finetune.py:992] (0/2) Epoch 10, batch 11550, loss[loss=0.2717, simple_loss=0.3321, pruned_loss=0.1057, over 6974.00 frames. ], tot_loss[loss=0.2329, simple_loss=0.3124, pruned_loss=0.07668, over 1829667.21 frames. ], batch size: 98, lr: 3.98e-03, grad_scale: 16.0 2023-05-16 16:07:00,387 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.705e+02 3.400e+02 4.212e+02 4.832e+02 8.048e+02, threshold=8.424e+02, percent-clipped=0.0 2023-05-16 16:07:30,291 INFO [finetune.py:992] (0/2) Epoch 10, batch 11600, loss[loss=0.2499, simple_loss=0.3189, pruned_loss=0.09047, over 6491.00 frames. ], tot_loss[loss=0.2348, simple_loss=0.3132, pruned_loss=0.07823, over 1793670.13 frames. ], batch size: 98, lr: 3.98e-03, grad_scale: 16.0 2023-05-16 16:07:36,813 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.3645, 2.9642, 4.9281, 2.6520, 2.5877, 3.9151, 3.0242, 3.8651], device='cuda:0'), covar=tensor([0.0601, 0.1540, 0.0214, 0.1356, 0.2221, 0.1307, 0.1702, 0.1045], device='cuda:0'), in_proj_covar=tensor([0.0229, 0.0226, 0.0235, 0.0176, 0.0227, 0.0282, 0.0215, 0.0250], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:0') 2023-05-16 16:08:07,789 INFO [finetune.py:992] (0/2) Epoch 10, batch 11650, loss[loss=0.2828, simple_loss=0.3482, pruned_loss=0.1086, over 7084.00 frames. ], tot_loss[loss=0.2353, simple_loss=0.3128, pruned_loss=0.07889, over 1777687.29 frames. ], batch size: 98, lr: 3.98e-03, grad_scale: 16.0 2023-05-16 16:08:10,144 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=221758.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 16:08:12,826 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.437e+02 3.386e+02 3.820e+02 4.524e+02 6.802e+02, threshold=7.640e+02, percent-clipped=0.0 2023-05-16 16:08:33,923 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.5308, 4.4928, 4.4092, 4.1082, 4.1878, 4.5145, 4.2466, 4.0948], device='cuda:0'), covar=tensor([0.0718, 0.0916, 0.0636, 0.1296, 0.1814, 0.0800, 0.1437, 0.1098], device='cuda:0'), in_proj_covar=tensor([0.0584, 0.0529, 0.0489, 0.0597, 0.0393, 0.0676, 0.0725, 0.0541], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 16:08:43,804 INFO [finetune.py:992] (0/2) Epoch 10, batch 11700, loss[loss=0.2218, simple_loss=0.3075, pruned_loss=0.06809, over 10314.00 frames. ], tot_loss[loss=0.2357, simple_loss=0.3124, pruned_loss=0.07952, over 1746067.10 frames. ], batch size: 68, lr: 3.97e-03, grad_scale: 16.0 2023-05-16 16:08:53,162 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.9920, 2.1304, 2.5668, 3.0027, 2.1755, 3.0524, 3.0890, 3.1244], device='cuda:0'), covar=tensor([0.0152, 0.1111, 0.0487, 0.0178, 0.1091, 0.0231, 0.0230, 0.0128], device='cuda:0'), in_proj_covar=tensor([0.0117, 0.0195, 0.0175, 0.0112, 0.0180, 0.0170, 0.0165, 0.0115], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 16:08:53,190 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=221819.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 16:08:59,260 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.43 vs. limit=2.0 2023-05-16 16:09:09,546 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.8677, 2.1013, 2.7431, 2.9130, 2.9594, 2.9589, 2.8670, 2.3627], device='cuda:0'), covar=tensor([0.0076, 0.0385, 0.0179, 0.0072, 0.0108, 0.0097, 0.0104, 0.0336], device='cuda:0'), in_proj_covar=tensor([0.0082, 0.0120, 0.0103, 0.0074, 0.0097, 0.0112, 0.0092, 0.0134], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 16:09:17,507 INFO [finetune.py:992] (0/2) Epoch 10, batch 11750, loss[loss=0.2646, simple_loss=0.326, pruned_loss=0.1016, over 7025.00 frames. ], tot_loss[loss=0.2377, simple_loss=0.3132, pruned_loss=0.08109, over 1699064.94 frames. ], batch size: 98, lr: 3.97e-03, grad_scale: 16.0 2023-05-16 16:09:22,866 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.552e+02 3.467e+02 3.945e+02 4.631e+02 8.258e+02, threshold=7.889e+02, percent-clipped=2.0 2023-05-16 16:09:23,692 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=221863.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 16:09:41,806 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=221890.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 16:09:51,662 INFO [finetune.py:992] (0/2) Epoch 10, batch 11800, loss[loss=0.2957, simple_loss=0.3462, pruned_loss=0.1226, over 6690.00 frames. ], tot_loss[loss=0.2406, simple_loss=0.3158, pruned_loss=0.08271, over 1686141.60 frames. ], batch size: 100, lr: 3.97e-03, grad_scale: 16.0 2023-05-16 16:10:23,930 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=221951.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 16:10:26,396 INFO [finetune.py:992] (0/2) Epoch 10, batch 11850, loss[loss=0.2316, simple_loss=0.3148, pruned_loss=0.07416, over 11741.00 frames. ], tot_loss[loss=0.2417, simple_loss=0.3173, pruned_loss=0.08306, over 1679418.32 frames. ], batch size: 44, lr: 3.97e-03, grad_scale: 16.0 2023-05-16 16:10:31,132 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.758e+02 3.642e+02 4.323e+02 4.856e+02 1.009e+03, threshold=8.645e+02, percent-clipped=1.0 2023-05-16 16:10:34,555 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.6674, 5.2389, 4.8483, 4.9569, 5.3476, 4.7500, 4.8171, 4.8049], device='cuda:0'), covar=tensor([0.1178, 0.0829, 0.1042, 0.1472, 0.0810, 0.1966, 0.1749, 0.0921], device='cuda:0'), in_proj_covar=tensor([0.0337, 0.0459, 0.0376, 0.0418, 0.0437, 0.0411, 0.0369, 0.0357], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002], device='cuda:0') 2023-05-16 16:10:35,962 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.7461, 3.4507, 3.5747, 3.7725, 3.3263, 3.7617, 3.7895, 3.7949], device='cuda:0'), covar=tensor([0.0261, 0.0191, 0.0216, 0.0355, 0.0740, 0.0380, 0.0191, 0.0279], device='cuda:0'), in_proj_covar=tensor([0.0169, 0.0170, 0.0165, 0.0216, 0.0211, 0.0189, 0.0150, 0.0201], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:0') 2023-05-16 16:10:40,460 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.99 vs. limit=2.0 2023-05-16 16:10:58,409 INFO [checkpoint.py:75] (0/2) Saving checkpoint to pruned_transducer_stateless7/exp_giga_finetune/checkpoint-122000.pt 2023-05-16 16:11:05,062 INFO [finetune.py:992] (0/2) Epoch 10, batch 11900, loss[loss=0.2317, simple_loss=0.3108, pruned_loss=0.07624, over 7198.00 frames. ], tot_loss[loss=0.2382, simple_loss=0.3151, pruned_loss=0.0806, over 1694987.36 frames. ], batch size: 97, lr: 3.97e-03, grad_scale: 8.0 2023-05-16 16:11:07,933 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=222009.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 16:11:35,484 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.7361, 2.7507, 4.4630, 4.5796, 2.9338, 2.6002, 2.9203, 1.9908], device='cuda:0'), covar=tensor([0.1488, 0.2884, 0.0401, 0.0389, 0.1240, 0.2507, 0.2793, 0.4607], device='cuda:0'), in_proj_covar=tensor([0.0279, 0.0369, 0.0261, 0.0286, 0.0256, 0.0290, 0.0364, 0.0360], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-16 16:11:39,914 INFO [finetune.py:992] (0/2) Epoch 10, batch 11950, loss[loss=0.1756, simple_loss=0.2739, pruned_loss=0.03869, over 11191.00 frames. ], tot_loss[loss=0.2334, simple_loss=0.3118, pruned_loss=0.07749, over 1697449.59 frames. ], batch size: 55, lr: 3.97e-03, grad_scale: 8.0 2023-05-16 16:11:45,245 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.111e+02 3.118e+02 3.605e+02 4.252e+02 6.600e+02, threshold=7.210e+02, percent-clipped=0.0 2023-05-16 16:11:50,208 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=222070.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 16:12:14,983 INFO [finetune.py:992] (0/2) Epoch 10, batch 12000, loss[loss=0.1872, simple_loss=0.2827, pruned_loss=0.04581, over 11691.00 frames. ], tot_loss[loss=0.2268, simple_loss=0.3068, pruned_loss=0.07343, over 1691594.52 frames. ], batch size: 44, lr: 3.97e-03, grad_scale: 8.0 2023-05-16 16:12:14,984 INFO [finetune.py:1017] (0/2) Computing validation loss 2023-05-16 16:12:33,168 INFO [finetune.py:1026] (0/2) Epoch 10, validation: loss=0.2883, simple_loss=0.3634, pruned_loss=0.1066, over 1020973.00 frames. 2023-05-16 16:12:33,169 INFO [finetune.py:1027] (0/2) Maximum memory allocated so far is 12827MB 2023-05-16 16:12:39,355 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=222114.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 16:13:07,739 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.8054, 3.1494, 2.4511, 2.2554, 2.8503, 2.2994, 3.0773, 2.6700], device='cuda:0'), covar=tensor([0.0584, 0.0588, 0.1028, 0.1467, 0.0274, 0.1193, 0.0519, 0.0728], device='cuda:0'), in_proj_covar=tensor([0.0181, 0.0237, 0.0170, 0.0192, 0.0133, 0.0175, 0.0185, 0.0169], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 16:13:08,115 INFO [finetune.py:992] (0/2) Epoch 10, batch 12050, loss[loss=0.2494, simple_loss=0.3197, pruned_loss=0.08953, over 7009.00 frames. ], tot_loss[loss=0.2216, simple_loss=0.3026, pruned_loss=0.07028, over 1692233.58 frames. ], batch size: 98, lr: 3.97e-03, grad_scale: 8.0 2023-05-16 16:13:13,304 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.964e+02 2.833e+02 3.308e+02 3.914e+02 7.827e+02, threshold=6.617e+02, percent-clipped=2.0 2023-05-16 16:13:13,452 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=222163.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 16:13:27,547 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=222185.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 16:13:30,978 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=2.15 vs. limit=2.0 2023-05-16 16:13:40,368 INFO [finetune.py:992] (0/2) Epoch 10, batch 12100, loss[loss=0.2251, simple_loss=0.308, pruned_loss=0.07115, over 11074.00 frames. ], tot_loss[loss=0.2199, simple_loss=0.3016, pruned_loss=0.06909, over 1697288.26 frames. ], batch size: 55, lr: 3.97e-03, grad_scale: 8.0 2023-05-16 16:13:40,574 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=222205.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 16:13:44,261 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=222211.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 16:13:51,005 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.99 vs. limit=2.0 2023-05-16 16:14:03,627 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2023-05-16 16:14:07,657 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=222246.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 16:14:07,750 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=222246.0, num_to_drop=1, layers_to_drop={3} 2023-05-16 16:14:13,168 INFO [finetune.py:992] (0/2) Epoch 10, batch 12150, loss[loss=0.2118, simple_loss=0.3006, pruned_loss=0.0615, over 12053.00 frames. ], tot_loss[loss=0.2197, simple_loss=0.3017, pruned_loss=0.06887, over 1706017.47 frames. ], batch size: 42, lr: 3.97e-03, grad_scale: 8.0 2023-05-16 16:14:18,159 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.270e+02 3.038e+02 3.502e+02 4.308e+02 1.208e+03, threshold=7.004e+02, percent-clipped=4.0 2023-05-16 16:14:20,226 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=222266.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 16:14:45,046 INFO [finetune.py:992] (0/2) Epoch 10, batch 12200, loss[loss=0.2398, simple_loss=0.3118, pruned_loss=0.0839, over 7269.00 frames. ], tot_loss[loss=0.2212, simple_loss=0.303, pruned_loss=0.06966, over 1693262.23 frames. ], batch size: 98, lr: 3.97e-03, grad_scale: 8.0 2023-05-16 16:15:02,715 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.5798, 4.5620, 4.4539, 4.1832, 4.1579, 4.5473, 4.2904, 4.1806], device='cuda:0'), covar=tensor([0.0851, 0.0913, 0.0667, 0.1233, 0.1917, 0.0869, 0.1510, 0.1048], device='cuda:0'), in_proj_covar=tensor([0.0568, 0.0518, 0.0479, 0.0578, 0.0384, 0.0658, 0.0704, 0.0524], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002], device='cuda:0') 2023-05-16 16:15:06,596 INFO [checkpoint.py:75] (0/2) Saving checkpoint to pruned_transducer_stateless7/exp_giga_finetune/epoch-10.pt 2023-05-16 16:15:28,750 INFO [finetune.py:992] (0/2) Epoch 11, batch 0, loss[loss=0.1744, simple_loss=0.2599, pruned_loss=0.04448, over 12289.00 frames. ], tot_loss[loss=0.1744, simple_loss=0.2599, pruned_loss=0.04448, over 12289.00 frames. ], batch size: 33, lr: 3.97e-03, grad_scale: 8.0 2023-05-16 16:15:28,751 INFO [finetune.py:1017] (0/2) Computing validation loss 2023-05-16 16:15:36,539 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.4687, 4.9507, 4.6612, 4.7859, 5.0639, 4.3432, 4.4404, 4.5312], device='cuda:0'), covar=tensor([0.1157, 0.1063, 0.0920, 0.1340, 0.0820, 0.2088, 0.2084, 0.1116], device='cuda:0'), in_proj_covar=tensor([0.0333, 0.0457, 0.0371, 0.0415, 0.0431, 0.0406, 0.0364, 0.0352], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002], device='cuda:0') 2023-05-16 16:15:45,297 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.6210, 3.4255, 3.5602, 3.6230, 3.3279, 3.5981, 3.6998, 3.7119], device='cuda:0'), covar=tensor([0.0243, 0.0243, 0.0222, 0.0369, 0.0602, 0.0444, 0.0202, 0.0284], device='cuda:0'), in_proj_covar=tensor([0.0165, 0.0166, 0.0161, 0.0210, 0.0205, 0.0185, 0.0147, 0.0197], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:0') 2023-05-16 16:15:47,008 INFO [finetune.py:1026] (0/2) Epoch 11, validation: loss=0.2863, simple_loss=0.3621, pruned_loss=0.1053, over 1020973.00 frames. 2023-05-16 16:15:47,008 INFO [finetune.py:1027] (0/2) Maximum memory allocated so far is 12827MB 2023-05-16 16:15:52,941 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.3820, 4.6851, 2.9579, 2.4117, 4.0641, 2.2804, 4.0152, 3.1439], device='cuda:0'), covar=tensor([0.0658, 0.0386, 0.1111, 0.1880, 0.0281, 0.1697, 0.0419, 0.0862], device='cuda:0'), in_proj_covar=tensor([0.0180, 0.0235, 0.0171, 0.0192, 0.0132, 0.0176, 0.0185, 0.0169], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 16:16:04,072 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.436e+02 3.276e+02 3.874e+02 4.757e+02 1.252e+03, threshold=7.749e+02, percent-clipped=3.0 2023-05-16 16:16:05,461 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=222365.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 16:16:22,435 INFO [finetune.py:992] (0/2) Epoch 11, batch 50, loss[loss=0.1508, simple_loss=0.2361, pruned_loss=0.03271, over 12137.00 frames. ], tot_loss[loss=0.178, simple_loss=0.2692, pruned_loss=0.04338, over 543711.71 frames. ], batch size: 30, lr: 3.97e-03, grad_scale: 8.0 2023-05-16 16:16:40,351 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=222414.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 16:16:57,995 INFO [finetune.py:992] (0/2) Epoch 11, batch 100, loss[loss=0.1447, simple_loss=0.2319, pruned_loss=0.02869, over 12361.00 frames. ], tot_loss[loss=0.1784, simple_loss=0.2692, pruned_loss=0.04379, over 951439.29 frames. ], batch size: 30, lr: 3.97e-03, grad_scale: 8.0 2023-05-16 16:17:14,273 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=222462.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 16:17:14,906 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.044e+02 2.885e+02 3.237e+02 3.919e+02 6.319e+02, threshold=6.473e+02, percent-clipped=0.0 2023-05-16 16:17:19,400 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.3696, 2.9255, 3.9880, 3.3447, 3.8243, 3.4952, 2.9213, 3.8907], device='cuda:0'), covar=tensor([0.0143, 0.0282, 0.0129, 0.0233, 0.0132, 0.0160, 0.0314, 0.0107], device='cuda:0'), in_proj_covar=tensor([0.0165, 0.0195, 0.0173, 0.0175, 0.0201, 0.0153, 0.0187, 0.0173], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 16:17:24,237 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.3708, 5.1910, 5.2783, 5.3493, 5.0404, 5.0381, 4.8321, 5.3161], device='cuda:0'), covar=tensor([0.0757, 0.0626, 0.0873, 0.0639, 0.1733, 0.1359, 0.0583, 0.0977], device='cuda:0'), in_proj_covar=tensor([0.0497, 0.0646, 0.0560, 0.0578, 0.0767, 0.0682, 0.0508, 0.0444], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0005, 0.0004, 0.0003, 0.0003], device='cuda:0') 2023-05-16 16:17:33,846 INFO [finetune.py:992] (0/2) Epoch 11, batch 150, loss[loss=0.1744, simple_loss=0.263, pruned_loss=0.04291, over 12251.00 frames. ], tot_loss[loss=0.1753, simple_loss=0.265, pruned_loss=0.04275, over 1271143.76 frames. ], batch size: 32, lr: 3.97e-03, grad_scale: 8.0 2023-05-16 16:17:55,933 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.4986, 5.0905, 5.4535, 4.8176, 5.1161, 4.9044, 5.5818, 5.1270], device='cuda:0'), covar=tensor([0.0306, 0.0329, 0.0312, 0.0255, 0.0335, 0.0302, 0.0201, 0.0237], device='cuda:0'), in_proj_covar=tensor([0.0243, 0.0245, 0.0267, 0.0242, 0.0242, 0.0239, 0.0218, 0.0204], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 16:17:58,824 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=222523.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 16:18:09,725 INFO [finetune.py:992] (0/2) Epoch 11, batch 200, loss[loss=0.1794, simple_loss=0.2857, pruned_loss=0.03659, over 12338.00 frames. ], tot_loss[loss=0.1737, simple_loss=0.2633, pruned_loss=0.04203, over 1519402.90 frames. ], batch size: 36, lr: 3.97e-03, grad_scale: 8.0 2023-05-16 16:18:11,188 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=222541.0, num_to_drop=1, layers_to_drop={1} 2023-05-16 16:18:14,832 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=222546.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 16:18:25,415 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=222561.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 16:18:26,779 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.057e+02 2.661e+02 3.019e+02 3.700e+02 1.022e+03, threshold=6.039e+02, percent-clipped=2.0 2023-05-16 16:18:32,074 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=222570.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 16:18:36,611 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.61 vs. limit=2.0 2023-05-16 16:18:42,006 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=222584.0, num_to_drop=1, layers_to_drop={2} 2023-05-16 16:18:45,174 INFO [finetune.py:992] (0/2) Epoch 11, batch 250, loss[loss=0.1685, simple_loss=0.2484, pruned_loss=0.0443, over 12294.00 frames. ], tot_loss[loss=0.1723, simple_loss=0.2614, pruned_loss=0.04163, over 1712873.65 frames. ], batch size: 33, lr: 3.97e-03, grad_scale: 8.0 2023-05-16 16:18:48,800 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=222594.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 16:19:08,997 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.9546, 4.2636, 4.0013, 4.4834, 3.4476, 4.0917, 2.6179, 4.5171], device='cuda:0'), covar=tensor([0.1135, 0.0607, 0.1362, 0.0876, 0.0927, 0.0505, 0.1719, 0.1199], device='cuda:0'), in_proj_covar=tensor([0.0230, 0.0270, 0.0298, 0.0355, 0.0238, 0.0242, 0.0262, 0.0378], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 16:19:15,233 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=222631.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 16:19:21,182 INFO [finetune.py:992] (0/2) Epoch 11, batch 300, loss[loss=0.2013, simple_loss=0.2847, pruned_loss=0.05897, over 12070.00 frames. ], tot_loss[loss=0.1723, simple_loss=0.261, pruned_loss=0.04176, over 1864296.98 frames. ], batch size: 42, lr: 3.96e-03, grad_scale: 8.0 2023-05-16 16:19:38,884 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.989e+02 2.847e+02 3.403e+02 4.061e+02 8.881e+02, threshold=6.805e+02, percent-clipped=3.0 2023-05-16 16:19:40,274 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=222665.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 16:19:44,542 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([6.0091, 5.9642, 5.7454, 5.2988, 5.1233, 5.8942, 5.4919, 5.2473], device='cuda:0'), covar=tensor([0.0705, 0.0992, 0.0650, 0.1561, 0.0643, 0.0683, 0.1475, 0.1097], device='cuda:0'), in_proj_covar=tensor([0.0591, 0.0541, 0.0498, 0.0607, 0.0398, 0.0685, 0.0740, 0.0547], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-05-16 16:19:50,732 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.48 vs. limit=2.0 2023-05-16 16:19:57,315 INFO [finetune.py:992] (0/2) Epoch 11, batch 350, loss[loss=0.1521, simple_loss=0.2282, pruned_loss=0.03806, over 11788.00 frames. ], tot_loss[loss=0.1733, simple_loss=0.2622, pruned_loss=0.0422, over 1976941.75 frames. ], batch size: 26, lr: 3.96e-03, grad_scale: 8.0 2023-05-16 16:20:14,646 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=222713.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 16:20:21,184 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.5573, 4.9060, 4.1164, 5.1702, 4.7039, 2.8244, 4.3609, 2.9674], device='cuda:0'), covar=tensor([0.0686, 0.0672, 0.1397, 0.0411, 0.0972, 0.1833, 0.0997, 0.3460], device='cuda:0'), in_proj_covar=tensor([0.0294, 0.0366, 0.0343, 0.0282, 0.0354, 0.0263, 0.0332, 0.0356], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 16:20:24,201 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=2.59 vs. limit=5.0 2023-05-16 16:20:33,064 INFO [finetune.py:992] (0/2) Epoch 11, batch 400, loss[loss=0.147, simple_loss=0.2286, pruned_loss=0.03266, over 11828.00 frames. ], tot_loss[loss=0.1726, simple_loss=0.2614, pruned_loss=0.0419, over 2057162.55 frames. ], batch size: 26, lr: 3.96e-03, grad_scale: 8.0 2023-05-16 16:20:47,727 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.43 vs. limit=2.0 2023-05-16 16:20:50,135 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.851e+02 2.690e+02 3.204e+02 3.852e+02 5.513e+02, threshold=6.408e+02, percent-clipped=1.0 2023-05-16 16:21:08,679 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.59 vs. limit=2.0 2023-05-16 16:21:09,022 INFO [finetune.py:992] (0/2) Epoch 11, batch 450, loss[loss=0.1501, simple_loss=0.2281, pruned_loss=0.03608, over 12288.00 frames. ], tot_loss[loss=0.1723, simple_loss=0.2611, pruned_loss=0.04175, over 2129330.58 frames. ], batch size: 28, lr: 3.96e-03, grad_scale: 8.0 2023-05-16 16:21:45,581 INFO [finetune.py:992] (0/2) Epoch 11, batch 500, loss[loss=0.1647, simple_loss=0.2635, pruned_loss=0.0329, over 12298.00 frames. ], tot_loss[loss=0.1714, simple_loss=0.2601, pruned_loss=0.04138, over 2189406.18 frames. ], batch size: 34, lr: 3.96e-03, grad_scale: 8.0 2023-05-16 16:21:47,173 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=222841.0, num_to_drop=1, layers_to_drop={2} 2023-05-16 16:22:01,183 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=222861.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 16:22:02,449 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.755e+02 2.741e+02 3.201e+02 3.737e+02 1.157e+03, threshold=6.403e+02, percent-clipped=2.0 2023-05-16 16:22:13,794 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=222879.0, num_to_drop=1, layers_to_drop={0} 2023-05-16 16:22:20,824 INFO [finetune.py:992] (0/2) Epoch 11, batch 550, loss[loss=0.1316, simple_loss=0.2161, pruned_loss=0.02358, over 12344.00 frames. ], tot_loss[loss=0.1706, simple_loss=0.2596, pruned_loss=0.04084, over 2234285.59 frames. ], batch size: 30, lr: 3.96e-03, grad_scale: 8.0 2023-05-16 16:22:20,891 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=222889.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 16:22:30,146 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=4.49 vs. limit=5.0 2023-05-16 16:22:35,370 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=222909.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 16:22:38,511 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.7962, 2.8398, 4.7650, 4.9661, 2.8390, 2.5716, 3.0266, 2.2368], device='cuda:0'), covar=tensor([0.1549, 0.3139, 0.0396, 0.0359, 0.1376, 0.2551, 0.2910, 0.4285], device='cuda:0'), in_proj_covar=tensor([0.0289, 0.0382, 0.0269, 0.0296, 0.0265, 0.0299, 0.0375, 0.0372], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-16 16:22:45,953 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.42 vs. limit=2.0 2023-05-16 16:22:47,608 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=222926.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 16:22:57,489 INFO [finetune.py:992] (0/2) Epoch 11, batch 600, loss[loss=0.2137, simple_loss=0.3057, pruned_loss=0.06086, over 11806.00 frames. ], tot_loss[loss=0.1701, simple_loss=0.2592, pruned_loss=0.04045, over 2275762.20 frames. ], batch size: 44, lr: 3.96e-03, grad_scale: 8.0 2023-05-16 16:22:57,863 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=2.80 vs. limit=5.0 2023-05-16 16:22:58,349 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=222940.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 16:23:15,002 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.030e+02 2.809e+02 3.240e+02 4.121e+02 5.870e+02, threshold=6.479e+02, percent-clipped=0.0 2023-05-16 16:23:19,534 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=222969.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 16:23:19,659 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.4362, 4.7618, 4.0292, 5.0012, 4.4823, 3.0604, 4.3299, 2.9989], device='cuda:0'), covar=tensor([0.0739, 0.0770, 0.1512, 0.0465, 0.1244, 0.1574, 0.1053, 0.3399], device='cuda:0'), in_proj_covar=tensor([0.0297, 0.0370, 0.0346, 0.0286, 0.0359, 0.0265, 0.0336, 0.0361], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 16:23:30,032 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=222984.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 16:23:32,431 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.38 vs. limit=2.0 2023-05-16 16:23:33,368 INFO [finetune.py:992] (0/2) Epoch 11, batch 650, loss[loss=0.1901, simple_loss=0.2787, pruned_loss=0.05077, over 10521.00 frames. ], tot_loss[loss=0.1704, simple_loss=0.2597, pruned_loss=0.04051, over 2298467.93 frames. ], batch size: 68, lr: 3.96e-03, grad_scale: 8.0 2023-05-16 16:23:42,316 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=223001.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 16:24:02,422 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.8269, 3.6360, 3.7555, 3.8197, 3.5451, 3.9003, 3.8543, 4.0038], device='cuda:0'), covar=tensor([0.0326, 0.0229, 0.0207, 0.0409, 0.0653, 0.0427, 0.0218, 0.0251], device='cuda:0'), in_proj_covar=tensor([0.0183, 0.0185, 0.0178, 0.0233, 0.0230, 0.0205, 0.0163, 0.0214], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:0') 2023-05-16 16:24:02,459 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=223030.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 16:24:08,664 INFO [finetune.py:992] (0/2) Epoch 11, batch 700, loss[loss=0.1575, simple_loss=0.2528, pruned_loss=0.03105, over 12365.00 frames. ], tot_loss[loss=0.1699, simple_loss=0.2593, pruned_loss=0.04025, over 2317127.66 frames. ], batch size: 36, lr: 3.96e-03, grad_scale: 8.0 2023-05-16 16:24:13,139 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=223045.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 16:24:25,711 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.768e+02 2.739e+02 3.214e+02 3.661e+02 1.810e+03, threshold=6.428e+02, percent-clipped=2.0 2023-05-16 16:24:26,642 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=223064.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 16:24:44,725 INFO [finetune.py:992] (0/2) Epoch 11, batch 750, loss[loss=0.1571, simple_loss=0.2454, pruned_loss=0.03439, over 12087.00 frames. ], tot_loss[loss=0.1701, simple_loss=0.2595, pruned_loss=0.04037, over 2322271.14 frames. ], batch size: 32, lr: 3.96e-03, grad_scale: 8.0 2023-05-16 16:25:11,224 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=223125.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 16:25:20,902 INFO [finetune.py:992] (0/2) Epoch 11, batch 800, loss[loss=0.1511, simple_loss=0.252, pruned_loss=0.02514, over 11778.00 frames. ], tot_loss[loss=0.1701, simple_loss=0.2596, pruned_loss=0.0403, over 2338840.74 frames. ], batch size: 44, lr: 3.96e-03, grad_scale: 8.0 2023-05-16 16:25:25,342 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.9433, 3.5272, 5.2508, 2.6525, 2.8174, 3.8937, 3.3588, 3.9320], device='cuda:0'), covar=tensor([0.0408, 0.1076, 0.0275, 0.1202, 0.1943, 0.1482, 0.1280, 0.1097], device='cuda:0'), in_proj_covar=tensor([0.0235, 0.0234, 0.0242, 0.0181, 0.0235, 0.0290, 0.0222, 0.0258], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-16 16:25:37,927 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.637e+02 2.699e+02 3.208e+02 3.913e+02 7.854e+02, threshold=6.416e+02, percent-clipped=1.0 2023-05-16 16:25:45,208 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=223173.0, num_to_drop=1, layers_to_drop={1} 2023-05-16 16:25:45,784 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.8998, 5.7504, 5.8286, 5.0584, 5.1175, 5.9623, 5.1222, 5.2911], device='cuda:0'), covar=tensor([0.1333, 0.1981, 0.1187, 0.2848, 0.1232, 0.1206, 0.3499, 0.2263], device='cuda:0'), in_proj_covar=tensor([0.0611, 0.0552, 0.0512, 0.0625, 0.0411, 0.0703, 0.0760, 0.0564], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0004, 0.0004, 0.0003], device='cuda:0') 2023-05-16 16:25:49,378 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=223179.0, num_to_drop=1, layers_to_drop={0} 2023-05-16 16:25:54,334 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.8192, 2.7439, 4.6886, 4.9163, 2.8865, 2.6502, 2.9783, 2.1396], device='cuda:0'), covar=tensor([0.1612, 0.3411, 0.0457, 0.0378, 0.1384, 0.2495, 0.2929, 0.4231], device='cuda:0'), in_proj_covar=tensor([0.0288, 0.0382, 0.0269, 0.0296, 0.0264, 0.0299, 0.0374, 0.0371], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-16 16:25:56,196 INFO [finetune.py:992] (0/2) Epoch 11, batch 850, loss[loss=0.1583, simple_loss=0.2418, pruned_loss=0.03738, over 11770.00 frames. ], tot_loss[loss=0.1703, simple_loss=0.2596, pruned_loss=0.04048, over 2341835.21 frames. ], batch size: 26, lr: 3.96e-03, grad_scale: 8.0 2023-05-16 16:26:18,824 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.4801, 2.4318, 3.5912, 4.3157, 3.7297, 4.3790, 3.8541, 2.9719], device='cuda:0'), covar=tensor([0.0030, 0.0417, 0.0161, 0.0046, 0.0126, 0.0079, 0.0114, 0.0361], device='cuda:0'), in_proj_covar=tensor([0.0084, 0.0122, 0.0104, 0.0075, 0.0099, 0.0113, 0.0093, 0.0137], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 16:26:23,678 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=223226.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 16:26:24,295 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=223227.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 16:26:30,322 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=223234.0, num_to_drop=1, layers_to_drop={1} 2023-05-16 16:26:33,552 INFO [finetune.py:992] (0/2) Epoch 11, batch 900, loss[loss=0.1465, simple_loss=0.2297, pruned_loss=0.03164, over 11996.00 frames. ], tot_loss[loss=0.1697, simple_loss=0.2591, pruned_loss=0.04021, over 2358591.25 frames. ], batch size: 28, lr: 3.96e-03, grad_scale: 8.0 2023-05-16 16:26:45,776 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.8342, 3.1799, 3.8720, 4.7264, 4.0889, 4.8268, 4.1829, 3.2939], device='cuda:0'), covar=tensor([0.0027, 0.0317, 0.0124, 0.0038, 0.0101, 0.0056, 0.0078, 0.0337], device='cuda:0'), in_proj_covar=tensor([0.0084, 0.0122, 0.0104, 0.0076, 0.0099, 0.0113, 0.0094, 0.0137], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 16:26:50,602 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.029e+02 2.660e+02 3.181e+02 3.660e+02 7.233e+02, threshold=6.361e+02, percent-clipped=3.0 2023-05-16 16:26:58,411 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=223274.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 16:27:09,005 INFO [finetune.py:992] (0/2) Epoch 11, batch 950, loss[loss=0.1654, simple_loss=0.2582, pruned_loss=0.03628, over 12031.00 frames. ], tot_loss[loss=0.1687, simple_loss=0.2577, pruned_loss=0.03983, over 2364559.79 frames. ], batch size: 42, lr: 3.96e-03, grad_scale: 8.0 2023-05-16 16:27:13,976 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=223296.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 16:27:33,942 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=3.06 vs. limit=5.0 2023-05-16 16:27:34,392 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=223325.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 16:27:44,296 INFO [finetune.py:992] (0/2) Epoch 11, batch 1000, loss[loss=0.1503, simple_loss=0.2288, pruned_loss=0.03594, over 11846.00 frames. ], tot_loss[loss=0.1687, simple_loss=0.2577, pruned_loss=0.03979, over 2375321.90 frames. ], batch size: 26, lr: 3.96e-03, grad_scale: 8.0 2023-05-16 16:27:45,147 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=223340.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 16:28:00,425 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.5369, 4.2867, 4.1407, 4.4923, 3.2431, 4.0914, 2.6784, 4.1681], device='cuda:0'), covar=tensor([0.1582, 0.0608, 0.0943, 0.0829, 0.1061, 0.0568, 0.1862, 0.1188], device='cuda:0'), in_proj_covar=tensor([0.0230, 0.0268, 0.0299, 0.0358, 0.0239, 0.0244, 0.0262, 0.0379], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 16:28:01,560 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.903e+02 2.720e+02 3.145e+02 3.668e+02 6.745e+02, threshold=6.290e+02, percent-clipped=1.0 2023-05-16 16:28:07,986 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([6.1822, 6.0916, 5.9193, 5.3544, 5.1599, 6.0430, 5.7093, 5.4279], device='cuda:0'), covar=tensor([0.0665, 0.0938, 0.0595, 0.1580, 0.0651, 0.0727, 0.1425, 0.1043], device='cuda:0'), in_proj_covar=tensor([0.0608, 0.0553, 0.0513, 0.0625, 0.0410, 0.0703, 0.0762, 0.0562], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0004, 0.0004, 0.0003], device='cuda:0') 2023-05-16 16:28:09,179 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.42 vs. limit=2.0 2023-05-16 16:28:21,552 INFO [finetune.py:992] (0/2) Epoch 11, batch 1050, loss[loss=0.1568, simple_loss=0.2376, pruned_loss=0.03799, over 12127.00 frames. ], tot_loss[loss=0.1687, simple_loss=0.2577, pruned_loss=0.03984, over 2372266.83 frames. ], batch size: 30, lr: 3.96e-03, grad_scale: 8.0 2023-05-16 16:28:43,648 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=223420.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 16:28:44,632 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=3.28 vs. limit=5.0 2023-05-16 16:28:57,079 INFO [finetune.py:992] (0/2) Epoch 11, batch 1100, loss[loss=0.1733, simple_loss=0.2674, pruned_loss=0.03966, over 12167.00 frames. ], tot_loss[loss=0.1689, simple_loss=0.258, pruned_loss=0.03994, over 2379318.93 frames. ], batch size: 36, lr: 3.96e-03, grad_scale: 8.0 2023-05-16 16:29:14,114 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.958e+02 2.738e+02 3.282e+02 3.607e+02 7.657e+02, threshold=6.564e+02, percent-clipped=2.0 2023-05-16 16:29:27,365 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.36 vs. limit=2.0 2023-05-16 16:29:32,508 INFO [finetune.py:992] (0/2) Epoch 11, batch 1150, loss[loss=0.1694, simple_loss=0.2592, pruned_loss=0.03982, over 12339.00 frames. ], tot_loss[loss=0.169, simple_loss=0.2582, pruned_loss=0.03986, over 2376224.17 frames. ], batch size: 36, lr: 3.96e-03, grad_scale: 8.0 2023-05-16 16:30:00,837 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=223529.0, num_to_drop=1, layers_to_drop={2} 2023-05-16 16:30:08,303 INFO [finetune.py:992] (0/2) Epoch 11, batch 1200, loss[loss=0.1497, simple_loss=0.2384, pruned_loss=0.03048, over 12102.00 frames. ], tot_loss[loss=0.1687, simple_loss=0.258, pruned_loss=0.03974, over 2381121.10 frames. ], batch size: 33, lr: 3.96e-03, grad_scale: 8.0 2023-05-16 16:30:25,837 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.958e+02 2.885e+02 3.352e+02 3.938e+02 8.392e+02, threshold=6.704e+02, percent-clipped=4.0 2023-05-16 16:30:36,598 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=223578.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 16:30:44,088 INFO [finetune.py:992] (0/2) Epoch 11, batch 1250, loss[loss=0.2193, simple_loss=0.2959, pruned_loss=0.07136, over 7802.00 frames. ], tot_loss[loss=0.1694, simple_loss=0.2585, pruned_loss=0.04014, over 2375882.70 frames. ], batch size: 97, lr: 3.96e-03, grad_scale: 8.0 2023-05-16 16:30:49,109 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=223596.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 16:31:09,704 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=223625.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 16:31:17,051 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.7132, 2.8861, 4.6471, 4.7357, 2.8382, 2.5930, 2.9746, 2.1225], device='cuda:0'), covar=tensor([0.1558, 0.2688, 0.0388, 0.0437, 0.1321, 0.2379, 0.2672, 0.4188], device='cuda:0'), in_proj_covar=tensor([0.0292, 0.0387, 0.0272, 0.0299, 0.0267, 0.0304, 0.0379, 0.0376], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-16 16:31:19,615 INFO [finetune.py:992] (0/2) Epoch 11, batch 1300, loss[loss=0.1678, simple_loss=0.2564, pruned_loss=0.0396, over 12180.00 frames. ], tot_loss[loss=0.1689, simple_loss=0.2579, pruned_loss=0.03992, over 2380153.40 frames. ], batch size: 31, lr: 3.96e-03, grad_scale: 8.0 2023-05-16 16:31:19,837 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=223639.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 16:31:20,461 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=223640.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 16:31:23,245 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=223644.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 16:31:36,579 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.418e+02 2.575e+02 3.052e+02 3.546e+02 8.927e+02, threshold=6.105e+02, percent-clipped=2.0 2023-05-16 16:31:43,763 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=223673.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 16:31:55,877 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=223688.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 16:31:56,520 INFO [finetune.py:992] (0/2) Epoch 11, batch 1350, loss[loss=0.1916, simple_loss=0.2766, pruned_loss=0.05324, over 12127.00 frames. ], tot_loss[loss=0.1694, simple_loss=0.2585, pruned_loss=0.04018, over 2379548.97 frames. ], batch size: 38, lr: 3.95e-03, grad_scale: 8.0 2023-05-16 16:32:18,437 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=223720.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 16:32:31,895 INFO [finetune.py:992] (0/2) Epoch 11, batch 1400, loss[loss=0.2068, simple_loss=0.2992, pruned_loss=0.05724, over 12035.00 frames. ], tot_loss[loss=0.17, simple_loss=0.2592, pruned_loss=0.04043, over 2384401.23 frames. ], batch size: 40, lr: 3.95e-03, grad_scale: 8.0 2023-05-16 16:32:32,324 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=3.43 vs. limit=5.0 2023-05-16 16:32:48,660 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.615e+02 2.710e+02 3.305e+02 4.108e+02 9.896e+02, threshold=6.610e+02, percent-clipped=5.0 2023-05-16 16:32:52,401 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=223768.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 16:33:07,223 INFO [finetune.py:992] (0/2) Epoch 11, batch 1450, loss[loss=0.2138, simple_loss=0.2979, pruned_loss=0.06485, over 7914.00 frames. ], tot_loss[loss=0.17, simple_loss=0.2591, pruned_loss=0.04042, over 2387142.30 frames. ], batch size: 98, lr: 3.95e-03, grad_scale: 8.0 2023-05-16 16:33:14,511 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.9932, 4.8509, 4.7989, 4.8271, 4.5028, 4.9413, 4.9646, 5.1263], device='cuda:0'), covar=tensor([0.0187, 0.0166, 0.0191, 0.0408, 0.0795, 0.0328, 0.0160, 0.0176], device='cuda:0'), in_proj_covar=tensor([0.0189, 0.0190, 0.0184, 0.0241, 0.0237, 0.0211, 0.0169, 0.0220], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0005, 0.0004, 0.0004, 0.0003, 0.0005], device='cuda:0') 2023-05-16 16:33:36,661 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=223829.0, num_to_drop=1, layers_to_drop={0} 2023-05-16 16:33:44,140 INFO [finetune.py:992] (0/2) Epoch 11, batch 1500, loss[loss=0.2152, simple_loss=0.2858, pruned_loss=0.0723, over 8476.00 frames. ], tot_loss[loss=0.1711, simple_loss=0.2598, pruned_loss=0.04117, over 2371706.00 frames. ], batch size: 99, lr: 3.95e-03, grad_scale: 8.0 2023-05-16 16:34:01,156 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.996e+02 2.778e+02 3.304e+02 3.965e+02 3.242e+03, threshold=6.607e+02, percent-clipped=6.0 2023-05-16 16:34:11,151 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=223877.0, num_to_drop=1, layers_to_drop={1} 2023-05-16 16:34:19,641 INFO [finetune.py:992] (0/2) Epoch 11, batch 1550, loss[loss=0.1935, simple_loss=0.2908, pruned_loss=0.04807, over 10649.00 frames. ], tot_loss[loss=0.1708, simple_loss=0.2593, pruned_loss=0.04116, over 2373521.44 frames. ], batch size: 68, lr: 3.95e-03, grad_scale: 8.0 2023-05-16 16:34:51,451 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=223934.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 16:34:54,975 INFO [finetune.py:992] (0/2) Epoch 11, batch 1600, loss[loss=0.1978, simple_loss=0.2807, pruned_loss=0.05744, over 12138.00 frames. ], tot_loss[loss=0.1713, simple_loss=0.2598, pruned_loss=0.04142, over 2370942.24 frames. ], batch size: 38, lr: 3.95e-03, grad_scale: 8.0 2023-05-16 16:35:11,589 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.746e+02 2.692e+02 3.217e+02 3.636e+02 1.307e+03, threshold=6.433e+02, percent-clipped=3.0 2023-05-16 16:35:31,545 INFO [finetune.py:992] (0/2) Epoch 11, batch 1650, loss[loss=0.1571, simple_loss=0.2537, pruned_loss=0.03022, over 12367.00 frames. ], tot_loss[loss=0.1702, simple_loss=0.2587, pruned_loss=0.04081, over 2376947.90 frames. ], batch size: 36, lr: 3.95e-03, grad_scale: 16.0 2023-05-16 16:35:35,971 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.3971, 4.9683, 5.3642, 4.6494, 4.9841, 4.7556, 5.4471, 5.1042], device='cuda:0'), covar=tensor([0.0310, 0.0398, 0.0279, 0.0270, 0.0374, 0.0343, 0.0212, 0.0225], device='cuda:0'), in_proj_covar=tensor([0.0256, 0.0260, 0.0282, 0.0256, 0.0256, 0.0256, 0.0231, 0.0213], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 16:35:39,718 INFO [checkpoint.py:75] (0/2) Saving checkpoint to pruned_transducer_stateless7/exp_giga_finetune/checkpoint-124000.pt 2023-05-16 16:36:10,075 INFO [finetune.py:992] (0/2) Epoch 11, batch 1700, loss[loss=0.1477, simple_loss=0.238, pruned_loss=0.02869, over 12009.00 frames. ], tot_loss[loss=0.1697, simple_loss=0.2581, pruned_loss=0.04061, over 2384275.92 frames. ], batch size: 28, lr: 3.95e-03, grad_scale: 16.0 2023-05-16 16:36:27,398 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.737e+02 2.666e+02 3.190e+02 3.702e+02 6.745e+02, threshold=6.379e+02, percent-clipped=1.0 2023-05-16 16:36:32,037 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.3806, 4.6085, 4.0904, 4.9219, 4.5397, 2.7580, 4.4259, 3.0371], device='cuda:0'), covar=tensor([0.0696, 0.0784, 0.1324, 0.0524, 0.1016, 0.1693, 0.0876, 0.3262], device='cuda:0'), in_proj_covar=tensor([0.0297, 0.0370, 0.0347, 0.0288, 0.0359, 0.0265, 0.0335, 0.0359], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 16:36:42,375 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.3066, 2.4325, 3.5510, 4.2427, 3.7711, 4.2433, 3.7087, 2.8312], device='cuda:0'), covar=tensor([0.0047, 0.0446, 0.0169, 0.0049, 0.0129, 0.0084, 0.0123, 0.0426], device='cuda:0'), in_proj_covar=tensor([0.0086, 0.0123, 0.0106, 0.0077, 0.0101, 0.0115, 0.0096, 0.0139], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 16:36:45,440 INFO [finetune.py:992] (0/2) Epoch 11, batch 1750, loss[loss=0.1629, simple_loss=0.24, pruned_loss=0.04291, over 11801.00 frames. ], tot_loss[loss=0.1698, simple_loss=0.2583, pruned_loss=0.04064, over 2379850.74 frames. ], batch size: 26, lr: 3.95e-03, grad_scale: 16.0 2023-05-16 16:36:49,200 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.8683, 2.4884, 3.4480, 2.8511, 3.2311, 3.0104, 2.4232, 3.3480], device='cuda:0'), covar=tensor([0.0136, 0.0382, 0.0157, 0.0253, 0.0180, 0.0181, 0.0367, 0.0147], device='cuda:0'), in_proj_covar=tensor([0.0173, 0.0203, 0.0184, 0.0184, 0.0210, 0.0161, 0.0195, 0.0182], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 16:37:02,662 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.67 vs. limit=2.0 2023-05-16 16:37:21,883 INFO [finetune.py:992] (0/2) Epoch 11, batch 1800, loss[loss=0.1747, simple_loss=0.2731, pruned_loss=0.03814, over 12051.00 frames. ], tot_loss[loss=0.1696, simple_loss=0.2584, pruned_loss=0.04041, over 2382497.99 frames. ], batch size: 40, lr: 3.95e-03, grad_scale: 16.0 2023-05-16 16:37:38,948 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.217e+02 2.842e+02 3.339e+02 3.794e+02 8.326e+02, threshold=6.677e+02, percent-clipped=2.0 2023-05-16 16:37:52,326 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.8859, 4.7097, 4.8232, 4.8563, 4.4960, 4.6027, 4.4062, 4.7858], device='cuda:0'), covar=tensor([0.0816, 0.0721, 0.0912, 0.0661, 0.2239, 0.1338, 0.0619, 0.1153], device='cuda:0'), in_proj_covar=tensor([0.0535, 0.0696, 0.0598, 0.0622, 0.0836, 0.0733, 0.0546, 0.0475], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0005, 0.0004, 0.0004, 0.0005, 0.0005, 0.0003, 0.0003], device='cuda:0') 2023-05-16 16:37:57,016 INFO [finetune.py:992] (0/2) Epoch 11, batch 1850, loss[loss=0.1785, simple_loss=0.2763, pruned_loss=0.04038, over 12157.00 frames. ], tot_loss[loss=0.1692, simple_loss=0.2583, pruned_loss=0.04004, over 2382741.41 frames. ], batch size: 39, lr: 3.95e-03, grad_scale: 16.0 2023-05-16 16:38:13,904 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=224212.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 16:38:29,459 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=224234.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 16:38:29,860 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.38 vs. limit=2.0 2023-05-16 16:38:32,777 INFO [finetune.py:992] (0/2) Epoch 11, batch 1900, loss[loss=0.1492, simple_loss=0.2346, pruned_loss=0.03189, over 12339.00 frames. ], tot_loss[loss=0.1692, simple_loss=0.2585, pruned_loss=0.03998, over 2387969.28 frames. ], batch size: 30, lr: 3.95e-03, grad_scale: 16.0 2023-05-16 16:38:34,403 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.7394, 3.3431, 5.1822, 2.7669, 3.0470, 3.8295, 3.3780, 3.8444], device='cuda:0'), covar=tensor([0.0444, 0.1201, 0.0284, 0.1124, 0.1798, 0.1657, 0.1287, 0.1230], device='cuda:0'), in_proj_covar=tensor([0.0233, 0.0231, 0.0242, 0.0179, 0.0233, 0.0288, 0.0221, 0.0259], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-16 16:38:48,375 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-05-16 16:38:49,363 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.899e+02 2.710e+02 3.083e+02 3.655e+02 7.153e+02, threshold=6.166e+02, percent-clipped=1.0 2023-05-16 16:38:58,158 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=224273.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 16:39:04,126 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=224282.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 16:39:09,045 INFO [finetune.py:992] (0/2) Epoch 11, batch 1950, loss[loss=0.1695, simple_loss=0.2562, pruned_loss=0.04142, over 12188.00 frames. ], tot_loss[loss=0.1698, simple_loss=0.259, pruned_loss=0.04032, over 2381611.69 frames. ], batch size: 31, lr: 3.95e-03, grad_scale: 16.0 2023-05-16 16:39:13,584 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.4347, 5.2621, 5.4126, 5.4131, 5.0080, 5.0965, 4.8833, 5.3696], device='cuda:0'), covar=tensor([0.0767, 0.0629, 0.0774, 0.0572, 0.2019, 0.1329, 0.0529, 0.1001], device='cuda:0'), in_proj_covar=tensor([0.0531, 0.0692, 0.0593, 0.0619, 0.0835, 0.0731, 0.0543, 0.0472], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0005, 0.0004, 0.0004, 0.0005, 0.0005, 0.0003, 0.0003], device='cuda:0') 2023-05-16 16:39:15,639 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.3560, 4.0131, 4.1757, 4.2049, 4.0539, 4.3401, 4.1821, 2.4145], device='cuda:0'), covar=tensor([0.0126, 0.0099, 0.0112, 0.0091, 0.0071, 0.0107, 0.0121, 0.0910], device='cuda:0'), in_proj_covar=tensor([0.0069, 0.0079, 0.0083, 0.0073, 0.0060, 0.0092, 0.0082, 0.0098], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 16:39:44,485 INFO [finetune.py:992] (0/2) Epoch 11, batch 2000, loss[loss=0.1767, simple_loss=0.2765, pruned_loss=0.03848, over 12353.00 frames. ], tot_loss[loss=0.1696, simple_loss=0.2586, pruned_loss=0.04024, over 2386912.61 frames. ], batch size: 36, lr: 3.95e-03, grad_scale: 16.0 2023-05-16 16:40:01,801 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.828e+02 2.747e+02 3.182e+02 3.685e+02 6.124e+02, threshold=6.365e+02, percent-clipped=0.0 2023-05-16 16:40:20,054 INFO [finetune.py:992] (0/2) Epoch 11, batch 2050, loss[loss=0.1648, simple_loss=0.2573, pruned_loss=0.03613, over 10760.00 frames. ], tot_loss[loss=0.1698, simple_loss=0.2589, pruned_loss=0.04039, over 2386617.54 frames. ], batch size: 69, lr: 3.95e-03, grad_scale: 16.0 2023-05-16 16:40:52,244 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.40 vs. limit=2.0 2023-05-16 16:40:56,649 INFO [finetune.py:992] (0/2) Epoch 11, batch 2100, loss[loss=0.1614, simple_loss=0.2393, pruned_loss=0.04179, over 11362.00 frames. ], tot_loss[loss=0.1704, simple_loss=0.2595, pruned_loss=0.04063, over 2382737.98 frames. ], batch size: 25, lr: 3.95e-03, grad_scale: 16.0 2023-05-16 16:41:13,657 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.905e+02 2.918e+02 3.370e+02 3.984e+02 9.339e+02, threshold=6.740e+02, percent-clipped=3.0 2023-05-16 16:41:24,253 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.8667, 4.7829, 4.7116, 4.7436, 4.3965, 4.8828, 4.8879, 5.0551], device='cuda:0'), covar=tensor([0.0227, 0.0146, 0.0181, 0.0308, 0.0818, 0.0239, 0.0118, 0.0159], device='cuda:0'), in_proj_covar=tensor([0.0193, 0.0195, 0.0189, 0.0246, 0.0243, 0.0215, 0.0174, 0.0224], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0005, 0.0004, 0.0004, 0.0003, 0.0005], device='cuda:0') 2023-05-16 16:41:31,931 INFO [finetune.py:992] (0/2) Epoch 11, batch 2150, loss[loss=0.1868, simple_loss=0.2727, pruned_loss=0.05045, over 11559.00 frames. ], tot_loss[loss=0.1699, simple_loss=0.259, pruned_loss=0.04043, over 2385199.49 frames. ], batch size: 48, lr: 3.95e-03, grad_scale: 16.0 2023-05-16 16:42:07,230 INFO [finetune.py:992] (0/2) Epoch 11, batch 2200, loss[loss=0.2476, simple_loss=0.3178, pruned_loss=0.08868, over 8192.00 frames. ], tot_loss[loss=0.1705, simple_loss=0.2593, pruned_loss=0.0408, over 2384684.07 frames. ], batch size: 97, lr: 3.95e-03, grad_scale: 16.0 2023-05-16 16:42:23,582 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=224562.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 16:42:24,044 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.040e+02 2.732e+02 3.415e+02 4.168e+02 1.123e+03, threshold=6.829e+02, percent-clipped=5.0 2023-05-16 16:42:29,026 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=224568.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 16:42:34,072 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.2069, 4.5457, 4.0003, 4.7128, 4.4980, 2.7931, 4.1791, 3.0204], device='cuda:0'), covar=tensor([0.0805, 0.0746, 0.1341, 0.0542, 0.1025, 0.1677, 0.0960, 0.3281], device='cuda:0'), in_proj_covar=tensor([0.0304, 0.0378, 0.0355, 0.0297, 0.0367, 0.0271, 0.0344, 0.0367], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 16:42:43,657 INFO [finetune.py:992] (0/2) Epoch 11, batch 2250, loss[loss=0.2022, simple_loss=0.2901, pruned_loss=0.05713, over 12029.00 frames. ], tot_loss[loss=0.1703, simple_loss=0.259, pruned_loss=0.04079, over 2385105.98 frames. ], batch size: 40, lr: 3.95e-03, grad_scale: 16.0 2023-05-16 16:42:49,479 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.1146, 6.0060, 5.5987, 5.6065, 6.0931, 5.4264, 5.4750, 5.6634], device='cuda:0'), covar=tensor([0.1781, 0.1043, 0.0975, 0.1921, 0.1075, 0.2304, 0.2106, 0.1148], device='cuda:0'), in_proj_covar=tensor([0.0357, 0.0493, 0.0395, 0.0443, 0.0464, 0.0437, 0.0394, 0.0384], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 16:42:57,739 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.3679, 4.6862, 4.1318, 4.8591, 4.6559, 2.8156, 4.2237, 3.0957], device='cuda:0'), covar=tensor([0.0786, 0.0768, 0.1355, 0.0554, 0.0984, 0.1772, 0.1072, 0.3232], device='cuda:0'), in_proj_covar=tensor([0.0305, 0.0379, 0.0357, 0.0298, 0.0369, 0.0272, 0.0345, 0.0368], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 16:43:08,358 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=224623.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 16:43:19,156 INFO [finetune.py:992] (0/2) Epoch 11, batch 2300, loss[loss=0.1668, simple_loss=0.2592, pruned_loss=0.03723, over 11249.00 frames. ], tot_loss[loss=0.1695, simple_loss=0.2582, pruned_loss=0.04036, over 2388597.24 frames. ], batch size: 55, lr: 3.95e-03, grad_scale: 16.0 2023-05-16 16:43:36,345 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.903e+02 2.733e+02 3.158e+02 3.825e+02 8.863e+02, threshold=6.315e+02, percent-clipped=2.0 2023-05-16 16:43:54,398 INFO [finetune.py:992] (0/2) Epoch 11, batch 2350, loss[loss=0.1471, simple_loss=0.2311, pruned_loss=0.03159, over 12087.00 frames. ], tot_loss[loss=0.1685, simple_loss=0.2569, pruned_loss=0.04002, over 2390530.99 frames. ], batch size: 32, lr: 3.95e-03, grad_scale: 16.0 2023-05-16 16:44:08,328 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=4.54 vs. limit=5.0 2023-05-16 16:44:21,967 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=224726.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 16:44:25,827 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.33 vs. limit=2.0 2023-05-16 16:44:30,984 INFO [finetune.py:992] (0/2) Epoch 11, batch 2400, loss[loss=0.1531, simple_loss=0.2385, pruned_loss=0.03383, over 12042.00 frames. ], tot_loss[loss=0.1686, simple_loss=0.2572, pruned_loss=0.04003, over 2386034.26 frames. ], batch size: 31, lr: 3.94e-03, grad_scale: 16.0 2023-05-16 16:44:31,880 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=224740.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 16:44:47,588 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.958e+02 2.760e+02 3.237e+02 3.987e+02 7.740e+02, threshold=6.473e+02, percent-clipped=4.0 2023-05-16 16:45:02,044 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.3431, 3.4831, 3.1408, 3.1587, 2.8429, 2.7112, 3.4610, 2.2764], device='cuda:0'), covar=tensor([0.0416, 0.0119, 0.0201, 0.0171, 0.0400, 0.0409, 0.0130, 0.0462], device='cuda:0'), in_proj_covar=tensor([0.0193, 0.0165, 0.0161, 0.0186, 0.0208, 0.0202, 0.0172, 0.0202], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 16:45:04,779 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=224787.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 16:45:05,986 INFO [finetune.py:992] (0/2) Epoch 11, batch 2450, loss[loss=0.1582, simple_loss=0.2515, pruned_loss=0.03249, over 12353.00 frames. ], tot_loss[loss=0.1696, simple_loss=0.2581, pruned_loss=0.04054, over 2376312.49 frames. ], batch size: 35, lr: 3.94e-03, grad_scale: 16.0 2023-05-16 16:45:14,869 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=224801.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 16:45:29,405 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.89 vs. limit=2.0 2023-05-16 16:45:41,785 INFO [finetune.py:992] (0/2) Epoch 11, batch 2500, loss[loss=0.175, simple_loss=0.2666, pruned_loss=0.04173, over 12354.00 frames. ], tot_loss[loss=0.1687, simple_loss=0.2574, pruned_loss=0.04002, over 2379794.79 frames. ], batch size: 36, lr: 3.94e-03, grad_scale: 16.0 2023-05-16 16:45:59,442 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.697e+02 2.514e+02 2.989e+02 3.631e+02 1.476e+03, threshold=5.979e+02, percent-clipped=2.0 2023-05-16 16:46:03,073 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=224868.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 16:46:08,775 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.5389, 2.9573, 3.6588, 4.5228, 3.9421, 4.5048, 3.8145, 3.2455], device='cuda:0'), covar=tensor([0.0036, 0.0337, 0.0154, 0.0033, 0.0121, 0.0074, 0.0121, 0.0329], device='cuda:0'), in_proj_covar=tensor([0.0086, 0.0123, 0.0105, 0.0077, 0.0102, 0.0114, 0.0097, 0.0138], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 16:46:18,569 INFO [finetune.py:992] (0/2) Epoch 11, batch 2550, loss[loss=0.1693, simple_loss=0.2588, pruned_loss=0.03987, over 12347.00 frames. ], tot_loss[loss=0.1679, simple_loss=0.2565, pruned_loss=0.03962, over 2388138.67 frames. ], batch size: 31, lr: 3.94e-03, grad_scale: 16.0 2023-05-16 16:46:37,455 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=224916.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 16:46:38,914 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=224918.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 16:46:53,574 INFO [finetune.py:992] (0/2) Epoch 11, batch 2600, loss[loss=0.1731, simple_loss=0.2635, pruned_loss=0.04132, over 12019.00 frames. ], tot_loss[loss=0.1687, simple_loss=0.2577, pruned_loss=0.0399, over 2381400.31 frames. ], batch size: 40, lr: 3.94e-03, grad_scale: 16.0 2023-05-16 16:47:10,462 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.748e+02 2.594e+02 2.935e+02 3.512e+02 9.403e+02, threshold=5.870e+02, percent-clipped=2.0 2023-05-16 16:47:18,431 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.3606, 5.1696, 5.2640, 5.3064, 4.9161, 4.9679, 4.7333, 5.2946], device='cuda:0'), covar=tensor([0.0704, 0.0660, 0.0896, 0.0685, 0.2223, 0.1397, 0.0622, 0.0992], device='cuda:0'), in_proj_covar=tensor([0.0537, 0.0703, 0.0602, 0.0624, 0.0844, 0.0739, 0.0547, 0.0477], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0005, 0.0004, 0.0004, 0.0005, 0.0005, 0.0003, 0.0003], device='cuda:0') 2023-05-16 16:47:28,781 INFO [finetune.py:992] (0/2) Epoch 11, batch 2650, loss[loss=0.1678, simple_loss=0.2524, pruned_loss=0.0416, over 12096.00 frames. ], tot_loss[loss=0.1685, simple_loss=0.2573, pruned_loss=0.03986, over 2384541.96 frames. ], batch size: 33, lr: 3.94e-03, grad_scale: 16.0 2023-05-16 16:47:43,005 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.58 vs. limit=2.0 2023-05-16 16:48:05,443 INFO [finetune.py:992] (0/2) Epoch 11, batch 2700, loss[loss=0.1433, simple_loss=0.2237, pruned_loss=0.03145, over 12347.00 frames. ], tot_loss[loss=0.1693, simple_loss=0.2579, pruned_loss=0.04029, over 2369700.43 frames. ], batch size: 30, lr: 3.94e-03, grad_scale: 16.0 2023-05-16 16:48:22,302 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.025e+02 2.698e+02 3.150e+02 3.686e+02 6.435e+02, threshold=6.301e+02, percent-clipped=2.0 2023-05-16 16:48:22,758 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.95 vs. limit=2.0 2023-05-16 16:48:35,994 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=225082.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 16:48:40,992 INFO [finetune.py:992] (0/2) Epoch 11, batch 2750, loss[loss=0.1419, simple_loss=0.2264, pruned_loss=0.02868, over 12295.00 frames. ], tot_loss[loss=0.1693, simple_loss=0.2581, pruned_loss=0.04023, over 2365676.01 frames. ], batch size: 28, lr: 3.94e-03, grad_scale: 16.0 2023-05-16 16:48:45,797 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=225096.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 16:48:50,883 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=225103.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 16:49:06,450 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.5038, 3.0051, 3.8104, 2.3445, 2.6533, 3.0679, 2.9063, 3.2559], device='cuda:0'), covar=tensor([0.0553, 0.0979, 0.0524, 0.1184, 0.1648, 0.1588, 0.1266, 0.1066], device='cuda:0'), in_proj_covar=tensor([0.0236, 0.0235, 0.0248, 0.0182, 0.0238, 0.0295, 0.0225, 0.0264], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-16 16:49:15,891 INFO [finetune.py:992] (0/2) Epoch 11, batch 2800, loss[loss=0.197, simple_loss=0.2919, pruned_loss=0.05104, over 12144.00 frames. ], tot_loss[loss=0.1688, simple_loss=0.2573, pruned_loss=0.04013, over 2368541.10 frames. ], batch size: 39, lr: 3.94e-03, grad_scale: 16.0 2023-05-16 16:49:18,866 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=225142.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 16:49:34,288 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.892e+02 2.666e+02 3.138e+02 3.783e+02 6.283e+02, threshold=6.276e+02, percent-clipped=0.0 2023-05-16 16:49:35,201 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=225164.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 16:49:39,918 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.90 vs. limit=2.0 2023-05-16 16:49:52,807 INFO [finetune.py:992] (0/2) Epoch 11, batch 2850, loss[loss=0.1515, simple_loss=0.2342, pruned_loss=0.03445, over 12134.00 frames. ], tot_loss[loss=0.1678, simple_loss=0.2561, pruned_loss=0.03977, over 2377131.34 frames. ], batch size: 30, lr: 3.94e-03, grad_scale: 16.0 2023-05-16 16:50:03,243 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=225203.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 16:50:12,402 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.3100, 3.7469, 3.3165, 3.3221, 3.0914, 2.9831, 3.7028, 2.2600], device='cuda:0'), covar=tensor([0.0490, 0.0155, 0.0188, 0.0163, 0.0353, 0.0302, 0.0132, 0.0494], device='cuda:0'), in_proj_covar=tensor([0.0191, 0.0164, 0.0161, 0.0185, 0.0205, 0.0198, 0.0171, 0.0200], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 16:50:13,695 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=225218.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 16:50:28,554 INFO [finetune.py:992] (0/2) Epoch 11, batch 2900, loss[loss=0.2211, simple_loss=0.3001, pruned_loss=0.071, over 7871.00 frames. ], tot_loss[loss=0.1677, simple_loss=0.2558, pruned_loss=0.03983, over 2380076.58 frames. ], batch size: 98, lr: 3.94e-03, grad_scale: 16.0 2023-05-16 16:50:45,576 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.803e+02 2.705e+02 3.062e+02 3.716e+02 6.501e+02, threshold=6.124e+02, percent-clipped=1.0 2023-05-16 16:50:47,664 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=225266.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 16:51:04,584 INFO [finetune.py:992] (0/2) Epoch 11, batch 2950, loss[loss=0.1453, simple_loss=0.2235, pruned_loss=0.03353, over 12183.00 frames. ], tot_loss[loss=0.1686, simple_loss=0.2569, pruned_loss=0.04015, over 2371373.33 frames. ], batch size: 29, lr: 3.94e-03, grad_scale: 16.0 2023-05-16 16:51:36,176 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2023-05-16 16:51:40,654 INFO [finetune.py:992] (0/2) Epoch 11, batch 3000, loss[loss=0.1564, simple_loss=0.2453, pruned_loss=0.03378, over 12088.00 frames. ], tot_loss[loss=0.1682, simple_loss=0.2567, pruned_loss=0.03991, over 2376434.46 frames. ], batch size: 32, lr: 3.94e-03, grad_scale: 16.0 2023-05-16 16:51:40,654 INFO [finetune.py:1017] (0/2) Computing validation loss 2023-05-16 16:51:59,595 INFO [finetune.py:1026] (0/2) Epoch 11, validation: loss=0.3119, simple_loss=0.3915, pruned_loss=0.1162, over 1020973.00 frames. 2023-05-16 16:51:59,596 INFO [finetune.py:1027] (0/2) Maximum memory allocated so far is 12827MB 2023-05-16 16:52:16,698 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.992e+02 2.700e+02 3.015e+02 3.753e+02 6.118e+02, threshold=6.030e+02, percent-clipped=0.0 2023-05-16 16:52:30,259 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=225382.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 16:52:35,512 INFO [finetune.py:992] (0/2) Epoch 11, batch 3050, loss[loss=0.2319, simple_loss=0.3036, pruned_loss=0.08015, over 8094.00 frames. ], tot_loss[loss=0.1681, simple_loss=0.2565, pruned_loss=0.03984, over 2375505.01 frames. ], batch size: 98, lr: 3.94e-03, grad_scale: 8.0 2023-05-16 16:52:40,467 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=225396.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 16:52:43,802 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.1779, 3.9542, 2.5715, 2.3888, 3.4537, 2.4592, 3.6280, 2.9664], device='cuda:0'), covar=tensor([0.0630, 0.0633, 0.1109, 0.1489, 0.0272, 0.1285, 0.0500, 0.0740], device='cuda:0'), in_proj_covar=tensor([0.0191, 0.0258, 0.0180, 0.0202, 0.0143, 0.0184, 0.0199, 0.0178], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 16:53:05,365 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=225430.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 16:53:11,618 INFO [finetune.py:992] (0/2) Epoch 11, batch 3100, loss[loss=0.143, simple_loss=0.234, pruned_loss=0.02597, over 12189.00 frames. ], tot_loss[loss=0.1676, simple_loss=0.256, pruned_loss=0.03963, over 2368499.59 frames. ], batch size: 31, lr: 3.94e-03, grad_scale: 8.0 2023-05-16 16:53:11,808 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=225439.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 16:53:15,178 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=225444.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 16:53:18,938 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.0451, 3.7096, 5.4060, 2.9456, 3.2681, 4.1041, 3.6255, 4.2303], device='cuda:0'), covar=tensor([0.0387, 0.0997, 0.0304, 0.1088, 0.1675, 0.1370, 0.1173, 0.0933], device='cuda:0'), in_proj_covar=tensor([0.0235, 0.0234, 0.0246, 0.0180, 0.0236, 0.0292, 0.0222, 0.0263], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-16 16:53:25,743 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=225459.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 16:53:27,982 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=225462.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 16:53:29,231 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.080e+02 2.825e+02 3.231e+02 3.810e+02 6.061e+02, threshold=6.463e+02, percent-clipped=1.0 2023-05-16 16:53:35,109 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.1837, 4.1953, 2.5783, 2.2545, 3.6745, 2.3886, 3.7879, 3.0015], device='cuda:0'), covar=tensor([0.0721, 0.0692, 0.1155, 0.1743, 0.0248, 0.1447, 0.0475, 0.0842], device='cuda:0'), in_proj_covar=tensor([0.0190, 0.0256, 0.0178, 0.0201, 0.0142, 0.0182, 0.0198, 0.0176], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 16:53:46,891 INFO [finetune.py:992] (0/2) Epoch 11, batch 3150, loss[loss=0.1342, simple_loss=0.2184, pruned_loss=0.02503, over 11915.00 frames. ], tot_loss[loss=0.1684, simple_loss=0.2569, pruned_loss=0.03993, over 2353810.04 frames. ], batch size: 26, lr: 3.94e-03, grad_scale: 8.0 2023-05-16 16:53:53,147 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=225498.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 16:53:54,733 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=225500.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 16:53:58,088 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=225505.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 16:54:09,580 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.7900, 4.0456, 3.5339, 4.2061, 3.9026, 2.6996, 3.7404, 2.9231], device='cuda:0'), covar=tensor([0.0857, 0.0866, 0.1390, 0.0712, 0.1150, 0.1621, 0.1121, 0.3091], device='cuda:0'), in_proj_covar=tensor([0.0302, 0.0377, 0.0357, 0.0297, 0.0366, 0.0270, 0.0342, 0.0365], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 16:54:10,197 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=225522.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 16:54:10,945 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=225523.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 16:54:15,095 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.6970, 4.6155, 4.4978, 4.5761, 4.2346, 4.6952, 4.6898, 4.8491], device='cuda:0'), covar=tensor([0.0273, 0.0182, 0.0250, 0.0376, 0.0783, 0.0367, 0.0167, 0.0207], device='cuda:0'), in_proj_covar=tensor([0.0193, 0.0197, 0.0190, 0.0246, 0.0243, 0.0217, 0.0176, 0.0225], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0005, 0.0004, 0.0005, 0.0003, 0.0005], device='cuda:0') 2023-05-16 16:54:22,565 INFO [finetune.py:992] (0/2) Epoch 11, batch 3200, loss[loss=0.1599, simple_loss=0.2393, pruned_loss=0.04022, over 12181.00 frames. ], tot_loss[loss=0.1686, simple_loss=0.2569, pruned_loss=0.04011, over 2353222.23 frames. ], batch size: 29, lr: 3.94e-03, grad_scale: 8.0 2023-05-16 16:54:25,895 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.40 vs. limit=2.0 2023-05-16 16:54:37,585 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=225559.0, num_to_drop=1, layers_to_drop={0} 2023-05-16 16:54:40,879 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.892e+02 2.979e+02 3.359e+02 3.894e+02 1.470e+03, threshold=6.719e+02, percent-clipped=3.0 2023-05-16 16:54:42,588 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=225566.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 16:54:47,492 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.4463, 4.8852, 3.0636, 2.8039, 4.1868, 2.7604, 4.2292, 3.4294], device='cuda:0'), covar=tensor([0.0788, 0.0523, 0.1138, 0.1604, 0.0314, 0.1389, 0.0482, 0.0825], device='cuda:0'), in_proj_covar=tensor([0.0190, 0.0255, 0.0178, 0.0201, 0.0143, 0.0183, 0.0197, 0.0176], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 16:54:54,523 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=225583.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 16:54:58,449 INFO [finetune.py:992] (0/2) Epoch 11, batch 3250, loss[loss=0.1566, simple_loss=0.2468, pruned_loss=0.03315, over 12342.00 frames. ], tot_loss[loss=0.1686, simple_loss=0.2571, pruned_loss=0.04009, over 2358806.98 frames. ], batch size: 36, lr: 3.94e-03, grad_scale: 8.0 2023-05-16 16:55:18,382 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.14 vs. limit=2.0 2023-05-16 16:55:21,091 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=225620.0, num_to_drop=1, layers_to_drop={2} 2023-05-16 16:55:23,154 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.0349, 5.8628, 5.3926, 5.3400, 5.9758, 5.2137, 5.4501, 5.4322], device='cuda:0'), covar=tensor([0.1490, 0.0987, 0.1056, 0.2190, 0.0915, 0.2186, 0.1711, 0.1313], device='cuda:0'), in_proj_covar=tensor([0.0356, 0.0491, 0.0393, 0.0448, 0.0463, 0.0436, 0.0396, 0.0385], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 16:55:34,491 INFO [finetune.py:992] (0/2) Epoch 11, batch 3300, loss[loss=0.1647, simple_loss=0.2667, pruned_loss=0.03138, over 12198.00 frames. ], tot_loss[loss=0.1679, simple_loss=0.2565, pruned_loss=0.03968, over 2366045.63 frames. ], batch size: 35, lr: 3.94e-03, grad_scale: 8.0 2023-05-16 16:55:52,485 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.021e+02 2.640e+02 3.098e+02 3.639e+02 5.733e+02, threshold=6.196e+02, percent-clipped=0.0 2023-05-16 16:56:02,954 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.62 vs. limit=2.0 2023-05-16 16:56:10,715 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.96 vs. limit=2.0 2023-05-16 16:56:10,942 INFO [finetune.py:992] (0/2) Epoch 11, batch 3350, loss[loss=0.1843, simple_loss=0.2762, pruned_loss=0.04618, over 12034.00 frames. ], tot_loss[loss=0.1677, simple_loss=0.2565, pruned_loss=0.03945, over 2363550.94 frames. ], batch size: 40, lr: 3.94e-03, grad_scale: 8.0 2023-05-16 16:56:47,201 INFO [finetune.py:992] (0/2) Epoch 11, batch 3400, loss[loss=0.1776, simple_loss=0.2619, pruned_loss=0.04671, over 12362.00 frames. ], tot_loss[loss=0.1677, simple_loss=0.2564, pruned_loss=0.0395, over 2371335.24 frames. ], batch size: 36, lr: 3.93e-03, grad_scale: 8.0 2023-05-16 16:57:01,251 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=225759.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 16:57:04,701 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.010e+02 2.798e+02 3.269e+02 3.862e+02 6.913e+02, threshold=6.537e+02, percent-clipped=1.0 2023-05-16 16:57:05,857 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2023-05-16 16:57:22,149 INFO [finetune.py:992] (0/2) Epoch 11, batch 3450, loss[loss=0.1827, simple_loss=0.2781, pruned_loss=0.04364, over 12060.00 frames. ], tot_loss[loss=0.1685, simple_loss=0.2573, pruned_loss=0.03982, over 2361694.79 frames. ], batch size: 37, lr: 3.93e-03, grad_scale: 8.0 2023-05-16 16:57:26,595 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=225795.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 16:57:28,803 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=225798.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 16:57:29,837 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.85 vs. limit=2.0 2023-05-16 16:57:35,247 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=225807.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 16:57:43,195 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=225818.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 16:57:54,577 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.5134, 4.2415, 4.1774, 4.6259, 3.3001, 4.1094, 2.8008, 4.2183], device='cuda:0'), covar=tensor([0.1457, 0.0627, 0.0929, 0.0602, 0.0963, 0.0549, 0.1672, 0.1322], device='cuda:0'), in_proj_covar=tensor([0.0225, 0.0266, 0.0295, 0.0353, 0.0236, 0.0241, 0.0259, 0.0375], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 16:57:58,490 INFO [finetune.py:992] (0/2) Epoch 11, batch 3500, loss[loss=0.231, simple_loss=0.3174, pruned_loss=0.07225, over 11341.00 frames. ], tot_loss[loss=0.1681, simple_loss=0.2572, pruned_loss=0.03948, over 2363689.32 frames. ], batch size: 55, lr: 3.93e-03, grad_scale: 8.0 2023-05-16 16:58:03,473 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=225846.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 16:58:14,554 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=225861.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 16:58:15,378 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=225862.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 16:58:16,560 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.689e+02 2.805e+02 3.086e+02 3.880e+02 7.528e+02, threshold=6.172e+02, percent-clipped=2.0 2023-05-16 16:58:26,489 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=225878.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 16:58:33,827 INFO [finetune.py:992] (0/2) Epoch 11, batch 3550, loss[loss=0.1916, simple_loss=0.2803, pruned_loss=0.05142, over 12283.00 frames. ], tot_loss[loss=0.1682, simple_loss=0.2574, pruned_loss=0.03956, over 2369941.83 frames. ], batch size: 37, lr: 3.93e-03, grad_scale: 8.0 2023-05-16 16:58:38,395 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=225895.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 16:58:52,144 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=225915.0, num_to_drop=1, layers_to_drop={0} 2023-05-16 16:58:57,932 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=225923.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 16:59:08,896 INFO [finetune.py:992] (0/2) Epoch 11, batch 3600, loss[loss=0.1462, simple_loss=0.2246, pruned_loss=0.03394, over 12270.00 frames. ], tot_loss[loss=0.1685, simple_loss=0.2575, pruned_loss=0.03978, over 2372463.15 frames. ], batch size: 28, lr: 3.93e-03, grad_scale: 8.0 2023-05-16 16:59:21,235 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=225956.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 16:59:26,636 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.936e+02 2.804e+02 3.266e+02 3.834e+02 8.704e+02, threshold=6.532e+02, percent-clipped=2.0 2023-05-16 16:59:39,399 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.6436, 2.8207, 3.7526, 4.6399, 4.0166, 4.6244, 3.9944, 3.2962], device='cuda:0'), covar=tensor([0.0036, 0.0377, 0.0144, 0.0037, 0.0135, 0.0065, 0.0106, 0.0337], device='cuda:0'), in_proj_covar=tensor([0.0087, 0.0124, 0.0106, 0.0078, 0.0104, 0.0117, 0.0097, 0.0140], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 16:59:44,809 INFO [finetune.py:992] (0/2) Epoch 11, batch 3650, loss[loss=0.1918, simple_loss=0.2773, pruned_loss=0.05311, over 11807.00 frames. ], tot_loss[loss=0.168, simple_loss=0.2569, pruned_loss=0.03957, over 2374530.66 frames. ], batch size: 44, lr: 3.93e-03, grad_scale: 8.0 2023-05-16 16:59:47,471 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.38 vs. limit=2.0 2023-05-16 16:59:53,608 INFO [checkpoint.py:75] (0/2) Saving checkpoint to pruned_transducer_stateless7/exp_giga_finetune/checkpoint-126000.pt 2023-05-16 17:00:11,653 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.1706, 4.1443, 4.1459, 4.5750, 2.9494, 4.0206, 2.8223, 4.1678], device='cuda:0'), covar=tensor([0.1601, 0.0654, 0.0789, 0.0568, 0.1123, 0.0577, 0.1585, 0.1121], device='cuda:0'), in_proj_covar=tensor([0.0223, 0.0264, 0.0291, 0.0350, 0.0235, 0.0240, 0.0256, 0.0373], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 17:00:24,376 INFO [finetune.py:992] (0/2) Epoch 11, batch 3700, loss[loss=0.1673, simple_loss=0.2473, pruned_loss=0.04361, over 12347.00 frames. ], tot_loss[loss=0.1675, simple_loss=0.2565, pruned_loss=0.03926, over 2376779.06 frames. ], batch size: 31, lr: 3.93e-03, grad_scale: 8.0 2023-05-16 17:00:25,946 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=226041.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 17:00:42,175 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.984e+02 2.762e+02 3.228e+02 3.974e+02 8.639e+02, threshold=6.455e+02, percent-clipped=4.0 2023-05-16 17:00:59,582 INFO [finetune.py:992] (0/2) Epoch 11, batch 3750, loss[loss=0.1563, simple_loss=0.2529, pruned_loss=0.02982, over 12276.00 frames. ], tot_loss[loss=0.1684, simple_loss=0.2572, pruned_loss=0.03979, over 2363233.48 frames. ], batch size: 33, lr: 3.93e-03, grad_scale: 8.0 2023-05-16 17:01:03,831 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=226095.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 17:01:08,926 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=226102.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 17:01:21,049 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=226118.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 17:01:35,799 INFO [finetune.py:992] (0/2) Epoch 11, batch 3800, loss[loss=0.1846, simple_loss=0.2764, pruned_loss=0.04645, over 11505.00 frames. ], tot_loss[loss=0.1673, simple_loss=0.2561, pruned_loss=0.03927, over 2369092.76 frames. ], batch size: 48, lr: 3.93e-03, grad_scale: 8.0 2023-05-16 17:01:39,358 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=226143.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 17:01:51,847 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=226161.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 17:01:53,854 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.007e+02 2.852e+02 3.373e+02 4.117e+02 5.606e+02, threshold=6.746e+02, percent-clipped=0.0 2023-05-16 17:01:55,322 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=226166.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 17:02:03,899 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=226178.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 17:02:11,337 INFO [finetune.py:992] (0/2) Epoch 11, batch 3850, loss[loss=0.2021, simple_loss=0.2967, pruned_loss=0.0537, over 12270.00 frames. ], tot_loss[loss=0.168, simple_loss=0.2569, pruned_loss=0.03957, over 2369665.00 frames. ], batch size: 37, lr: 3.93e-03, grad_scale: 8.0 2023-05-16 17:02:25,895 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=226209.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 17:02:30,344 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=226215.0, num_to_drop=1, layers_to_drop={2} 2023-05-16 17:02:32,264 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=226218.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 17:02:37,993 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=226226.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 17:02:42,829 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=226233.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 17:02:46,685 INFO [finetune.py:992] (0/2) Epoch 11, batch 3900, loss[loss=0.1542, simple_loss=0.2385, pruned_loss=0.03493, over 12074.00 frames. ], tot_loss[loss=0.1676, simple_loss=0.2563, pruned_loss=0.03943, over 2372262.62 frames. ], batch size: 32, lr: 3.93e-03, grad_scale: 8.0 2023-05-16 17:02:51,134 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.2378, 4.8464, 5.1183, 5.0834, 4.9532, 5.1195, 5.0138, 2.8963], device='cuda:0'), covar=tensor([0.0099, 0.0063, 0.0063, 0.0050, 0.0035, 0.0087, 0.0059, 0.0721], device='cuda:0'), in_proj_covar=tensor([0.0068, 0.0078, 0.0081, 0.0072, 0.0059, 0.0091, 0.0080, 0.0097], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-16 17:02:55,268 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=226251.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 17:03:04,691 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=226263.0, num_to_drop=1, layers_to_drop={1} 2023-05-16 17:03:05,282 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.996e+02 2.742e+02 3.251e+02 4.151e+02 6.960e+02, threshold=6.503e+02, percent-clipped=2.0 2023-05-16 17:03:16,899 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.7079, 2.7950, 4.6099, 4.9469, 2.9103, 2.6889, 2.9502, 2.2907], device='cuda:0'), covar=tensor([0.1603, 0.3210, 0.0536, 0.0372, 0.1358, 0.2358, 0.2722, 0.4007], device='cuda:0'), in_proj_covar=tensor([0.0292, 0.0390, 0.0275, 0.0302, 0.0270, 0.0305, 0.0380, 0.0376], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-16 17:03:20,270 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.3268, 4.6277, 4.2034, 4.9368, 4.4820, 3.0531, 4.2961, 3.0736], device='cuda:0'), covar=tensor([0.0850, 0.0891, 0.1413, 0.0583, 0.1256, 0.1663, 0.1073, 0.3293], device='cuda:0'), in_proj_covar=tensor([0.0300, 0.0374, 0.0353, 0.0294, 0.0364, 0.0269, 0.0339, 0.0361], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 17:03:22,895 INFO [finetune.py:992] (0/2) Epoch 11, batch 3950, loss[loss=0.1649, simple_loss=0.2605, pruned_loss=0.03465, over 12151.00 frames. ], tot_loss[loss=0.1679, simple_loss=0.2571, pruned_loss=0.03937, over 2373797.50 frames. ], batch size: 34, lr: 3.93e-03, grad_scale: 8.0 2023-05-16 17:03:27,190 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=226294.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 17:03:32,924 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.54 vs. limit=2.0 2023-05-16 17:03:58,769 INFO [finetune.py:992] (0/2) Epoch 11, batch 4000, loss[loss=0.1518, simple_loss=0.2463, pruned_loss=0.02868, over 12266.00 frames. ], tot_loss[loss=0.1675, simple_loss=0.2567, pruned_loss=0.03915, over 2382120.72 frames. ], batch size: 37, lr: 3.93e-03, grad_scale: 8.0 2023-05-16 17:04:16,251 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.044e+02 2.667e+02 2.947e+02 3.775e+02 7.891e+02, threshold=5.895e+02, percent-clipped=1.0 2023-05-16 17:04:20,777 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.1214, 4.9372, 5.0407, 5.0691, 4.5074, 4.6295, 4.5088, 4.9978], device='cuda:0'), covar=tensor([0.1036, 0.0992, 0.1153, 0.0978, 0.3346, 0.1893, 0.0815, 0.1353], device='cuda:0'), in_proj_covar=tensor([0.0542, 0.0704, 0.0609, 0.0624, 0.0847, 0.0742, 0.0552, 0.0484], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0005, 0.0004, 0.0004, 0.0005, 0.0005, 0.0003, 0.0003], device='cuda:0') 2023-05-16 17:04:33,980 INFO [finetune.py:992] (0/2) Epoch 11, batch 4050, loss[loss=0.1711, simple_loss=0.2547, pruned_loss=0.04375, over 12295.00 frames. ], tot_loss[loss=0.1674, simple_loss=0.2563, pruned_loss=0.03926, over 2379970.30 frames. ], batch size: 34, lr: 3.93e-03, grad_scale: 8.0 2023-05-16 17:04:34,988 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.8752, 2.9672, 4.7542, 4.9863, 3.0086, 2.7823, 3.0526, 2.4041], device='cuda:0'), covar=tensor([0.1493, 0.2956, 0.0427, 0.0362, 0.1280, 0.2236, 0.2671, 0.3740], device='cuda:0'), in_proj_covar=tensor([0.0291, 0.0389, 0.0274, 0.0302, 0.0269, 0.0304, 0.0379, 0.0375], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-16 17:04:39,751 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=226397.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 17:04:45,819 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.91 vs. limit=2.0 2023-05-16 17:05:10,049 INFO [finetune.py:992] (0/2) Epoch 11, batch 4100, loss[loss=0.1735, simple_loss=0.2625, pruned_loss=0.0422, over 12107.00 frames. ], tot_loss[loss=0.1674, simple_loss=0.2563, pruned_loss=0.0392, over 2376816.39 frames. ], batch size: 38, lr: 3.93e-03, grad_scale: 8.0 2023-05-16 17:05:27,032 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.1486, 2.5695, 3.7624, 3.1800, 3.5298, 3.3065, 2.7199, 3.6418], device='cuda:0'), covar=tensor([0.0161, 0.0346, 0.0156, 0.0255, 0.0169, 0.0171, 0.0339, 0.0110], device='cuda:0'), in_proj_covar=tensor([0.0176, 0.0204, 0.0187, 0.0188, 0.0215, 0.0164, 0.0197, 0.0185], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 17:05:28,199 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.182e+02 2.822e+02 3.304e+02 3.943e+02 5.377e+02, threshold=6.609e+02, percent-clipped=0.0 2023-05-16 17:05:30,727 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.53 vs. limit=2.0 2023-05-16 17:05:45,790 INFO [finetune.py:992] (0/2) Epoch 11, batch 4150, loss[loss=0.1513, simple_loss=0.2425, pruned_loss=0.03003, over 12239.00 frames. ], tot_loss[loss=0.1675, simple_loss=0.2559, pruned_loss=0.03956, over 2364417.90 frames. ], batch size: 32, lr: 3.93e-03, grad_scale: 8.0 2023-05-16 17:05:54,942 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.0300, 5.9370, 5.5202, 5.4534, 5.9539, 5.2854, 5.4134, 5.4527], device='cuda:0'), covar=tensor([0.1400, 0.0802, 0.1235, 0.1695, 0.0885, 0.1963, 0.1964, 0.1106], device='cuda:0'), in_proj_covar=tensor([0.0349, 0.0480, 0.0384, 0.0440, 0.0454, 0.0425, 0.0387, 0.0376], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 17:06:06,060 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=226518.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 17:06:11,217 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.3454, 3.5101, 3.1441, 3.1083, 2.8051, 2.6196, 3.5565, 2.3278], device='cuda:0'), covar=tensor([0.0433, 0.0143, 0.0211, 0.0198, 0.0395, 0.0355, 0.0122, 0.0461], device='cuda:0'), in_proj_covar=tensor([0.0191, 0.0164, 0.0163, 0.0185, 0.0206, 0.0199, 0.0172, 0.0201], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 17:06:20,781 INFO [finetune.py:992] (0/2) Epoch 11, batch 4200, loss[loss=0.2087, simple_loss=0.2865, pruned_loss=0.06543, over 12380.00 frames. ], tot_loss[loss=0.1676, simple_loss=0.2561, pruned_loss=0.03953, over 2371200.92 frames. ], batch size: 38, lr: 3.93e-03, grad_scale: 8.0 2023-05-16 17:06:30,102 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=226551.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 17:06:38,981 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.859e+02 2.677e+02 3.287e+02 3.935e+02 7.058e+02, threshold=6.573e+02, percent-clipped=2.0 2023-05-16 17:06:40,404 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=226566.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 17:06:49,285 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=3.06 vs. limit=5.0 2023-05-16 17:06:53,962 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.6591, 2.4793, 3.2838, 4.5050, 2.7036, 4.5933, 4.6589, 4.7063], device='cuda:0'), covar=tensor([0.0116, 0.1251, 0.0479, 0.0155, 0.1185, 0.0230, 0.0155, 0.0090], device='cuda:0'), in_proj_covar=tensor([0.0119, 0.0197, 0.0177, 0.0114, 0.0182, 0.0171, 0.0170, 0.0117], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 17:06:57,256 INFO [finetune.py:992] (0/2) Epoch 11, batch 4250, loss[loss=0.1868, simple_loss=0.2837, pruned_loss=0.04498, over 12126.00 frames. ], tot_loss[loss=0.1681, simple_loss=0.2567, pruned_loss=0.03978, over 2373346.62 frames. ], batch size: 38, lr: 3.93e-03, grad_scale: 8.0 2023-05-16 17:06:57,401 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=226589.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 17:07:04,460 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=226599.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 17:07:07,956 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-05-16 17:07:33,239 INFO [finetune.py:992] (0/2) Epoch 11, batch 4300, loss[loss=0.1781, simple_loss=0.2736, pruned_loss=0.04129, over 12352.00 frames. ], tot_loss[loss=0.1688, simple_loss=0.2574, pruned_loss=0.04009, over 2373027.54 frames. ], batch size: 35, lr: 3.93e-03, grad_scale: 8.0 2023-05-16 17:07:46,132 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.9927, 5.9905, 5.7005, 5.2276, 5.1016, 5.8580, 5.4741, 5.1474], device='cuda:0'), covar=tensor([0.0660, 0.0835, 0.0669, 0.1690, 0.0721, 0.0761, 0.1461, 0.1091], device='cuda:0'), in_proj_covar=tensor([0.0616, 0.0560, 0.0514, 0.0630, 0.0412, 0.0712, 0.0767, 0.0573], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0004, 0.0004, 0.0003], device='cuda:0') 2023-05-16 17:07:50,827 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.746e+02 2.747e+02 3.282e+02 3.834e+02 8.024e+02, threshold=6.565e+02, percent-clipped=1.0 2023-05-16 17:08:09,131 INFO [finetune.py:992] (0/2) Epoch 11, batch 4350, loss[loss=0.2237, simple_loss=0.2988, pruned_loss=0.07433, over 7470.00 frames. ], tot_loss[loss=0.1686, simple_loss=0.2572, pruned_loss=0.04002, over 2363963.29 frames. ], batch size: 98, lr: 3.93e-03, grad_scale: 8.0 2023-05-16 17:08:14,891 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=226697.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 17:08:32,527 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.39 vs. limit=2.0 2023-05-16 17:08:39,088 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.2235, 4.5547, 2.7917, 2.6158, 3.9136, 2.3441, 4.0408, 2.9883], device='cuda:0'), covar=tensor([0.0749, 0.0617, 0.1299, 0.1488, 0.0273, 0.1504, 0.0413, 0.0978], device='cuda:0'), in_proj_covar=tensor([0.0192, 0.0258, 0.0180, 0.0203, 0.0145, 0.0185, 0.0201, 0.0178], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 17:08:45,346 INFO [finetune.py:992] (0/2) Epoch 11, batch 4400, loss[loss=0.1424, simple_loss=0.229, pruned_loss=0.0279, over 12334.00 frames. ], tot_loss[loss=0.1677, simple_loss=0.2563, pruned_loss=0.03956, over 2366250.86 frames. ], batch size: 31, lr: 3.93e-03, grad_scale: 8.0 2023-05-16 17:08:49,699 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=226745.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 17:08:52,002 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.3927, 2.3599, 3.1532, 4.2768, 2.3794, 4.3929, 4.3866, 4.4678], device='cuda:0'), covar=tensor([0.0119, 0.1394, 0.0509, 0.0134, 0.1305, 0.0240, 0.0152, 0.0087], device='cuda:0'), in_proj_covar=tensor([0.0121, 0.0201, 0.0181, 0.0116, 0.0187, 0.0175, 0.0175, 0.0119], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 17:09:03,160 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.812e+02 2.735e+02 3.143e+02 3.705e+02 6.632e+02, threshold=6.287e+02, percent-clipped=1.0 2023-05-16 17:09:07,355 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=4.90 vs. limit=5.0 2023-05-16 17:09:20,963 INFO [finetune.py:992] (0/2) Epoch 11, batch 4450, loss[loss=0.1774, simple_loss=0.2641, pruned_loss=0.04536, over 12109.00 frames. ], tot_loss[loss=0.168, simple_loss=0.2565, pruned_loss=0.03978, over 2371012.10 frames. ], batch size: 38, lr: 3.92e-03, grad_scale: 8.0 2023-05-16 17:09:33,634 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.32 vs. limit=2.0 2023-05-16 17:09:52,554 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 2023-05-16 17:09:57,083 INFO [finetune.py:992] (0/2) Epoch 11, batch 4500, loss[loss=0.1673, simple_loss=0.2402, pruned_loss=0.04719, over 12272.00 frames. ], tot_loss[loss=0.1686, simple_loss=0.2572, pruned_loss=0.04002, over 2368020.35 frames. ], batch size: 28, lr: 3.92e-03, grad_scale: 8.0 2023-05-16 17:10:14,845 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.996e+02 2.671e+02 3.036e+02 3.765e+02 8.868e+02, threshold=6.071e+02, percent-clipped=2.0 2023-05-16 17:10:26,471 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([1.9568, 4.1254, 4.1572, 4.3605, 2.8118, 4.0505, 2.8429, 4.0503], device='cuda:0'), covar=tensor([0.1737, 0.0689, 0.0852, 0.0513, 0.1269, 0.0605, 0.1672, 0.1108], device='cuda:0'), in_proj_covar=tensor([0.0225, 0.0266, 0.0295, 0.0355, 0.0239, 0.0240, 0.0258, 0.0376], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 17:10:27,742 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.9395, 4.8471, 4.7727, 4.7795, 4.4920, 4.9641, 4.8775, 5.1409], device='cuda:0'), covar=tensor([0.0194, 0.0157, 0.0178, 0.0326, 0.0756, 0.0301, 0.0141, 0.0170], device='cuda:0'), in_proj_covar=tensor([0.0190, 0.0195, 0.0186, 0.0243, 0.0240, 0.0212, 0.0174, 0.0221], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0005, 0.0004, 0.0004, 0.0003, 0.0005], device='cuda:0') 2023-05-16 17:10:33,105 INFO [finetune.py:992] (0/2) Epoch 11, batch 4550, loss[loss=0.1575, simple_loss=0.2475, pruned_loss=0.03379, over 12264.00 frames. ], tot_loss[loss=0.1685, simple_loss=0.2569, pruned_loss=0.04004, over 2368003.31 frames. ], batch size: 32, lr: 3.92e-03, grad_scale: 8.0 2023-05-16 17:10:33,274 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=226889.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 17:10:47,129 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.38 vs. limit=2.0 2023-05-16 17:11:01,668 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.6617, 2.8128, 3.6782, 4.6739, 3.8925, 4.5436, 4.0407, 3.4709], device='cuda:0'), covar=tensor([0.0032, 0.0365, 0.0152, 0.0029, 0.0138, 0.0075, 0.0088, 0.0285], device='cuda:0'), in_proj_covar=tensor([0.0084, 0.0121, 0.0105, 0.0077, 0.0102, 0.0116, 0.0094, 0.0136], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 17:11:07,358 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=226937.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 17:11:08,653 INFO [finetune.py:992] (0/2) Epoch 11, batch 4600, loss[loss=0.1719, simple_loss=0.2625, pruned_loss=0.04068, over 12264.00 frames. ], tot_loss[loss=0.1676, simple_loss=0.256, pruned_loss=0.0396, over 2371378.09 frames. ], batch size: 32, lr: 3.92e-03, grad_scale: 8.0 2023-05-16 17:11:26,237 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.850e+02 2.738e+02 3.287e+02 4.050e+02 6.950e+02, threshold=6.573e+02, percent-clipped=1.0 2023-05-16 17:11:44,306 INFO [finetune.py:992] (0/2) Epoch 11, batch 4650, loss[loss=0.1634, simple_loss=0.257, pruned_loss=0.03489, over 12190.00 frames. ], tot_loss[loss=0.1686, simple_loss=0.2573, pruned_loss=0.03998, over 2378307.47 frames. ], batch size: 35, lr: 3.92e-03, grad_scale: 8.0 2023-05-16 17:11:44,510 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=226989.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 17:12:09,410 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.1950, 3.4278, 3.0855, 3.0204, 2.7456, 2.6855, 3.4326, 2.2335], device='cuda:0'), covar=tensor([0.0452, 0.0150, 0.0224, 0.0224, 0.0431, 0.0339, 0.0138, 0.0479], device='cuda:0'), in_proj_covar=tensor([0.0194, 0.0164, 0.0164, 0.0187, 0.0209, 0.0201, 0.0172, 0.0204], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 17:12:10,491 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=4.94 vs. limit=5.0 2023-05-16 17:12:20,738 INFO [finetune.py:992] (0/2) Epoch 11, batch 4700, loss[loss=0.1645, simple_loss=0.2446, pruned_loss=0.04225, over 12285.00 frames. ], tot_loss[loss=0.1694, simple_loss=0.258, pruned_loss=0.04041, over 2375857.41 frames. ], batch size: 28, lr: 3.92e-03, grad_scale: 8.0 2023-05-16 17:12:28,797 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=227050.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 17:12:38,482 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.002e+02 2.827e+02 3.209e+02 4.002e+02 7.614e+02, threshold=6.418e+02, percent-clipped=2.0 2023-05-16 17:12:52,848 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.9818, 4.4233, 3.8993, 4.6518, 4.2125, 2.8272, 3.9793, 3.0392], device='cuda:0'), covar=tensor([0.0893, 0.0764, 0.1495, 0.0486, 0.1134, 0.1618, 0.1223, 0.3017], device='cuda:0'), in_proj_covar=tensor([0.0300, 0.0375, 0.0355, 0.0295, 0.0365, 0.0268, 0.0340, 0.0363], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 17:12:53,123 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.35 vs. limit=2.0 2023-05-16 17:12:53,448 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.7290, 3.8154, 3.2922, 3.2805, 3.0103, 2.9288, 3.8267, 2.4954], device='cuda:0'), covar=tensor([0.0379, 0.0155, 0.0223, 0.0182, 0.0393, 0.0378, 0.0113, 0.0460], device='cuda:0'), in_proj_covar=tensor([0.0194, 0.0165, 0.0165, 0.0188, 0.0210, 0.0202, 0.0173, 0.0205], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 17:12:56,132 INFO [finetune.py:992] (0/2) Epoch 11, batch 4750, loss[loss=0.1575, simple_loss=0.2493, pruned_loss=0.03286, over 12364.00 frames. ], tot_loss[loss=0.1689, simple_loss=0.2576, pruned_loss=0.04011, over 2374365.73 frames. ], batch size: 35, lr: 3.92e-03, grad_scale: 8.0 2023-05-16 17:13:04,448 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 2023-05-16 17:13:32,443 INFO [finetune.py:992] (0/2) Epoch 11, batch 4800, loss[loss=0.1627, simple_loss=0.2495, pruned_loss=0.0379, over 12333.00 frames. ], tot_loss[loss=0.17, simple_loss=0.2583, pruned_loss=0.0408, over 2353694.56 frames. ], batch size: 30, lr: 3.92e-03, grad_scale: 8.0 2023-05-16 17:13:50,613 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.883e+02 2.763e+02 3.237e+02 3.912e+02 6.148e+02, threshold=6.474e+02, percent-clipped=0.0 2023-05-16 17:13:56,537 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.6025, 3.8927, 3.3680, 3.3398, 3.0509, 2.9885, 3.8934, 2.4674], device='cuda:0'), covar=tensor([0.0410, 0.0115, 0.0222, 0.0199, 0.0415, 0.0334, 0.0125, 0.0474], device='cuda:0'), in_proj_covar=tensor([0.0195, 0.0164, 0.0166, 0.0188, 0.0211, 0.0202, 0.0173, 0.0205], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 17:13:57,243 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.5674, 4.9330, 3.0644, 3.1349, 4.1399, 2.8062, 4.2147, 3.6812], device='cuda:0'), covar=tensor([0.0555, 0.0522, 0.1105, 0.1217, 0.0306, 0.1272, 0.0463, 0.0683], device='cuda:0'), in_proj_covar=tensor([0.0192, 0.0258, 0.0180, 0.0203, 0.0146, 0.0185, 0.0201, 0.0179], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 17:14:02,230 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.7305, 3.8272, 3.8905, 4.3703, 3.2939, 3.8608, 2.5834, 4.1847], device='cuda:0'), covar=tensor([0.1200, 0.0795, 0.1238, 0.0926, 0.1004, 0.0633, 0.1665, 0.1444], device='cuda:0'), in_proj_covar=tensor([0.0227, 0.0266, 0.0296, 0.0356, 0.0239, 0.0242, 0.0258, 0.0379], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 17:14:05,738 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.1858, 4.0921, 4.0668, 4.3070, 2.9329, 4.0283, 2.7567, 4.0590], device='cuda:0'), covar=tensor([0.1590, 0.0656, 0.0963, 0.0714, 0.1186, 0.0562, 0.1636, 0.1299], device='cuda:0'), in_proj_covar=tensor([0.0226, 0.0266, 0.0295, 0.0356, 0.0239, 0.0242, 0.0258, 0.0378], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 17:14:08,356 INFO [finetune.py:992] (0/2) Epoch 11, batch 4850, loss[loss=0.1535, simple_loss=0.2461, pruned_loss=0.03047, over 12294.00 frames. ], tot_loss[loss=0.1692, simple_loss=0.2575, pruned_loss=0.04046, over 2361043.02 frames. ], batch size: 33, lr: 3.92e-03, grad_scale: 8.0 2023-05-16 17:14:43,917 INFO [finetune.py:992] (0/2) Epoch 11, batch 4900, loss[loss=0.2141, simple_loss=0.2888, pruned_loss=0.0697, over 8284.00 frames. ], tot_loss[loss=0.1699, simple_loss=0.2584, pruned_loss=0.04073, over 2352836.45 frames. ], batch size: 99, lr: 3.92e-03, grad_scale: 8.0 2023-05-16 17:14:44,868 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.2064, 2.6799, 3.7733, 3.2458, 3.5451, 3.3511, 2.7046, 3.6825], device='cuda:0'), covar=tensor([0.0133, 0.0324, 0.0157, 0.0242, 0.0158, 0.0167, 0.0342, 0.0127], device='cuda:0'), in_proj_covar=tensor([0.0178, 0.0206, 0.0189, 0.0190, 0.0218, 0.0166, 0.0199, 0.0188], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 17:14:53,503 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.90 vs. limit=2.0 2023-05-16 17:14:59,567 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.4140, 5.0397, 5.3877, 4.7639, 5.0808, 4.6916, 5.3875, 5.1480], device='cuda:0'), covar=tensor([0.0421, 0.0477, 0.0448, 0.0274, 0.0451, 0.0390, 0.0393, 0.0257], device='cuda:0'), in_proj_covar=tensor([0.0255, 0.0259, 0.0286, 0.0257, 0.0257, 0.0258, 0.0232, 0.0217], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 17:15:01,559 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.997e+02 2.765e+02 3.274e+02 4.144e+02 1.019e+03, threshold=6.547e+02, percent-clipped=2.0 2023-05-16 17:15:19,763 INFO [finetune.py:992] (0/2) Epoch 11, batch 4950, loss[loss=0.1764, simple_loss=0.2647, pruned_loss=0.04401, over 12122.00 frames. ], tot_loss[loss=0.1696, simple_loss=0.2586, pruned_loss=0.04035, over 2360690.85 frames. ], batch size: 30, lr: 3.92e-03, grad_scale: 8.0 2023-05-16 17:15:55,676 INFO [finetune.py:992] (0/2) Epoch 11, batch 5000, loss[loss=0.1593, simple_loss=0.2546, pruned_loss=0.032, over 10658.00 frames. ], tot_loss[loss=0.1698, simple_loss=0.259, pruned_loss=0.04032, over 2361037.25 frames. ], batch size: 68, lr: 3.92e-03, grad_scale: 8.0 2023-05-16 17:15:59,935 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=227345.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 17:16:09,871 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.2183, 5.1332, 5.0227, 5.0186, 4.7479, 5.2739, 5.1736, 5.4431], device='cuda:0'), covar=tensor([0.0250, 0.0148, 0.0186, 0.0354, 0.0750, 0.0359, 0.0162, 0.0153], device='cuda:0'), in_proj_covar=tensor([0.0191, 0.0197, 0.0188, 0.0247, 0.0242, 0.0215, 0.0175, 0.0223], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0005, 0.0004, 0.0004, 0.0003, 0.0005], device='cuda:0') 2023-05-16 17:16:13,207 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.944e+02 2.824e+02 3.448e+02 4.234e+02 1.052e+03, threshold=6.897e+02, percent-clipped=5.0 2023-05-16 17:16:31,080 INFO [finetune.py:992] (0/2) Epoch 11, batch 5050, loss[loss=0.1645, simple_loss=0.2545, pruned_loss=0.03729, over 12175.00 frames. ], tot_loss[loss=0.1692, simple_loss=0.2586, pruned_loss=0.03994, over 2362302.42 frames. ], batch size: 31, lr: 3.92e-03, grad_scale: 16.0 2023-05-16 17:16:44,476 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2023-05-16 17:16:47,757 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=227411.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 17:17:07,526 INFO [finetune.py:992] (0/2) Epoch 11, batch 5100, loss[loss=0.1893, simple_loss=0.2746, pruned_loss=0.05198, over 12303.00 frames. ], tot_loss[loss=0.1698, simple_loss=0.2587, pruned_loss=0.04043, over 2365504.92 frames. ], batch size: 34, lr: 3.92e-03, grad_scale: 16.0 2023-05-16 17:17:23,098 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.4214, 2.5108, 3.6347, 4.5051, 3.8821, 4.3387, 3.8998, 3.0336], device='cuda:0'), covar=tensor([0.0040, 0.0378, 0.0156, 0.0033, 0.0117, 0.0080, 0.0112, 0.0362], device='cuda:0'), in_proj_covar=tensor([0.0085, 0.0123, 0.0105, 0.0078, 0.0103, 0.0117, 0.0096, 0.0138], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 17:17:25,714 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.985e+02 2.628e+02 2.984e+02 3.553e+02 6.120e+02, threshold=5.968e+02, percent-clipped=0.0 2023-05-16 17:17:31,610 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=227472.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 17:17:32,350 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.8459, 3.8995, 3.2896, 3.4087, 3.1577, 2.8959, 3.9491, 2.5350], device='cuda:0'), covar=tensor([0.0389, 0.0122, 0.0234, 0.0186, 0.0335, 0.0392, 0.0113, 0.0453], device='cuda:0'), in_proj_covar=tensor([0.0192, 0.0163, 0.0165, 0.0186, 0.0207, 0.0200, 0.0171, 0.0204], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 17:17:43,459 INFO [finetune.py:992] (0/2) Epoch 11, batch 5150, loss[loss=0.1832, simple_loss=0.271, pruned_loss=0.04773, over 12047.00 frames. ], tot_loss[loss=0.1697, simple_loss=0.2586, pruned_loss=0.04041, over 2365991.67 frames. ], batch size: 42, lr: 3.92e-03, grad_scale: 16.0 2023-05-16 17:17:52,390 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=4.26 vs. limit=5.0 2023-05-16 17:18:18,557 INFO [finetune.py:992] (0/2) Epoch 11, batch 5200, loss[loss=0.1702, simple_loss=0.2719, pruned_loss=0.03425, over 12278.00 frames. ], tot_loss[loss=0.1696, simple_loss=0.2585, pruned_loss=0.04036, over 2363309.69 frames. ], batch size: 37, lr: 3.92e-03, grad_scale: 16.0 2023-05-16 17:18:36,913 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.908e+02 2.643e+02 3.110e+02 3.966e+02 7.037e+02, threshold=6.219e+02, percent-clipped=5.0 2023-05-16 17:18:54,570 INFO [finetune.py:992] (0/2) Epoch 11, batch 5250, loss[loss=0.1538, simple_loss=0.2432, pruned_loss=0.03218, over 12419.00 frames. ], tot_loss[loss=0.1687, simple_loss=0.2575, pruned_loss=0.04, over 2374326.02 frames. ], batch size: 32, lr: 3.92e-03, grad_scale: 16.0 2023-05-16 17:19:30,893 INFO [finetune.py:992] (0/2) Epoch 11, batch 5300, loss[loss=0.1891, simple_loss=0.2794, pruned_loss=0.04937, over 11268.00 frames. ], tot_loss[loss=0.1682, simple_loss=0.2569, pruned_loss=0.0398, over 2374675.23 frames. ], batch size: 55, lr: 3.92e-03, grad_scale: 16.0 2023-05-16 17:19:35,386 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=227645.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 17:19:48,924 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.109e+02 2.684e+02 3.181e+02 3.822e+02 7.017e+02, threshold=6.362e+02, percent-clipped=2.0 2023-05-16 17:19:49,991 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 2023-05-16 17:20:06,764 INFO [finetune.py:992] (0/2) Epoch 11, batch 5350, loss[loss=0.1656, simple_loss=0.2491, pruned_loss=0.04106, over 12257.00 frames. ], tot_loss[loss=0.1688, simple_loss=0.2578, pruned_loss=0.0399, over 2378723.63 frames. ], batch size: 32, lr: 3.92e-03, grad_scale: 16.0 2023-05-16 17:20:09,629 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=227693.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 17:20:09,728 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=227693.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 17:20:26,902 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.9575, 5.9468, 5.7563, 5.1967, 5.1049, 5.8727, 5.4516, 5.2522], device='cuda:0'), covar=tensor([0.0692, 0.0983, 0.0743, 0.1657, 0.0711, 0.0722, 0.1529, 0.1106], device='cuda:0'), in_proj_covar=tensor([0.0616, 0.0557, 0.0515, 0.0629, 0.0411, 0.0710, 0.0766, 0.0570], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0004, 0.0004, 0.0003], device='cuda:0') 2023-05-16 17:20:42,920 INFO [finetune.py:992] (0/2) Epoch 11, batch 5400, loss[loss=0.1885, simple_loss=0.2819, pruned_loss=0.04756, over 12131.00 frames. ], tot_loss[loss=0.1698, simple_loss=0.2589, pruned_loss=0.04031, over 2377634.14 frames. ], batch size: 39, lr: 3.92e-03, grad_scale: 16.0 2023-05-16 17:20:53,958 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=227754.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 17:21:01,144 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 2023-05-16 17:21:01,442 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.824e+02 2.711e+02 3.096e+02 3.743e+02 6.834e+02, threshold=6.192e+02, percent-clipped=3.0 2023-05-16 17:21:03,583 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=227767.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 17:21:19,124 INFO [finetune.py:992] (0/2) Epoch 11, batch 5450, loss[loss=0.1583, simple_loss=0.2523, pruned_loss=0.03217, over 12092.00 frames. ], tot_loss[loss=0.1693, simple_loss=0.2586, pruned_loss=0.04, over 2379736.58 frames. ], batch size: 32, lr: 3.92e-03, grad_scale: 16.0 2023-05-16 17:21:20,871 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.8327, 2.9003, 4.7392, 4.9387, 2.8978, 2.6138, 2.9791, 2.3364], device='cuda:0'), covar=tensor([0.1512, 0.2943, 0.0475, 0.0405, 0.1361, 0.2381, 0.2803, 0.3931], device='cuda:0'), in_proj_covar=tensor([0.0290, 0.0387, 0.0275, 0.0301, 0.0269, 0.0304, 0.0378, 0.0374], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-16 17:21:46,769 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=227827.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 17:21:54,868 INFO [finetune.py:992] (0/2) Epoch 11, batch 5500, loss[loss=0.1571, simple_loss=0.2492, pruned_loss=0.03253, over 12299.00 frames. ], tot_loss[loss=0.1695, simple_loss=0.2591, pruned_loss=0.0399, over 2384867.12 frames. ], batch size: 33, lr: 3.91e-03, grad_scale: 16.0 2023-05-16 17:22:12,511 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.791e+02 2.638e+02 3.047e+02 3.700e+02 7.213e+02, threshold=6.093e+02, percent-clipped=2.0 2023-05-16 17:22:13,028 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.35 vs. limit=2.0 2023-05-16 17:22:18,474 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.6303, 3.6036, 3.2416, 3.2048, 2.9577, 2.8039, 3.6979, 2.2297], device='cuda:0'), covar=tensor([0.0362, 0.0123, 0.0184, 0.0192, 0.0363, 0.0333, 0.0106, 0.0551], device='cuda:0'), in_proj_covar=tensor([0.0193, 0.0164, 0.0165, 0.0186, 0.0208, 0.0201, 0.0171, 0.0203], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 17:22:30,502 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=227888.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 17:22:30,989 INFO [finetune.py:992] (0/2) Epoch 11, batch 5550, loss[loss=0.1504, simple_loss=0.2371, pruned_loss=0.03185, over 12178.00 frames. ], tot_loss[loss=0.1694, simple_loss=0.2588, pruned_loss=0.04001, over 2383153.19 frames. ], batch size: 31, lr: 3.91e-03, grad_scale: 16.0 2023-05-16 17:22:35,333 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.1405, 5.9530, 5.5406, 5.4929, 6.0602, 5.3513, 5.6274, 5.4780], device='cuda:0'), covar=tensor([0.1297, 0.0971, 0.0932, 0.1931, 0.0844, 0.2198, 0.1567, 0.1177], device='cuda:0'), in_proj_covar=tensor([0.0353, 0.0487, 0.0388, 0.0442, 0.0461, 0.0429, 0.0388, 0.0380], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 17:23:07,318 INFO [finetune.py:992] (0/2) Epoch 11, batch 5600, loss[loss=0.1552, simple_loss=0.2427, pruned_loss=0.03382, over 12030.00 frames. ], tot_loss[loss=0.1691, simple_loss=0.258, pruned_loss=0.04006, over 2383143.02 frames. ], batch size: 31, lr: 3.91e-03, grad_scale: 16.0 2023-05-16 17:23:24,840 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.862e+02 2.752e+02 3.255e+02 3.896e+02 8.050e+02, threshold=6.511e+02, percent-clipped=3.0 2023-05-16 17:23:42,509 INFO [finetune.py:992] (0/2) Epoch 11, batch 5650, loss[loss=0.1992, simple_loss=0.2838, pruned_loss=0.05727, over 12001.00 frames. ], tot_loss[loss=0.17, simple_loss=0.2592, pruned_loss=0.04035, over 2384432.83 frames. ], batch size: 42, lr: 3.91e-03, grad_scale: 16.0 2023-05-16 17:23:51,337 INFO [checkpoint.py:75] (0/2) Saving checkpoint to pruned_transducer_stateless7/exp_giga_finetune/checkpoint-128000.pt 2023-05-16 17:23:58,989 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.39 vs. limit=2.0 2023-05-16 17:23:59,343 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=228007.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 17:24:06,616 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.31 vs. limit=2.0 2023-05-16 17:24:13,970 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.2236, 2.1135, 3.3792, 4.2221, 3.6519, 4.1693, 3.6139, 2.8085], device='cuda:0'), covar=tensor([0.0043, 0.0492, 0.0158, 0.0040, 0.0144, 0.0083, 0.0131, 0.0431], device='cuda:0'), in_proj_covar=tensor([0.0084, 0.0122, 0.0104, 0.0077, 0.0102, 0.0115, 0.0095, 0.0137], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 17:24:21,535 INFO [finetune.py:992] (0/2) Epoch 11, batch 5700, loss[loss=0.2232, simple_loss=0.2943, pruned_loss=0.07603, over 7882.00 frames. ], tot_loss[loss=0.1691, simple_loss=0.2585, pruned_loss=0.03986, over 2380925.42 frames. ], batch size: 98, lr: 3.91e-03, grad_scale: 16.0 2023-05-16 17:24:23,123 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.9751, 3.0529, 4.4349, 2.3666, 2.6880, 3.4021, 2.9931, 3.4963], device='cuda:0'), covar=tensor([0.0610, 0.1240, 0.0455, 0.1237, 0.1893, 0.1461, 0.1376, 0.1217], device='cuda:0'), in_proj_covar=tensor([0.0235, 0.0233, 0.0249, 0.0181, 0.0237, 0.0293, 0.0221, 0.0264], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-16 17:24:29,342 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=228049.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 17:24:34,414 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.6724, 5.5065, 5.6109, 5.6629, 5.2658, 5.3150, 5.0753, 5.5765], device='cuda:0'), covar=tensor([0.0687, 0.0585, 0.0718, 0.0530, 0.1969, 0.1188, 0.0553, 0.0904], device='cuda:0'), in_proj_covar=tensor([0.0536, 0.0699, 0.0600, 0.0617, 0.0837, 0.0730, 0.0546, 0.0475], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0005, 0.0004, 0.0004, 0.0005, 0.0005, 0.0003, 0.0003], device='cuda:0') 2023-05-16 17:24:39,951 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.723e+02 2.655e+02 3.162e+02 3.777e+02 6.559e+02, threshold=6.324e+02, percent-clipped=1.0 2023-05-16 17:24:42,178 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=228067.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 17:24:43,027 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=228068.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 17:24:51,530 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=228080.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 17:24:57,689 INFO [finetune.py:992] (0/2) Epoch 11, batch 5750, loss[loss=0.1599, simple_loss=0.2455, pruned_loss=0.03721, over 12345.00 frames. ], tot_loss[loss=0.1682, simple_loss=0.2574, pruned_loss=0.03954, over 2381991.51 frames. ], batch size: 30, lr: 3.91e-03, grad_scale: 16.0 2023-05-16 17:25:16,230 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=228115.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 17:25:23,976 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.9106, 5.5859, 5.1404, 5.2402, 5.7182, 5.0190, 5.1390, 5.1847], device='cuda:0'), covar=tensor([0.1441, 0.1125, 0.1317, 0.1845, 0.1066, 0.2309, 0.2165, 0.1230], device='cuda:0'), in_proj_covar=tensor([0.0358, 0.0493, 0.0393, 0.0445, 0.0465, 0.0433, 0.0391, 0.0383], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 17:25:27,585 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.5640, 3.6300, 3.2620, 3.2347, 2.8818, 2.7380, 3.6901, 2.3247], device='cuda:0'), covar=tensor([0.0371, 0.0132, 0.0201, 0.0200, 0.0407, 0.0372, 0.0117, 0.0495], device='cuda:0'), in_proj_covar=tensor([0.0191, 0.0162, 0.0163, 0.0184, 0.0207, 0.0199, 0.0170, 0.0202], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 17:25:33,597 INFO [finetune.py:992] (0/2) Epoch 11, batch 5800, loss[loss=0.1642, simple_loss=0.2599, pruned_loss=0.03422, over 12048.00 frames. ], tot_loss[loss=0.1678, simple_loss=0.2566, pruned_loss=0.03949, over 2381802.80 frames. ], batch size: 40, lr: 3.91e-03, grad_scale: 16.0 2023-05-16 17:25:35,269 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=228141.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 17:25:38,852 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.5978, 5.3728, 5.5347, 5.5616, 5.1556, 5.2191, 4.9800, 5.4719], device='cuda:0'), covar=tensor([0.0663, 0.0620, 0.0834, 0.0506, 0.1835, 0.1250, 0.0541, 0.1008], device='cuda:0'), in_proj_covar=tensor([0.0535, 0.0699, 0.0600, 0.0616, 0.0837, 0.0732, 0.0545, 0.0478], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0005, 0.0004, 0.0004, 0.0005, 0.0005, 0.0003, 0.0003], device='cuda:0') 2023-05-16 17:25:51,483 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.991e+02 2.760e+02 3.160e+02 4.031e+02 1.052e+03, threshold=6.321e+02, percent-clipped=2.0 2023-05-16 17:26:04,901 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=228183.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 17:26:09,242 INFO [finetune.py:992] (0/2) Epoch 11, batch 5850, loss[loss=0.1857, simple_loss=0.2724, pruned_loss=0.04952, over 12138.00 frames. ], tot_loss[loss=0.1688, simple_loss=0.2575, pruned_loss=0.04008, over 2377208.70 frames. ], batch size: 38, lr: 3.91e-03, grad_scale: 16.0 2023-05-16 17:26:45,400 INFO [finetune.py:992] (0/2) Epoch 11, batch 5900, loss[loss=0.1714, simple_loss=0.2622, pruned_loss=0.04025, over 12088.00 frames. ], tot_loss[loss=0.1698, simple_loss=0.2587, pruned_loss=0.04042, over 2373198.62 frames. ], batch size: 42, lr: 3.91e-03, grad_scale: 16.0 2023-05-16 17:27:02,783 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.805e+02 2.734e+02 3.192e+02 3.742e+02 7.430e+02, threshold=6.385e+02, percent-clipped=3.0 2023-05-16 17:27:05,128 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.4897, 5.3296, 5.3973, 5.4597, 5.0357, 5.1171, 4.9096, 5.3345], device='cuda:0'), covar=tensor([0.0674, 0.0566, 0.0743, 0.0524, 0.1878, 0.1300, 0.0545, 0.1081], device='cuda:0'), in_proj_covar=tensor([0.0536, 0.0699, 0.0601, 0.0614, 0.0839, 0.0733, 0.0546, 0.0478], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0005, 0.0004, 0.0004, 0.0005, 0.0005, 0.0003, 0.0003], device='cuda:0') 2023-05-16 17:27:10,517 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.36 vs. limit=2.0 2023-05-16 17:27:14,455 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.1277, 4.9418, 5.0576, 5.1028, 4.6892, 4.7666, 4.5827, 5.0386], device='cuda:0'), covar=tensor([0.0677, 0.0656, 0.0724, 0.0565, 0.1847, 0.1279, 0.0551, 0.1022], device='cuda:0'), in_proj_covar=tensor([0.0536, 0.0698, 0.0600, 0.0614, 0.0837, 0.0732, 0.0545, 0.0477], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0005, 0.0004, 0.0004, 0.0005, 0.0005, 0.0003, 0.0003], device='cuda:0') 2023-05-16 17:27:21,369 INFO [finetune.py:992] (0/2) Epoch 11, batch 5950, loss[loss=0.1916, simple_loss=0.2845, pruned_loss=0.04929, over 12093.00 frames. ], tot_loss[loss=0.1699, simple_loss=0.2588, pruned_loss=0.04048, over 2370473.09 frames. ], batch size: 39, lr: 3.91e-03, grad_scale: 16.0 2023-05-16 17:27:48,807 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.87 vs. limit=2.0 2023-05-16 17:27:49,777 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=228329.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 17:27:53,349 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=228333.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 17:27:57,351 INFO [finetune.py:992] (0/2) Epoch 11, batch 6000, loss[loss=0.1481, simple_loss=0.2329, pruned_loss=0.03162, over 12359.00 frames. ], tot_loss[loss=0.1691, simple_loss=0.2576, pruned_loss=0.04034, over 2365754.48 frames. ], batch size: 30, lr: 3.91e-03, grad_scale: 16.0 2023-05-16 17:27:57,352 INFO [finetune.py:1017] (0/2) Computing validation loss 2023-05-16 17:28:15,633 INFO [finetune.py:1026] (0/2) Epoch 11, validation: loss=0.3167, simple_loss=0.3935, pruned_loss=0.12, over 1020973.00 frames. 2023-05-16 17:28:15,633 INFO [finetune.py:1027] (0/2) Maximum memory allocated so far is 12827MB 2023-05-16 17:28:21,836 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 2023-05-16 17:28:22,730 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=228349.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 17:28:32,768 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=228363.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 17:28:33,395 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.822e+02 2.594e+02 3.119e+02 3.794e+02 1.855e+03, threshold=6.238e+02, percent-clipped=1.0 2023-05-16 17:28:51,426 INFO [finetune.py:992] (0/2) Epoch 11, batch 6050, loss[loss=0.1823, simple_loss=0.2785, pruned_loss=0.04309, over 11132.00 frames. ], tot_loss[loss=0.1689, simple_loss=0.2575, pruned_loss=0.04018, over 2364354.58 frames. ], batch size: 55, lr: 3.91e-03, grad_scale: 16.0 2023-05-16 17:28:52,296 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=228390.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 17:28:55,138 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=228394.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 17:28:56,989 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=228397.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 17:29:06,978 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.35 vs. limit=2.0 2023-05-16 17:29:20,417 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=228429.0, num_to_drop=1, layers_to_drop={0} 2023-05-16 17:29:25,361 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=228436.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 17:29:27,443 INFO [finetune.py:992] (0/2) Epoch 11, batch 6100, loss[loss=0.1686, simple_loss=0.258, pruned_loss=0.03964, over 12353.00 frames. ], tot_loss[loss=0.1688, simple_loss=0.2575, pruned_loss=0.04, over 2371786.89 frames. ], batch size: 36, lr: 3.91e-03, grad_scale: 16.0 2023-05-16 17:29:44,825 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.704e+02 2.773e+02 3.303e+02 3.873e+02 8.250e+02, threshold=6.606e+02, percent-clipped=3.0 2023-05-16 17:29:56,737 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-05-16 17:29:58,032 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=228483.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 17:30:02,165 INFO [finetune.py:992] (0/2) Epoch 11, batch 6150, loss[loss=0.1613, simple_loss=0.2507, pruned_loss=0.03594, over 12345.00 frames. ], tot_loss[loss=0.1696, simple_loss=0.2585, pruned_loss=0.04032, over 2367365.15 frames. ], batch size: 36, lr: 3.91e-03, grad_scale: 16.0 2023-05-16 17:30:03,023 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=228490.0, num_to_drop=1, layers_to_drop={1} 2023-05-16 17:30:20,679 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.37 vs. limit=2.0 2023-05-16 17:30:32,195 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=228531.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 17:30:37,754 INFO [finetune.py:992] (0/2) Epoch 11, batch 6200, loss[loss=0.1691, simple_loss=0.2617, pruned_loss=0.03823, over 12109.00 frames. ], tot_loss[loss=0.1703, simple_loss=0.2594, pruned_loss=0.04058, over 2372167.42 frames. ], batch size: 39, lr: 3.91e-03, grad_scale: 16.0 2023-05-16 17:30:51,279 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.8904, 5.8357, 5.6810, 5.0733, 5.0292, 5.7750, 5.3744, 5.1964], device='cuda:0'), covar=tensor([0.0746, 0.0943, 0.0630, 0.1558, 0.0769, 0.0738, 0.1513, 0.0984], device='cuda:0'), in_proj_covar=tensor([0.0613, 0.0555, 0.0508, 0.0626, 0.0410, 0.0707, 0.0763, 0.0570], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0004, 0.0004, 0.0003], device='cuda:0') 2023-05-16 17:30:55,315 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.984e+02 2.786e+02 3.277e+02 4.123e+02 6.920e+02, threshold=6.554e+02, percent-clipped=1.0 2023-05-16 17:31:13,924 INFO [finetune.py:992] (0/2) Epoch 11, batch 6250, loss[loss=0.1619, simple_loss=0.2481, pruned_loss=0.03785, over 12094.00 frames. ], tot_loss[loss=0.1701, simple_loss=0.2594, pruned_loss=0.04036, over 2370444.46 frames. ], batch size: 32, lr: 3.91e-03, grad_scale: 16.0 2023-05-16 17:31:17,001 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 2023-05-16 17:31:49,558 INFO [finetune.py:992] (0/2) Epoch 11, batch 6300, loss[loss=0.1896, simple_loss=0.2798, pruned_loss=0.04966, over 11375.00 frames. ], tot_loss[loss=0.1697, simple_loss=0.2588, pruned_loss=0.0403, over 2373525.96 frames. ], batch size: 55, lr: 3.91e-03, grad_scale: 16.0 2023-05-16 17:31:53,833 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.3225, 2.3750, 3.5677, 4.2510, 3.6904, 4.3093, 3.7413, 3.0069], device='cuda:0'), covar=tensor([0.0041, 0.0433, 0.0149, 0.0055, 0.0153, 0.0079, 0.0133, 0.0371], device='cuda:0'), in_proj_covar=tensor([0.0084, 0.0120, 0.0104, 0.0077, 0.0101, 0.0114, 0.0094, 0.0136], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 17:31:57,000 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.71 vs. limit=2.0 2023-05-16 17:32:06,747 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=228663.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 17:32:07,357 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.923e+02 2.706e+02 3.069e+02 3.698e+02 7.170e+02, threshold=6.138e+02, percent-clipped=2.0 2023-05-16 17:32:19,413 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.0277, 5.7202, 5.2395, 5.2596, 5.8146, 5.1001, 5.3371, 5.3272], device='cuda:0'), covar=tensor([0.1439, 0.0876, 0.1131, 0.1786, 0.0892, 0.2134, 0.1674, 0.1086], device='cuda:0'), in_proj_covar=tensor([0.0356, 0.0493, 0.0392, 0.0442, 0.0462, 0.0431, 0.0390, 0.0384], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 17:32:23,027 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=228685.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 17:32:25,785 INFO [finetune.py:992] (0/2) Epoch 11, batch 6350, loss[loss=0.1432, simple_loss=0.2327, pruned_loss=0.02681, over 12288.00 frames. ], tot_loss[loss=0.1688, simple_loss=0.2579, pruned_loss=0.0398, over 2375562.65 frames. ], batch size: 33, lr: 3.91e-03, grad_scale: 16.0 2023-05-16 17:32:25,871 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=228689.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 17:32:41,150 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=228711.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 17:32:51,165 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.94 vs. limit=2.0 2023-05-16 17:32:59,371 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=228736.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 17:33:01,382 INFO [finetune.py:992] (0/2) Epoch 11, batch 6400, loss[loss=0.1741, simple_loss=0.269, pruned_loss=0.03962, over 12285.00 frames. ], tot_loss[loss=0.169, simple_loss=0.2583, pruned_loss=0.03986, over 2381621.62 frames. ], batch size: 37, lr: 3.91e-03, grad_scale: 16.0 2023-05-16 17:33:18,897 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.866e+02 2.844e+02 3.278e+02 3.824e+02 1.219e+03, threshold=6.557e+02, percent-clipped=5.0 2023-05-16 17:33:33,019 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=228784.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 17:33:33,771 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=228785.0, num_to_drop=1, layers_to_drop={2} 2023-05-16 17:33:36,317 INFO [finetune.py:992] (0/2) Epoch 11, batch 6450, loss[loss=0.2144, simple_loss=0.301, pruned_loss=0.06391, over 11125.00 frames. ], tot_loss[loss=0.1695, simple_loss=0.2583, pruned_loss=0.04034, over 2370247.51 frames. ], batch size: 55, lr: 3.91e-03, grad_scale: 16.0 2023-05-16 17:34:12,185 INFO [finetune.py:992] (0/2) Epoch 11, batch 6500, loss[loss=0.1621, simple_loss=0.2427, pruned_loss=0.0407, over 12383.00 frames. ], tot_loss[loss=0.1696, simple_loss=0.258, pruned_loss=0.04061, over 2361441.28 frames. ], batch size: 30, lr: 3.91e-03, grad_scale: 16.0 2023-05-16 17:34:30,684 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.116e+02 2.836e+02 3.404e+02 4.180e+02 7.755e+02, threshold=6.809e+02, percent-clipped=2.0 2023-05-16 17:34:48,414 INFO [finetune.py:992] (0/2) Epoch 11, batch 6550, loss[loss=0.206, simple_loss=0.2909, pruned_loss=0.06057, over 12349.00 frames. ], tot_loss[loss=0.1686, simple_loss=0.2572, pruned_loss=0.04003, over 2369826.48 frames. ], batch size: 35, lr: 3.90e-03, grad_scale: 16.0 2023-05-16 17:34:54,913 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.1879, 2.0460, 2.6764, 3.1676, 2.1852, 3.2976, 3.2170, 3.3634], device='cuda:0'), covar=tensor([0.0187, 0.1167, 0.0539, 0.0224, 0.1162, 0.0333, 0.0301, 0.0144], device='cuda:0'), in_proj_covar=tensor([0.0125, 0.0204, 0.0186, 0.0119, 0.0191, 0.0179, 0.0177, 0.0120], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 17:35:19,318 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.4411, 4.6820, 4.2135, 5.0110, 4.5495, 2.8782, 4.2200, 3.1896], device='cuda:0'), covar=tensor([0.0697, 0.0823, 0.1387, 0.0507, 0.1165, 0.1691, 0.1020, 0.3051], device='cuda:0'), in_proj_covar=tensor([0.0305, 0.0385, 0.0362, 0.0305, 0.0374, 0.0273, 0.0346, 0.0370], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 17:35:21,400 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=228935.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 17:35:24,024 INFO [finetune.py:992] (0/2) Epoch 11, batch 6600, loss[loss=0.1722, simple_loss=0.2648, pruned_loss=0.03985, over 12306.00 frames. ], tot_loss[loss=0.1681, simple_loss=0.2569, pruned_loss=0.03963, over 2370101.77 frames. ], batch size: 34, lr: 3.90e-03, grad_scale: 16.0 2023-05-16 17:35:41,803 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.802e+02 2.806e+02 3.298e+02 4.086e+02 1.232e+03, threshold=6.595e+02, percent-clipped=2.0 2023-05-16 17:35:47,752 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.2568, 5.1544, 4.9602, 5.0898, 4.7252, 5.1909, 5.2117, 5.3753], device='cuda:0'), covar=tensor([0.0164, 0.0117, 0.0180, 0.0283, 0.0655, 0.0226, 0.0116, 0.0145], device='cuda:0'), in_proj_covar=tensor([0.0193, 0.0197, 0.0189, 0.0246, 0.0244, 0.0217, 0.0176, 0.0222], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0005, 0.0004, 0.0005, 0.0003, 0.0005], device='cuda:0') 2023-05-16 17:35:56,445 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.6994, 3.4768, 5.2175, 2.6943, 2.8888, 3.7686, 3.2983, 3.8707], device='cuda:0'), covar=tensor([0.0550, 0.1088, 0.0287, 0.1195, 0.1991, 0.1710, 0.1310, 0.1198], device='cuda:0'), in_proj_covar=tensor([0.0237, 0.0234, 0.0251, 0.0182, 0.0238, 0.0295, 0.0223, 0.0266], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-16 17:35:57,708 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=228985.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 17:36:00,416 INFO [finetune.py:992] (0/2) Epoch 11, batch 6650, loss[loss=0.2522, simple_loss=0.3174, pruned_loss=0.09352, over 8345.00 frames. ], tot_loss[loss=0.1688, simple_loss=0.2573, pruned_loss=0.04019, over 2362245.48 frames. ], batch size: 100, lr: 3.90e-03, grad_scale: 16.0 2023-05-16 17:36:00,550 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=228989.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 17:36:05,525 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=228996.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 17:36:30,465 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.7166, 3.8791, 3.3584, 3.3315, 3.1228, 2.9248, 3.8976, 2.4487], device='cuda:0'), covar=tensor([0.0330, 0.0113, 0.0218, 0.0178, 0.0347, 0.0365, 0.0112, 0.0454], device='cuda:0'), in_proj_covar=tensor([0.0194, 0.0164, 0.0164, 0.0186, 0.0207, 0.0203, 0.0173, 0.0204], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 17:36:32,507 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=229033.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 17:36:35,378 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=229037.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 17:36:36,755 INFO [finetune.py:992] (0/2) Epoch 11, batch 6700, loss[loss=0.1593, simple_loss=0.2446, pruned_loss=0.03698, over 12258.00 frames. ], tot_loss[loss=0.1688, simple_loss=0.2574, pruned_loss=0.04011, over 2370299.96 frames. ], batch size: 32, lr: 3.90e-03, grad_scale: 16.0 2023-05-16 17:36:44,026 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.1413, 2.4082, 3.6734, 3.0960, 3.4655, 3.2548, 2.5974, 3.6259], device='cuda:0'), covar=tensor([0.0137, 0.0358, 0.0168, 0.0249, 0.0148, 0.0167, 0.0367, 0.0117], device='cuda:0'), in_proj_covar=tensor([0.0180, 0.0208, 0.0193, 0.0193, 0.0222, 0.0170, 0.0202, 0.0190], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 17:36:54,253 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.841e+02 2.704e+02 3.020e+02 3.596e+02 6.128e+02, threshold=6.041e+02, percent-clipped=0.0 2023-05-16 17:37:09,225 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=229085.0, num_to_drop=1, layers_to_drop={2} 2023-05-16 17:37:11,848 INFO [finetune.py:992] (0/2) Epoch 11, batch 6750, loss[loss=0.1834, simple_loss=0.2753, pruned_loss=0.04576, over 10418.00 frames. ], tot_loss[loss=0.1688, simple_loss=0.2575, pruned_loss=0.04003, over 2374183.11 frames. ], batch size: 68, lr: 3.90e-03, grad_scale: 16.0 2023-05-16 17:37:13,465 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.8588, 4.7983, 4.6576, 4.7058, 4.4082, 4.8603, 4.8099, 5.0285], device='cuda:0'), covar=tensor([0.0252, 0.0149, 0.0181, 0.0326, 0.0691, 0.0256, 0.0158, 0.0180], device='cuda:0'), in_proj_covar=tensor([0.0193, 0.0196, 0.0188, 0.0245, 0.0242, 0.0215, 0.0175, 0.0222], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0005, 0.0004, 0.0004, 0.0003, 0.0005], device='cuda:0') 2023-05-16 17:37:17,534 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.81 vs. limit=2.0 2023-05-16 17:37:25,935 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=2.17 vs. limit=5.0 2023-05-16 17:37:44,133 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=229133.0, num_to_drop=1, layers_to_drop={1} 2023-05-16 17:37:48,229 INFO [finetune.py:992] (0/2) Epoch 11, batch 6800, loss[loss=0.1777, simple_loss=0.2719, pruned_loss=0.0417, over 12355.00 frames. ], tot_loss[loss=0.1682, simple_loss=0.2571, pruned_loss=0.03963, over 2376216.62 frames. ], batch size: 36, lr: 3.90e-03, grad_scale: 16.0 2023-05-16 17:38:05,932 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.2659, 4.7647, 5.1233, 5.1071, 4.9053, 5.1329, 4.9430, 2.7194], device='cuda:0'), covar=tensor([0.0103, 0.0071, 0.0074, 0.0062, 0.0045, 0.0085, 0.0080, 0.0706], device='cuda:0'), in_proj_covar=tensor([0.0071, 0.0081, 0.0084, 0.0076, 0.0062, 0.0094, 0.0084, 0.0100], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 17:38:06,451 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.994e+02 2.625e+02 3.230e+02 3.758e+02 8.293e+02, threshold=6.460e+02, percent-clipped=1.0 2023-05-16 17:38:13,003 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.4488, 5.2717, 5.3541, 5.3942, 5.0401, 5.0807, 4.8272, 5.3092], device='cuda:0'), covar=tensor([0.0730, 0.0609, 0.0769, 0.0607, 0.1872, 0.1304, 0.0524, 0.1078], device='cuda:0'), in_proj_covar=tensor([0.0539, 0.0701, 0.0603, 0.0622, 0.0844, 0.0735, 0.0549, 0.0476], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0005, 0.0004, 0.0004, 0.0005, 0.0005, 0.0003, 0.0003], device='cuda:0') 2023-05-16 17:38:22,818 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.83 vs. limit=2.0 2023-05-16 17:38:24,414 INFO [finetune.py:992] (0/2) Epoch 11, batch 6850, loss[loss=0.1523, simple_loss=0.24, pruned_loss=0.0323, over 12109.00 frames. ], tot_loss[loss=0.1677, simple_loss=0.2568, pruned_loss=0.03932, over 2380708.18 frames. ], batch size: 30, lr: 3.90e-03, grad_scale: 16.0 2023-05-16 17:39:00,056 INFO [finetune.py:992] (0/2) Epoch 11, batch 6900, loss[loss=0.1457, simple_loss=0.2337, pruned_loss=0.02886, over 12017.00 frames. ], tot_loss[loss=0.1679, simple_loss=0.2572, pruned_loss=0.03932, over 2378863.56 frames. ], batch size: 31, lr: 3.90e-03, grad_scale: 16.0 2023-05-16 17:39:02,389 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.6035, 2.7460, 3.8980, 4.5884, 3.9829, 4.5889, 3.9398, 3.1718], device='cuda:0'), covar=tensor([0.0032, 0.0331, 0.0128, 0.0033, 0.0113, 0.0057, 0.0103, 0.0350], device='cuda:0'), in_proj_covar=tensor([0.0084, 0.0121, 0.0105, 0.0078, 0.0101, 0.0116, 0.0095, 0.0137], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 17:39:11,046 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.0622, 4.9750, 4.7921, 4.8486, 4.4623, 4.9757, 4.9729, 5.1716], device='cuda:0'), covar=tensor([0.0245, 0.0153, 0.0203, 0.0340, 0.0849, 0.0317, 0.0163, 0.0181], device='cuda:0'), in_proj_covar=tensor([0.0195, 0.0198, 0.0189, 0.0249, 0.0246, 0.0218, 0.0177, 0.0224], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0005, 0.0004, 0.0005, 0.0003, 0.0005], device='cuda:0') 2023-05-16 17:39:18,635 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.913e+02 2.742e+02 3.213e+02 3.908e+02 1.203e+03, threshold=6.427e+02, percent-clipped=1.0 2023-05-16 17:39:36,536 INFO [finetune.py:992] (0/2) Epoch 11, batch 6950, loss[loss=0.2015, simple_loss=0.2841, pruned_loss=0.05943, over 10506.00 frames. ], tot_loss[loss=0.1674, simple_loss=0.2565, pruned_loss=0.03911, over 2379416.80 frames. ], batch size: 68, lr: 3.90e-03, grad_scale: 16.0 2023-05-16 17:39:38,033 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=229291.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 17:39:45,183 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=229300.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 17:40:12,870 INFO [finetune.py:992] (0/2) Epoch 11, batch 7000, loss[loss=0.1832, simple_loss=0.2798, pruned_loss=0.04325, over 12258.00 frames. ], tot_loss[loss=0.1686, simple_loss=0.2574, pruned_loss=0.03987, over 2371572.27 frames. ], batch size: 37, lr: 3.90e-03, grad_scale: 16.0 2023-05-16 17:40:28,262 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=229361.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 17:40:30,050 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.088e+02 2.844e+02 3.336e+02 3.992e+02 8.859e+02, threshold=6.672e+02, percent-clipped=3.0 2023-05-16 17:40:30,931 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.6021, 5.4205, 5.4976, 5.5618, 5.1792, 5.2064, 5.0197, 5.4981], device='cuda:0'), covar=tensor([0.0551, 0.0555, 0.0657, 0.0486, 0.1749, 0.1169, 0.0456, 0.0847], device='cuda:0'), in_proj_covar=tensor([0.0539, 0.0708, 0.0604, 0.0624, 0.0847, 0.0741, 0.0551, 0.0479], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0005, 0.0004, 0.0004, 0.0005, 0.0005, 0.0003, 0.0003], device='cuda:0') 2023-05-16 17:40:31,675 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.0719, 4.0290, 4.0413, 4.1244, 3.9042, 3.8968, 3.7860, 4.0469], device='cuda:0'), covar=tensor([0.0898, 0.0582, 0.1131, 0.0584, 0.1413, 0.1162, 0.0525, 0.0834], device='cuda:0'), in_proj_covar=tensor([0.0539, 0.0708, 0.0604, 0.0624, 0.0847, 0.0740, 0.0551, 0.0479], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0005, 0.0004, 0.0004, 0.0005, 0.0005, 0.0003, 0.0003], device='cuda:0') 2023-05-16 17:40:47,960 INFO [finetune.py:992] (0/2) Epoch 11, batch 7050, loss[loss=0.1857, simple_loss=0.2843, pruned_loss=0.0435, over 12243.00 frames. ], tot_loss[loss=0.1698, simple_loss=0.2584, pruned_loss=0.04054, over 2363820.17 frames. ], batch size: 32, lr: 3.90e-03, grad_scale: 32.0 2023-05-16 17:41:23,994 INFO [finetune.py:992] (0/2) Epoch 11, batch 7100, loss[loss=0.1652, simple_loss=0.2582, pruned_loss=0.03613, over 12147.00 frames. ], tot_loss[loss=0.1697, simple_loss=0.2584, pruned_loss=0.04044, over 2372752.67 frames. ], batch size: 34, lr: 3.90e-03, grad_scale: 32.0 2023-05-16 17:41:27,786 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=229443.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 17:41:36,059 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.4479, 4.3192, 4.2901, 4.2985, 3.9620, 4.4611, 4.3880, 4.5537], device='cuda:0'), covar=tensor([0.0251, 0.0164, 0.0189, 0.0331, 0.0755, 0.0304, 0.0167, 0.0218], device='cuda:0'), in_proj_covar=tensor([0.0195, 0.0198, 0.0189, 0.0248, 0.0244, 0.0216, 0.0177, 0.0224], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0005, 0.0004, 0.0004, 0.0003, 0.0005], device='cuda:0') 2023-05-16 17:41:42,119 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.678e+02 2.714e+02 3.188e+02 3.777e+02 1.024e+03, threshold=6.376e+02, percent-clipped=2.0 2023-05-16 17:41:51,581 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=229477.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 17:41:59,989 INFO [finetune.py:992] (0/2) Epoch 11, batch 7150, loss[loss=0.1858, simple_loss=0.2726, pruned_loss=0.04953, over 12296.00 frames. ], tot_loss[loss=0.1699, simple_loss=0.2588, pruned_loss=0.04051, over 2376677.07 frames. ], batch size: 34, lr: 3.90e-03, grad_scale: 32.0 2023-05-16 17:42:10,638 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=229504.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 17:42:34,847 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=229538.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 17:42:35,365 INFO [finetune.py:992] (0/2) Epoch 11, batch 7200, loss[loss=0.1436, simple_loss=0.2331, pruned_loss=0.02699, over 12174.00 frames. ], tot_loss[loss=0.1691, simple_loss=0.258, pruned_loss=0.04005, over 2382492.38 frames. ], batch size: 31, lr: 3.90e-03, grad_scale: 32.0 2023-05-16 17:42:53,798 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.105e+02 2.606e+02 3.037e+02 3.666e+02 6.952e+02, threshold=6.074e+02, percent-clipped=1.0 2023-05-16 17:43:03,348 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=229577.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 17:43:11,578 INFO [finetune.py:992] (0/2) Epoch 11, batch 7250, loss[loss=0.1549, simple_loss=0.2327, pruned_loss=0.03851, over 12272.00 frames. ], tot_loss[loss=0.1684, simple_loss=0.2576, pruned_loss=0.03965, over 2385651.17 frames. ], batch size: 28, lr: 3.90e-03, grad_scale: 16.0 2023-05-16 17:43:13,159 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=229591.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 17:43:43,866 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.4226, 2.8758, 3.8319, 2.1999, 2.5013, 3.0633, 2.7750, 3.2202], device='cuda:0'), covar=tensor([0.0748, 0.1233, 0.0534, 0.1387, 0.1954, 0.1406, 0.1370, 0.1162], device='cuda:0'), in_proj_covar=tensor([0.0235, 0.0233, 0.0250, 0.0181, 0.0237, 0.0292, 0.0221, 0.0264], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-16 17:43:44,547 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.2186, 3.7510, 3.8777, 4.3046, 2.7807, 3.7841, 2.4363, 3.9043], device='cuda:0'), covar=tensor([0.1684, 0.0815, 0.1059, 0.0718, 0.1269, 0.0679, 0.1967, 0.1206], device='cuda:0'), in_proj_covar=tensor([0.0225, 0.0265, 0.0290, 0.0354, 0.0236, 0.0240, 0.0256, 0.0376], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 17:43:47,413 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=229638.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 17:43:47,893 INFO [finetune.py:992] (0/2) Epoch 11, batch 7300, loss[loss=0.1869, simple_loss=0.2731, pruned_loss=0.05032, over 11328.00 frames. ], tot_loss[loss=0.1689, simple_loss=0.2575, pruned_loss=0.04012, over 2376887.25 frames. ], batch size: 55, lr: 3.90e-03, grad_scale: 8.0 2023-05-16 17:43:47,962 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=229639.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 17:43:56,422 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.9208, 2.3568, 3.5480, 2.9644, 3.3632, 3.0579, 2.4119, 3.4214], device='cuda:0'), covar=tensor([0.0150, 0.0359, 0.0149, 0.0269, 0.0161, 0.0209, 0.0375, 0.0132], device='cuda:0'), in_proj_covar=tensor([0.0177, 0.0205, 0.0190, 0.0190, 0.0218, 0.0167, 0.0199, 0.0188], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 17:43:59,905 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=229656.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 17:44:02,139 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=229659.0, num_to_drop=1, layers_to_drop={1} 2023-05-16 17:44:03,566 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.0521, 3.7339, 3.8929, 4.3781, 2.6694, 3.7510, 2.5036, 3.9024], device='cuda:0'), covar=tensor([0.1852, 0.0925, 0.0973, 0.0524, 0.1442, 0.0739, 0.2033, 0.1139], device='cuda:0'), in_proj_covar=tensor([0.0225, 0.0265, 0.0291, 0.0355, 0.0237, 0.0241, 0.0256, 0.0377], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 17:44:06,892 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.882e+02 2.730e+02 3.253e+02 4.003e+02 5.938e+02, threshold=6.506e+02, percent-clipped=0.0 2023-05-16 17:44:23,036 INFO [finetune.py:992] (0/2) Epoch 11, batch 7350, loss[loss=0.1896, simple_loss=0.2796, pruned_loss=0.0498, over 12153.00 frames. ], tot_loss[loss=0.1693, simple_loss=0.2579, pruned_loss=0.04031, over 2381126.70 frames. ], batch size: 36, lr: 3.90e-03, grad_scale: 8.0 2023-05-16 17:44:39,538 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.8938, 4.7850, 4.6797, 4.8021, 4.4667, 4.9560, 4.8149, 5.0245], device='cuda:0'), covar=tensor([0.0261, 0.0178, 0.0208, 0.0329, 0.0753, 0.0292, 0.0184, 0.0205], device='cuda:0'), in_proj_covar=tensor([0.0195, 0.0198, 0.0189, 0.0248, 0.0244, 0.0216, 0.0176, 0.0224], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0005, 0.0004, 0.0004, 0.0003, 0.0005], device='cuda:0') 2023-05-16 17:44:46,085 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=229720.0, num_to_drop=1, layers_to_drop={1} 2023-05-16 17:44:59,246 INFO [finetune.py:992] (0/2) Epoch 11, batch 7400, loss[loss=0.1834, simple_loss=0.2692, pruned_loss=0.0488, over 12072.00 frames. ], tot_loss[loss=0.1691, simple_loss=0.258, pruned_loss=0.04017, over 2382693.90 frames. ], batch size: 40, lr: 3.90e-03, grad_scale: 8.0 2023-05-16 17:45:18,589 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.960e+02 2.667e+02 3.178e+02 3.720e+02 7.407e+02, threshold=6.356e+02, percent-clipped=1.0 2023-05-16 17:45:34,711 INFO [finetune.py:992] (0/2) Epoch 11, batch 7450, loss[loss=0.1628, simple_loss=0.2577, pruned_loss=0.03397, over 12146.00 frames. ], tot_loss[loss=0.1691, simple_loss=0.2576, pruned_loss=0.04029, over 2376021.08 frames. ], batch size: 34, lr: 3.90e-03, grad_scale: 8.0 2023-05-16 17:45:40,647 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.7807, 2.9800, 4.6295, 4.8936, 2.8141, 2.7065, 3.0700, 2.2401], device='cuda:0'), covar=tensor([0.1419, 0.2840, 0.0443, 0.0370, 0.1305, 0.2113, 0.2464, 0.3624], device='cuda:0'), in_proj_covar=tensor([0.0288, 0.0385, 0.0272, 0.0300, 0.0268, 0.0301, 0.0374, 0.0369], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-16 17:45:41,892 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=229799.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 17:46:06,023 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=229833.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 17:46:10,201 INFO [finetune.py:992] (0/2) Epoch 11, batch 7500, loss[loss=0.1496, simple_loss=0.239, pruned_loss=0.03006, over 12346.00 frames. ], tot_loss[loss=0.1696, simple_loss=0.2581, pruned_loss=0.04052, over 2376976.45 frames. ], batch size: 30, lr: 3.90e-03, grad_scale: 8.0 2023-05-16 17:46:24,564 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.41 vs. limit=2.0 2023-05-16 17:46:29,849 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.955e+02 2.803e+02 3.268e+02 4.138e+02 6.772e+02, threshold=6.536e+02, percent-clipped=2.0 2023-05-16 17:46:34,406 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.9006, 2.2583, 3.5791, 2.9364, 3.4500, 3.0769, 2.3618, 3.3866], device='cuda:0'), covar=tensor([0.0157, 0.0466, 0.0162, 0.0286, 0.0156, 0.0208, 0.0465, 0.0154], device='cuda:0'), in_proj_covar=tensor([0.0177, 0.0206, 0.0191, 0.0189, 0.0217, 0.0167, 0.0199, 0.0188], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 17:46:46,286 INFO [finetune.py:992] (0/2) Epoch 11, batch 7550, loss[loss=0.1969, simple_loss=0.2756, pruned_loss=0.0591, over 12045.00 frames. ], tot_loss[loss=0.1698, simple_loss=0.2582, pruned_loss=0.04067, over 2369608.98 frames. ], batch size: 37, lr: 3.90e-03, grad_scale: 8.0 2023-05-16 17:46:50,001 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=229894.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 17:47:18,599 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=229933.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 17:47:22,838 INFO [finetune.py:992] (0/2) Epoch 11, batch 7600, loss[loss=0.1881, simple_loss=0.2688, pruned_loss=0.05367, over 11583.00 frames. ], tot_loss[loss=0.1699, simple_loss=0.2584, pruned_loss=0.04071, over 2370993.15 frames. ], batch size: 48, lr: 3.90e-03, grad_scale: 8.0 2023-05-16 17:47:24,555 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.4344, 4.1622, 4.3159, 4.5352, 3.0685, 4.0161, 3.0119, 4.3027], device='cuda:0'), covar=tensor([0.1518, 0.0649, 0.0686, 0.0522, 0.1130, 0.0599, 0.1482, 0.1119], device='cuda:0'), in_proj_covar=tensor([0.0227, 0.0267, 0.0293, 0.0357, 0.0239, 0.0243, 0.0258, 0.0379], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 17:47:34,641 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=229955.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 17:47:35,286 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=229956.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 17:47:42,262 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.918e+02 2.765e+02 3.248e+02 4.055e+02 1.373e+03, threshold=6.497e+02, percent-clipped=4.0 2023-05-16 17:47:56,924 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=229987.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 17:47:58,159 INFO [finetune.py:992] (0/2) Epoch 11, batch 7650, loss[loss=0.1642, simple_loss=0.2599, pruned_loss=0.03423, over 12148.00 frames. ], tot_loss[loss=0.1709, simple_loss=0.2594, pruned_loss=0.04119, over 2368297.45 frames. ], batch size: 36, lr: 3.89e-03, grad_scale: 8.0 2023-05-16 17:48:06,586 INFO [checkpoint.py:75] (0/2) Saving checkpoint to pruned_transducer_stateless7/exp_giga_finetune/checkpoint-130000.pt 2023-05-16 17:48:12,809 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=230004.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 17:48:20,736 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=230015.0, num_to_drop=1, layers_to_drop={2} 2023-05-16 17:48:37,932 INFO [finetune.py:992] (0/2) Epoch 11, batch 7700, loss[loss=0.1407, simple_loss=0.2234, pruned_loss=0.02897, over 12191.00 frames. ], tot_loss[loss=0.1706, simple_loss=0.2592, pruned_loss=0.04101, over 2370631.10 frames. ], batch size: 29, lr: 3.89e-03, grad_scale: 8.0 2023-05-16 17:48:44,488 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=230048.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 17:48:51,505 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.8762, 4.2083, 3.7091, 4.4592, 4.0307, 2.7641, 3.8051, 2.8201], device='cuda:0'), covar=tensor([0.0930, 0.0896, 0.1470, 0.0561, 0.1156, 0.1679, 0.1086, 0.3378], device='cuda:0'), in_proj_covar=tensor([0.0300, 0.0378, 0.0356, 0.0303, 0.0368, 0.0268, 0.0343, 0.0363], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 17:48:52,884 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-05-16 17:48:56,781 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.911e+02 2.911e+02 3.424e+02 4.220e+02 6.889e+02, threshold=6.848e+02, percent-clipped=3.0 2023-05-16 17:49:13,228 INFO [finetune.py:992] (0/2) Epoch 11, batch 7750, loss[loss=0.1742, simple_loss=0.2696, pruned_loss=0.03943, over 11196.00 frames. ], tot_loss[loss=0.1715, simple_loss=0.2602, pruned_loss=0.04145, over 2361684.44 frames. ], batch size: 55, lr: 3.89e-03, grad_scale: 8.0 2023-05-16 17:49:20,193 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=230099.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 17:49:31,511 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([6.2197, 6.1680, 6.0102, 5.5504, 5.2946, 6.1257, 5.7316, 5.5297], device='cuda:0'), covar=tensor([0.0608, 0.0785, 0.0604, 0.1560, 0.0646, 0.0679, 0.1394, 0.0921], device='cuda:0'), in_proj_covar=tensor([0.0609, 0.0553, 0.0510, 0.0624, 0.0412, 0.0709, 0.0768, 0.0567], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0004, 0.0004, 0.0003], device='cuda:0') 2023-05-16 17:49:35,449 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=3.19 vs. limit=5.0 2023-05-16 17:49:44,298 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=230133.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 17:49:48,991 INFO [finetune.py:992] (0/2) Epoch 11, batch 7800, loss[loss=0.1542, simple_loss=0.2511, pruned_loss=0.02866, over 12099.00 frames. ], tot_loss[loss=0.1712, simple_loss=0.2593, pruned_loss=0.04151, over 2361937.58 frames. ], batch size: 33, lr: 3.89e-03, grad_scale: 8.0 2023-05-16 17:49:54,832 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=230147.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 17:49:56,995 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=230150.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 17:50:07,997 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.987e+02 2.602e+02 3.039e+02 3.458e+02 5.413e+02, threshold=6.079e+02, percent-clipped=0.0 2023-05-16 17:50:19,603 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=230181.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 17:50:25,209 INFO [finetune.py:992] (0/2) Epoch 11, batch 7850, loss[loss=0.1567, simple_loss=0.2435, pruned_loss=0.03491, over 12344.00 frames. ], tot_loss[loss=0.1719, simple_loss=0.2604, pruned_loss=0.0417, over 2352917.51 frames. ], batch size: 31, lr: 3.89e-03, grad_scale: 8.0 2023-05-16 17:50:41,186 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=230211.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 17:50:56,828 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=230233.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 17:51:00,884 INFO [finetune.py:992] (0/2) Epoch 11, batch 7900, loss[loss=0.1924, simple_loss=0.2745, pruned_loss=0.05514, over 12353.00 frames. ], tot_loss[loss=0.1713, simple_loss=0.2593, pruned_loss=0.04166, over 2362379.91 frames. ], batch size: 36, lr: 3.89e-03, grad_scale: 8.0 2023-05-16 17:51:02,580 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=230241.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 17:51:09,051 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=230250.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 17:51:20,371 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.106e+02 2.861e+02 3.357e+02 3.950e+02 8.975e+02, threshold=6.714e+02, percent-clipped=3.0 2023-05-16 17:51:31,776 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=230281.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 17:51:33,355 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.2928, 3.1038, 3.0759, 3.5469, 2.6906, 3.1829, 2.6326, 3.0290], device='cuda:0'), covar=tensor([0.1362, 0.0854, 0.0838, 0.0579, 0.0908, 0.0697, 0.1417, 0.0863], device='cuda:0'), in_proj_covar=tensor([0.0228, 0.0269, 0.0295, 0.0360, 0.0240, 0.0243, 0.0260, 0.0381], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 17:51:37,363 INFO [finetune.py:992] (0/2) Epoch 11, batch 7950, loss[loss=0.1625, simple_loss=0.2482, pruned_loss=0.03841, over 12190.00 frames. ], tot_loss[loss=0.172, simple_loss=0.2595, pruned_loss=0.04229, over 2346591.58 frames. ], batch size: 31, lr: 3.89e-03, grad_scale: 8.0 2023-05-16 17:51:46,936 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=230302.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 17:51:55,951 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=230315.0, num_to_drop=1, layers_to_drop={0} 2023-05-16 17:52:13,564 INFO [finetune.py:992] (0/2) Epoch 11, batch 8000, loss[loss=0.175, simple_loss=0.2588, pruned_loss=0.04563, over 10627.00 frames. ], tot_loss[loss=0.1715, simple_loss=0.2592, pruned_loss=0.04189, over 2356950.34 frames. ], batch size: 68, lr: 3.89e-03, grad_scale: 8.0 2023-05-16 17:52:16,421 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=230343.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 17:52:30,651 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=230363.0, num_to_drop=1, layers_to_drop={1} 2023-05-16 17:52:32,618 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.719e+02 2.609e+02 2.925e+02 3.511e+02 7.142e+02, threshold=5.850e+02, percent-clipped=2.0 2023-05-16 17:52:49,070 INFO [finetune.py:992] (0/2) Epoch 11, batch 8050, loss[loss=0.1434, simple_loss=0.2296, pruned_loss=0.02861, over 12185.00 frames. ], tot_loss[loss=0.1711, simple_loss=0.2592, pruned_loss=0.04153, over 2360336.74 frames. ], batch size: 29, lr: 3.89e-03, grad_scale: 8.0 2023-05-16 17:52:54,773 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.5248, 4.3262, 4.3166, 4.5420, 3.0219, 4.2243, 2.8979, 4.1825], device='cuda:0'), covar=tensor([0.1392, 0.0608, 0.0732, 0.0646, 0.1170, 0.0491, 0.1589, 0.1190], device='cuda:0'), in_proj_covar=tensor([0.0226, 0.0267, 0.0292, 0.0357, 0.0238, 0.0241, 0.0257, 0.0378], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 17:53:21,895 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.4525, 2.4747, 3.7242, 4.4372, 3.8508, 4.4547, 3.7695, 2.8756], device='cuda:0'), covar=tensor([0.0038, 0.0395, 0.0131, 0.0040, 0.0109, 0.0074, 0.0113, 0.0387], device='cuda:0'), in_proj_covar=tensor([0.0085, 0.0121, 0.0104, 0.0077, 0.0102, 0.0115, 0.0096, 0.0136], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 17:53:25,165 INFO [finetune.py:992] (0/2) Epoch 11, batch 8100, loss[loss=0.1531, simple_loss=0.2508, pruned_loss=0.0277, over 12262.00 frames. ], tot_loss[loss=0.1707, simple_loss=0.2588, pruned_loss=0.04131, over 2364707.36 frames. ], batch size: 37, lr: 3.89e-03, grad_scale: 8.0 2023-05-16 17:53:41,878 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.7497, 5.4123, 4.9659, 4.9514, 5.5117, 4.8808, 4.9817, 4.9913], device='cuda:0'), covar=tensor([0.1494, 0.1088, 0.1205, 0.2308, 0.1068, 0.2333, 0.1921, 0.1272], device='cuda:0'), in_proj_covar=tensor([0.0364, 0.0498, 0.0397, 0.0448, 0.0465, 0.0441, 0.0398, 0.0391], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 17:53:44,596 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.745e+02 2.665e+02 3.226e+02 3.957e+02 7.795e+02, threshold=6.451e+02, percent-clipped=2.0 2023-05-16 17:53:46,253 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=230468.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 17:54:00,771 INFO [finetune.py:992] (0/2) Epoch 11, batch 8150, loss[loss=0.1929, simple_loss=0.2892, pruned_loss=0.04832, over 10674.00 frames. ], tot_loss[loss=0.1717, simple_loss=0.2602, pruned_loss=0.04156, over 2359020.37 frames. ], batch size: 68, lr: 3.89e-03, grad_scale: 8.0 2023-05-16 17:54:12,802 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=230506.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 17:54:26,363 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=230525.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 17:54:29,158 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=230529.0, num_to_drop=1, layers_to_drop={3} 2023-05-16 17:54:31,792 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=230533.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 17:54:35,906 INFO [finetune.py:992] (0/2) Epoch 11, batch 8200, loss[loss=0.1638, simple_loss=0.264, pruned_loss=0.03177, over 10620.00 frames. ], tot_loss[loss=0.1729, simple_loss=0.2614, pruned_loss=0.04226, over 2358347.47 frames. ], batch size: 68, lr: 3.89e-03, grad_scale: 8.0 2023-05-16 17:54:43,914 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=230550.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 17:54:55,578 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.898e+02 2.736e+02 3.152e+02 4.001e+02 7.952e+02, threshold=6.304e+02, percent-clipped=2.0 2023-05-16 17:55:09,958 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=230586.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 17:55:11,888 INFO [finetune.py:992] (0/2) Epoch 11, batch 8250, loss[loss=0.1634, simple_loss=0.2511, pruned_loss=0.03788, over 12337.00 frames. ], tot_loss[loss=0.1726, simple_loss=0.2609, pruned_loss=0.0421, over 2360998.07 frames. ], batch size: 30, lr: 3.89e-03, grad_scale: 8.0 2023-05-16 17:55:15,601 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=230594.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 17:55:17,692 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=230597.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 17:55:18,382 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=230598.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 17:55:45,962 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.7669, 2.8677, 4.6497, 4.8162, 2.8974, 2.5704, 2.9903, 2.3175], device='cuda:0'), covar=tensor([0.1482, 0.3214, 0.0448, 0.0383, 0.1267, 0.2517, 0.2662, 0.3896], device='cuda:0'), in_proj_covar=tensor([0.0290, 0.0388, 0.0277, 0.0303, 0.0270, 0.0304, 0.0378, 0.0373], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-16 17:55:48,508 INFO [finetune.py:992] (0/2) Epoch 11, batch 8300, loss[loss=0.1468, simple_loss=0.233, pruned_loss=0.0303, over 12185.00 frames. ], tot_loss[loss=0.1712, simple_loss=0.2599, pruned_loss=0.04128, over 2367796.69 frames. ], batch size: 31, lr: 3.89e-03, grad_scale: 8.0 2023-05-16 17:55:51,397 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=230643.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 17:56:07,451 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.816e+02 2.789e+02 3.208e+02 4.062e+02 2.917e+03, threshold=6.416e+02, percent-clipped=6.0 2023-05-16 17:56:23,404 INFO [finetune.py:992] (0/2) Epoch 11, batch 8350, loss[loss=0.1962, simple_loss=0.2929, pruned_loss=0.04978, over 12121.00 frames. ], tot_loss[loss=0.172, simple_loss=0.2608, pruned_loss=0.04153, over 2372373.62 frames. ], batch size: 39, lr: 3.89e-03, grad_scale: 8.0 2023-05-16 17:56:24,905 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=230691.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 17:56:28,850 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=2.57 vs. limit=5.0 2023-05-16 17:56:43,352 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.6788, 2.7889, 4.5424, 4.7569, 2.8551, 2.5026, 2.8130, 2.1368], device='cuda:0'), covar=tensor([0.1564, 0.2948, 0.0461, 0.0417, 0.1266, 0.2470, 0.2762, 0.3893], device='cuda:0'), in_proj_covar=tensor([0.0289, 0.0386, 0.0275, 0.0301, 0.0269, 0.0302, 0.0377, 0.0371], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-16 17:56:56,362 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.39 vs. limit=2.0 2023-05-16 17:57:00,091 INFO [finetune.py:992] (0/2) Epoch 11, batch 8400, loss[loss=0.1635, simple_loss=0.2393, pruned_loss=0.04384, over 11822.00 frames. ], tot_loss[loss=0.1725, simple_loss=0.2613, pruned_loss=0.0418, over 2358984.00 frames. ], batch size: 26, lr: 3.89e-03, grad_scale: 8.0 2023-05-16 17:57:19,204 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.9821, 4.9711, 4.8152, 4.9896, 3.7846, 5.1756, 4.9803, 5.1196], device='cuda:0'), covar=tensor([0.0293, 0.0198, 0.0216, 0.0351, 0.1484, 0.0299, 0.0203, 0.0235], device='cuda:0'), in_proj_covar=tensor([0.0194, 0.0196, 0.0188, 0.0245, 0.0242, 0.0214, 0.0176, 0.0223], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0005, 0.0004, 0.0004, 0.0003, 0.0005], device='cuda:0') 2023-05-16 17:57:19,680 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.982e+02 2.634e+02 3.184e+02 3.774e+02 7.292e+02, threshold=6.369e+02, percent-clipped=3.0 2023-05-16 17:57:35,734 INFO [finetune.py:992] (0/2) Epoch 11, batch 8450, loss[loss=0.1833, simple_loss=0.2748, pruned_loss=0.04594, over 12098.00 frames. ], tot_loss[loss=0.1723, simple_loss=0.2612, pruned_loss=0.04167, over 2368147.84 frames. ], batch size: 38, lr: 3.89e-03, grad_scale: 8.0 2023-05-16 17:57:37,960 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=230792.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 17:57:47,919 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=230806.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 17:58:00,560 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=230824.0, num_to_drop=1, layers_to_drop={0} 2023-05-16 17:58:11,077 INFO [finetune.py:992] (0/2) Epoch 11, batch 8500, loss[loss=0.1802, simple_loss=0.2696, pruned_loss=0.04539, over 12354.00 frames. ], tot_loss[loss=0.1711, simple_loss=0.2601, pruned_loss=0.0411, over 2377418.01 frames. ], batch size: 36, lr: 3.89e-03, grad_scale: 8.0 2023-05-16 17:58:21,220 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=230853.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 17:58:21,807 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=230854.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 17:58:30,852 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.930e+02 2.607e+02 3.151e+02 3.693e+02 1.366e+03, threshold=6.301e+02, percent-clipped=2.0 2023-05-16 17:58:41,382 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=230881.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 17:58:47,002 INFO [finetune.py:992] (0/2) Epoch 11, batch 8550, loss[loss=0.2396, simple_loss=0.3052, pruned_loss=0.08697, over 8056.00 frames. ], tot_loss[loss=0.1704, simple_loss=0.2591, pruned_loss=0.04084, over 2378680.12 frames. ], batch size: 97, lr: 3.89e-03, grad_scale: 8.0 2023-05-16 17:58:47,085 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=230889.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 17:58:47,860 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=230890.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 17:58:52,998 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=230897.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 17:59:13,248 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.1964, 2.5733, 3.7536, 3.1605, 3.5267, 3.3235, 2.7392, 3.5731], device='cuda:0'), covar=tensor([0.0134, 0.0350, 0.0145, 0.0250, 0.0170, 0.0165, 0.0326, 0.0145], device='cuda:0'), in_proj_covar=tensor([0.0179, 0.0208, 0.0194, 0.0191, 0.0221, 0.0171, 0.0202, 0.0191], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 17:59:22,826 INFO [finetune.py:992] (0/2) Epoch 11, batch 8600, loss[loss=0.2211, simple_loss=0.2987, pruned_loss=0.07178, over 8395.00 frames. ], tot_loss[loss=0.1716, simple_loss=0.2602, pruned_loss=0.04151, over 2363279.06 frames. ], batch size: 98, lr: 3.89e-03, grad_scale: 8.0 2023-05-16 17:59:27,046 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=230945.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 17:59:31,509 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=230951.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 17:59:41,762 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.076e+02 2.776e+02 3.433e+02 4.051e+02 7.621e+02, threshold=6.865e+02, percent-clipped=5.0 2023-05-16 17:59:57,887 INFO [finetune.py:992] (0/2) Epoch 11, batch 8650, loss[loss=0.1785, simple_loss=0.2698, pruned_loss=0.04356, over 10383.00 frames. ], tot_loss[loss=0.1708, simple_loss=0.2593, pruned_loss=0.04115, over 2359083.48 frames. ], batch size: 68, lr: 3.89e-03, grad_scale: 8.0 2023-05-16 18:00:32,377 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.5039, 3.2121, 4.8947, 2.5044, 2.6323, 3.5437, 2.9847, 3.6109], device='cuda:0'), covar=tensor([0.0453, 0.1198, 0.0386, 0.1266, 0.2060, 0.1623, 0.1404, 0.1276], device='cuda:0'), in_proj_covar=tensor([0.0232, 0.0229, 0.0247, 0.0179, 0.0235, 0.0290, 0.0218, 0.0260], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-16 18:00:34,263 INFO [finetune.py:992] (0/2) Epoch 11, batch 8700, loss[loss=0.1744, simple_loss=0.2627, pruned_loss=0.04311, over 12196.00 frames. ], tot_loss[loss=0.1703, simple_loss=0.2589, pruned_loss=0.04084, over 2356781.57 frames. ], batch size: 35, lr: 3.88e-03, grad_scale: 8.0 2023-05-16 18:00:37,434 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.57 vs. limit=2.0 2023-05-16 18:00:54,144 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.687e+02 2.745e+02 3.296e+02 3.890e+02 9.227e+02, threshold=6.593e+02, percent-clipped=2.0 2023-05-16 18:01:10,427 INFO [finetune.py:992] (0/2) Epoch 11, batch 8750, loss[loss=0.1402, simple_loss=0.223, pruned_loss=0.02869, over 12342.00 frames. ], tot_loss[loss=0.1697, simple_loss=0.2583, pruned_loss=0.04054, over 2364902.41 frames. ], batch size: 30, lr: 3.88e-03, grad_scale: 8.0 2023-05-16 18:01:35,391 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=231124.0, num_to_drop=1, layers_to_drop={0} 2023-05-16 18:01:45,814 INFO [finetune.py:992] (0/2) Epoch 11, batch 8800, loss[loss=0.1774, simple_loss=0.254, pruned_loss=0.05035, over 12287.00 frames. ], tot_loss[loss=0.1696, simple_loss=0.2581, pruned_loss=0.04049, over 2368249.94 frames. ], batch size: 33, lr: 3.88e-03, grad_scale: 8.0 2023-05-16 18:01:52,136 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=231148.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 18:02:05,562 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.994e+02 2.928e+02 3.202e+02 4.207e+02 8.695e+02, threshold=6.405e+02, percent-clipped=1.0 2023-05-16 18:02:08,980 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.31 vs. limit=2.0 2023-05-16 18:02:10,020 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=231172.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 18:02:16,268 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=231181.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 18:02:22,463 INFO [finetune.py:992] (0/2) Epoch 11, batch 8850, loss[loss=0.2077, simple_loss=0.2914, pruned_loss=0.06202, over 12064.00 frames. ], tot_loss[loss=0.1696, simple_loss=0.2583, pruned_loss=0.04048, over 2369227.03 frames. ], batch size: 37, lr: 3.88e-03, grad_scale: 8.0 2023-05-16 18:02:22,567 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=231189.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 18:02:51,083 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=231229.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 18:02:51,866 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.7079, 4.6268, 4.4975, 4.5378, 4.2686, 4.7218, 4.6944, 4.8755], device='cuda:0'), covar=tensor([0.0221, 0.0168, 0.0207, 0.0374, 0.0766, 0.0377, 0.0186, 0.0193], device='cuda:0'), in_proj_covar=tensor([0.0194, 0.0197, 0.0189, 0.0247, 0.0243, 0.0215, 0.0176, 0.0224], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0005, 0.0004, 0.0004, 0.0003, 0.0005], device='cuda:0') 2023-05-16 18:02:52,632 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.6742, 2.6610, 3.3634, 4.5935, 2.5914, 4.4599, 4.6470, 4.7708], device='cuda:0'), covar=tensor([0.0119, 0.1137, 0.0436, 0.0158, 0.1118, 0.0252, 0.0144, 0.0101], device='cuda:0'), in_proj_covar=tensor([0.0121, 0.0201, 0.0183, 0.0118, 0.0189, 0.0177, 0.0175, 0.0120], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 18:02:56,672 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=231237.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 18:02:58,107 INFO [finetune.py:992] (0/2) Epoch 11, batch 8900, loss[loss=0.1813, simple_loss=0.2662, pruned_loss=0.04826, over 11384.00 frames. ], tot_loss[loss=0.1689, simple_loss=0.2576, pruned_loss=0.04005, over 2372835.72 frames. ], batch size: 55, lr: 3.88e-03, grad_scale: 8.0 2023-05-16 18:03:03,062 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=231246.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 18:03:06,592 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=231251.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 18:03:16,927 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.867e+02 2.751e+02 3.110e+02 3.886e+02 7.148e+02, threshold=6.221e+02, percent-clipped=1.0 2023-05-16 18:03:33,148 INFO [finetune.py:992] (0/2) Epoch 11, batch 8950, loss[loss=0.1807, simple_loss=0.2757, pruned_loss=0.04284, over 11598.00 frames. ], tot_loss[loss=0.1685, simple_loss=0.2573, pruned_loss=0.03981, over 2380478.71 frames. ], batch size: 48, lr: 3.88e-03, grad_scale: 8.0 2023-05-16 18:03:50,246 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=231312.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 18:04:08,729 INFO [finetune.py:992] (0/2) Epoch 11, batch 9000, loss[loss=0.2415, simple_loss=0.3092, pruned_loss=0.08685, over 8090.00 frames. ], tot_loss[loss=0.1682, simple_loss=0.2571, pruned_loss=0.03968, over 2376265.45 frames. ], batch size: 98, lr: 3.88e-03, grad_scale: 8.0 2023-05-16 18:04:08,730 INFO [finetune.py:1017] (0/2) Computing validation loss 2023-05-16 18:04:23,211 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.4941, 4.7779, 2.8999, 2.3942, 4.3985, 3.1004, 4.3706, 3.3962], device='cuda:0'), covar=tensor([0.0567, 0.0160, 0.1069, 0.1948, 0.0133, 0.1052, 0.0233, 0.0755], device='cuda:0'), in_proj_covar=tensor([0.0187, 0.0254, 0.0176, 0.0200, 0.0139, 0.0180, 0.0198, 0.0173], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 18:04:26,868 INFO [finetune.py:1026] (0/2) Epoch 11, validation: loss=0.3233, simple_loss=0.3971, pruned_loss=0.1248, over 1020973.00 frames. 2023-05-16 18:04:26,869 INFO [finetune.py:1027] (0/2) Maximum memory allocated so far is 12827MB 2023-05-16 18:04:45,841 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.984e+02 2.737e+02 3.170e+02 3.842e+02 1.182e+03, threshold=6.339e+02, percent-clipped=5.0 2023-05-16 18:04:55,915 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.4852, 4.1306, 4.2030, 4.4586, 3.1516, 4.1269, 2.9644, 4.1536], device='cuda:0'), covar=tensor([0.1456, 0.0655, 0.0846, 0.0679, 0.1046, 0.0537, 0.1540, 0.1352], device='cuda:0'), in_proj_covar=tensor([0.0227, 0.0266, 0.0291, 0.0355, 0.0237, 0.0241, 0.0255, 0.0379], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 18:05:02,082 INFO [finetune.py:992] (0/2) Epoch 11, batch 9050, loss[loss=0.1785, simple_loss=0.2859, pruned_loss=0.03552, over 12362.00 frames. ], tot_loss[loss=0.1682, simple_loss=0.2573, pruned_loss=0.03952, over 2377881.62 frames. ], batch size: 35, lr: 3.88e-03, grad_scale: 8.0 2023-05-16 18:05:39,125 INFO [finetune.py:992] (0/2) Epoch 11, batch 9100, loss[loss=0.1537, simple_loss=0.2372, pruned_loss=0.0351, over 12291.00 frames. ], tot_loss[loss=0.1694, simple_loss=0.2583, pruned_loss=0.04021, over 2372484.85 frames. ], batch size: 28, lr: 3.88e-03, grad_scale: 8.0 2023-05-16 18:05:45,555 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=231448.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 18:05:58,184 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.861e+02 2.760e+02 3.235e+02 3.788e+02 1.428e+03, threshold=6.471e+02, percent-clipped=2.0 2023-05-16 18:06:14,295 INFO [finetune.py:992] (0/2) Epoch 11, batch 9150, loss[loss=0.1765, simple_loss=0.2704, pruned_loss=0.04133, over 11821.00 frames. ], tot_loss[loss=0.1696, simple_loss=0.2586, pruned_loss=0.04033, over 2372038.33 frames. ], batch size: 44, lr: 3.88e-03, grad_scale: 8.0 2023-05-16 18:06:19,345 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=231496.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 18:06:40,274 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=3.35 vs. limit=5.0 2023-05-16 18:06:49,688 INFO [finetune.py:992] (0/2) Epoch 11, batch 9200, loss[loss=0.1346, simple_loss=0.2243, pruned_loss=0.02244, over 12274.00 frames. ], tot_loss[loss=0.1689, simple_loss=0.2579, pruned_loss=0.03989, over 2373106.73 frames. ], batch size: 28, lr: 3.88e-03, grad_scale: 8.0 2023-05-16 18:06:54,551 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=231546.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 18:07:09,109 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.715e+02 2.671e+02 3.119e+02 3.638e+02 8.693e+02, threshold=6.238e+02, percent-clipped=1.0 2023-05-16 18:07:26,041 INFO [finetune.py:992] (0/2) Epoch 11, batch 9250, loss[loss=0.1388, simple_loss=0.2197, pruned_loss=0.02897, over 12128.00 frames. ], tot_loss[loss=0.1696, simple_loss=0.2585, pruned_loss=0.04036, over 2367577.35 frames. ], batch size: 30, lr: 3.88e-03, grad_scale: 8.0 2023-05-16 18:07:29,737 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=231594.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 18:07:39,309 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=231607.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 18:08:01,678 INFO [finetune.py:992] (0/2) Epoch 11, batch 9300, loss[loss=0.1916, simple_loss=0.2858, pruned_loss=0.04871, over 11337.00 frames. ], tot_loss[loss=0.1694, simple_loss=0.2585, pruned_loss=0.04022, over 2371530.08 frames. ], batch size: 55, lr: 3.88e-03, grad_scale: 16.0 2023-05-16 18:08:07,472 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=231647.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 18:08:20,737 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.779e+02 2.565e+02 3.061e+02 3.853e+02 6.352e+02, threshold=6.123e+02, percent-clipped=2.0 2023-05-16 18:08:37,116 INFO [finetune.py:992] (0/2) Epoch 11, batch 9350, loss[loss=0.1552, simple_loss=0.2337, pruned_loss=0.03831, over 12200.00 frames. ], tot_loss[loss=0.1687, simple_loss=0.2574, pruned_loss=0.03999, over 2374072.95 frames. ], batch size: 29, lr: 3.88e-03, grad_scale: 16.0 2023-05-16 18:08:38,659 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.9097, 5.6880, 5.3481, 5.1526, 5.8050, 5.1483, 5.2588, 5.2620], device='cuda:0'), covar=tensor([0.1493, 0.1013, 0.1017, 0.2443, 0.0962, 0.2315, 0.1893, 0.1340], device='cuda:0'), in_proj_covar=tensor([0.0361, 0.0501, 0.0394, 0.0450, 0.0467, 0.0440, 0.0399, 0.0394], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 18:08:51,717 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=231708.0, num_to_drop=1, layers_to_drop={2} 2023-05-16 18:08:52,526 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.4385, 3.0997, 4.7833, 2.3995, 2.6663, 3.5082, 3.0390, 3.6485], device='cuda:0'), covar=tensor([0.0471, 0.1209, 0.0371, 0.1343, 0.2037, 0.1601, 0.1438, 0.1223], device='cuda:0'), in_proj_covar=tensor([0.0233, 0.0230, 0.0247, 0.0181, 0.0236, 0.0291, 0.0220, 0.0262], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-16 18:08:56,674 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.7369, 2.3091, 3.2556, 3.6486, 3.4714, 3.6882, 3.3155, 2.6497], device='cuda:0'), covar=tensor([0.0061, 0.0406, 0.0165, 0.0067, 0.0117, 0.0081, 0.0128, 0.0389], device='cuda:0'), in_proj_covar=tensor([0.0086, 0.0122, 0.0105, 0.0078, 0.0102, 0.0116, 0.0096, 0.0137], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 18:09:14,126 INFO [finetune.py:992] (0/2) Epoch 11, batch 9400, loss[loss=0.1507, simple_loss=0.2411, pruned_loss=0.03016, over 12182.00 frames. ], tot_loss[loss=0.1686, simple_loss=0.2576, pruned_loss=0.03983, over 2372064.84 frames. ], batch size: 31, lr: 3.88e-03, grad_scale: 16.0 2023-05-16 18:09:16,352 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.7754, 3.6728, 3.3402, 3.2808, 3.0338, 2.8926, 3.7958, 2.4397], device='cuda:0'), covar=tensor([0.0342, 0.0144, 0.0192, 0.0196, 0.0371, 0.0381, 0.0138, 0.0398], device='cuda:0'), in_proj_covar=tensor([0.0194, 0.0165, 0.0166, 0.0188, 0.0207, 0.0204, 0.0171, 0.0203], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 18:09:33,276 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.705e+02 2.554e+02 2.912e+02 3.615e+02 7.355e+02, threshold=5.823e+02, percent-clipped=1.0 2023-05-16 18:09:49,449 INFO [finetune.py:992] (0/2) Epoch 11, batch 9450, loss[loss=0.151, simple_loss=0.2319, pruned_loss=0.03509, over 12285.00 frames. ], tot_loss[loss=0.169, simple_loss=0.2578, pruned_loss=0.04007, over 2366024.47 frames. ], batch size: 28, lr: 3.88e-03, grad_scale: 16.0 2023-05-16 18:10:24,881 INFO [finetune.py:992] (0/2) Epoch 11, batch 9500, loss[loss=0.1709, simple_loss=0.2635, pruned_loss=0.03916, over 11685.00 frames. ], tot_loss[loss=0.1699, simple_loss=0.259, pruned_loss=0.04042, over 2354599.93 frames. ], batch size: 48, lr: 3.88e-03, grad_scale: 16.0 2023-05-16 18:10:38,719 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.76 vs. limit=2.0 2023-05-16 18:10:45,347 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.029e+02 2.719e+02 3.179e+02 3.902e+02 9.199e+02, threshold=6.358e+02, percent-clipped=2.0 2023-05-16 18:10:56,904 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.2441, 3.8824, 4.0228, 4.3530, 2.9232, 3.7881, 2.4081, 4.0227], device='cuda:0'), covar=tensor([0.1498, 0.0751, 0.0794, 0.0497, 0.1076, 0.0642, 0.1857, 0.1095], device='cuda:0'), in_proj_covar=tensor([0.0228, 0.0266, 0.0292, 0.0356, 0.0238, 0.0242, 0.0257, 0.0379], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 18:11:01,619 INFO [finetune.py:992] (0/2) Epoch 11, batch 9550, loss[loss=0.1751, simple_loss=0.2687, pruned_loss=0.04081, over 12109.00 frames. ], tot_loss[loss=0.1684, simple_loss=0.2574, pruned_loss=0.0397, over 2366129.47 frames. ], batch size: 33, lr: 3.88e-03, grad_scale: 16.0 2023-05-16 18:11:04,639 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.3445, 2.2097, 3.1059, 4.2849, 2.3056, 4.2872, 4.3595, 4.4369], device='cuda:0'), covar=tensor([0.0131, 0.1395, 0.0535, 0.0158, 0.1324, 0.0220, 0.0139, 0.0093], device='cuda:0'), in_proj_covar=tensor([0.0122, 0.0202, 0.0184, 0.0117, 0.0190, 0.0177, 0.0176, 0.0120], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 18:11:14,481 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=231907.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 18:11:36,727 INFO [finetune.py:992] (0/2) Epoch 11, batch 9600, loss[loss=0.1772, simple_loss=0.2761, pruned_loss=0.03918, over 10683.00 frames. ], tot_loss[loss=0.1694, simple_loss=0.2587, pruned_loss=0.04007, over 2363290.96 frames. ], batch size: 69, lr: 3.88e-03, grad_scale: 16.0 2023-05-16 18:11:47,844 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=231955.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 18:11:55,416 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.952e+02 2.672e+02 3.125e+02 3.880e+02 1.029e+03, threshold=6.250e+02, percent-clipped=2.0 2023-05-16 18:12:08,444 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.8269, 4.7276, 4.6137, 4.6735, 4.3196, 4.8546, 4.7958, 4.9498], device='cuda:0'), covar=tensor([0.0242, 0.0166, 0.0200, 0.0343, 0.0775, 0.0292, 0.0184, 0.0185], device='cuda:0'), in_proj_covar=tensor([0.0195, 0.0198, 0.0190, 0.0249, 0.0246, 0.0216, 0.0177, 0.0225], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0005, 0.0004, 0.0004, 0.0003, 0.0005], device='cuda:0') 2023-05-16 18:12:11,938 INFO [finetune.py:992] (0/2) Epoch 11, batch 9650, loss[loss=0.1875, simple_loss=0.2776, pruned_loss=0.04873, over 11816.00 frames. ], tot_loss[loss=0.1684, simple_loss=0.2573, pruned_loss=0.03974, over 2370701.19 frames. ], batch size: 44, lr: 3.88e-03, grad_scale: 16.0 2023-05-16 18:12:12,289 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.15 vs. limit=2.0 2023-05-16 18:12:18,056 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=4.91 vs. limit=5.0 2023-05-16 18:12:20,548 INFO [checkpoint.py:75] (0/2) Saving checkpoint to pruned_transducer_stateless7/exp_giga_finetune/checkpoint-132000.pt 2023-05-16 18:12:25,404 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=232003.0, num_to_drop=1, layers_to_drop={2} 2023-05-16 18:12:43,715 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.2299, 3.6945, 3.7592, 4.0819, 2.9912, 3.6481, 2.5689, 3.7456], device='cuda:0'), covar=tensor([0.1500, 0.0762, 0.0929, 0.0636, 0.1050, 0.0662, 0.1732, 0.1037], device='cuda:0'), in_proj_covar=tensor([0.0229, 0.0267, 0.0294, 0.0358, 0.0239, 0.0243, 0.0259, 0.0380], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 18:12:51,277 INFO [finetune.py:992] (0/2) Epoch 11, batch 9700, loss[loss=0.1712, simple_loss=0.2548, pruned_loss=0.04383, over 12183.00 frames. ], tot_loss[loss=0.1693, simple_loss=0.2582, pruned_loss=0.04018, over 2375267.09 frames. ], batch size: 31, lr: 3.88e-03, grad_scale: 16.0 2023-05-16 18:12:57,648 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.5512, 5.3957, 5.5026, 5.5359, 5.1688, 5.2241, 4.9711, 5.4867], device='cuda:0'), covar=tensor([0.0663, 0.0538, 0.0735, 0.0488, 0.1868, 0.1261, 0.0489, 0.0942], device='cuda:0'), in_proj_covar=tensor([0.0545, 0.0708, 0.0612, 0.0627, 0.0851, 0.0752, 0.0556, 0.0484], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0005, 0.0004, 0.0004, 0.0005, 0.0005, 0.0003, 0.0003], device='cuda:0') 2023-05-16 18:13:10,112 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.881e+02 2.676e+02 3.220e+02 3.837e+02 6.586e+02, threshold=6.439e+02, percent-clipped=2.0 2023-05-16 18:13:20,248 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.2418, 2.5174, 3.7425, 3.2194, 3.5968, 3.2707, 2.6636, 3.5565], device='cuda:0'), covar=tensor([0.0142, 0.0375, 0.0218, 0.0244, 0.0138, 0.0183, 0.0351, 0.0174], device='cuda:0'), in_proj_covar=tensor([0.0179, 0.0206, 0.0191, 0.0189, 0.0219, 0.0168, 0.0197, 0.0189], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 18:13:26,481 INFO [finetune.py:992] (0/2) Epoch 11, batch 9750, loss[loss=0.1665, simple_loss=0.2569, pruned_loss=0.038, over 12296.00 frames. ], tot_loss[loss=0.1696, simple_loss=0.2582, pruned_loss=0.04049, over 2378297.00 frames. ], batch size: 34, lr: 3.87e-03, grad_scale: 16.0 2023-05-16 18:14:01,855 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.6344, 4.4102, 4.2407, 4.5143, 3.5139, 4.0252, 2.6731, 4.3127], device='cuda:0'), covar=tensor([0.1339, 0.0550, 0.0838, 0.0622, 0.0950, 0.0582, 0.1658, 0.1314], device='cuda:0'), in_proj_covar=tensor([0.0228, 0.0267, 0.0293, 0.0359, 0.0239, 0.0243, 0.0258, 0.0380], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 18:14:02,361 INFO [finetune.py:992] (0/2) Epoch 11, batch 9800, loss[loss=0.1332, simple_loss=0.213, pruned_loss=0.02671, over 11987.00 frames. ], tot_loss[loss=0.17, simple_loss=0.2587, pruned_loss=0.04059, over 2372121.51 frames. ], batch size: 28, lr: 3.87e-03, grad_scale: 16.0 2023-05-16 18:14:22,139 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.557e+02 2.630e+02 3.078e+02 3.591e+02 8.413e+02, threshold=6.157e+02, percent-clipped=1.0 2023-05-16 18:14:38,081 INFO [finetune.py:992] (0/2) Epoch 11, batch 9850, loss[loss=0.1796, simple_loss=0.2726, pruned_loss=0.0433, over 12021.00 frames. ], tot_loss[loss=0.1694, simple_loss=0.2582, pruned_loss=0.04026, over 2374892.47 frames. ], batch size: 40, lr: 3.87e-03, grad_scale: 16.0 2023-05-16 18:15:13,663 INFO [finetune.py:992] (0/2) Epoch 11, batch 9900, loss[loss=0.2176, simple_loss=0.2938, pruned_loss=0.07068, over 7894.00 frames. ], tot_loss[loss=0.1685, simple_loss=0.2572, pruned_loss=0.03986, over 2370928.08 frames. ], batch size: 98, lr: 3.87e-03, grad_scale: 16.0 2023-05-16 18:15:19,476 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=232247.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 18:15:32,849 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.850e+02 2.591e+02 3.141e+02 3.790e+02 6.209e+02, threshold=6.282e+02, percent-clipped=1.0 2023-05-16 18:15:49,743 INFO [finetune.py:992] (0/2) Epoch 11, batch 9950, loss[loss=0.1559, simple_loss=0.2406, pruned_loss=0.03556, over 12322.00 frames. ], tot_loss[loss=0.1691, simple_loss=0.2578, pruned_loss=0.04017, over 2364438.29 frames. ], batch size: 30, lr: 3.87e-03, grad_scale: 16.0 2023-05-16 18:15:52,331 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 2023-05-16 18:16:00,424 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=232303.0, num_to_drop=1, layers_to_drop={1} 2023-05-16 18:16:04,215 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=232308.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 18:16:25,507 INFO [finetune.py:992] (0/2) Epoch 11, batch 10000, loss[loss=0.1715, simple_loss=0.2672, pruned_loss=0.03792, over 12317.00 frames. ], tot_loss[loss=0.17, simple_loss=0.2588, pruned_loss=0.04055, over 2362676.80 frames. ], batch size: 34, lr: 3.87e-03, grad_scale: 16.0 2023-05-16 18:16:30,649 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.1548, 4.7650, 5.0453, 5.0678, 4.9140, 5.0875, 4.9061, 2.5123], device='cuda:0'), covar=tensor([0.0080, 0.0064, 0.0069, 0.0058, 0.0041, 0.0099, 0.0096, 0.0748], device='cuda:0'), in_proj_covar=tensor([0.0071, 0.0081, 0.0083, 0.0076, 0.0061, 0.0093, 0.0083, 0.0100], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 18:16:33,913 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=232351.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 18:16:44,835 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.624e+02 2.705e+02 3.211e+02 4.031e+02 1.098e+03, threshold=6.422e+02, percent-clipped=4.0 2023-05-16 18:16:55,828 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=232381.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 18:17:01,442 INFO [finetune.py:992] (0/2) Epoch 11, batch 10050, loss[loss=0.1589, simple_loss=0.258, pruned_loss=0.02992, over 12264.00 frames. ], tot_loss[loss=0.1691, simple_loss=0.2579, pruned_loss=0.04015, over 2364901.80 frames. ], batch size: 37, lr: 3.87e-03, grad_scale: 16.0 2023-05-16 18:17:38,394 INFO [finetune.py:992] (0/2) Epoch 11, batch 10100, loss[loss=0.1282, simple_loss=0.2161, pruned_loss=0.02016, over 11984.00 frames. ], tot_loss[loss=0.1697, simple_loss=0.2585, pruned_loss=0.04047, over 2363738.90 frames. ], batch size: 28, lr: 3.87e-03, grad_scale: 16.0 2023-05-16 18:17:40,770 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=232442.0, num_to_drop=1, layers_to_drop={0} 2023-05-16 18:17:57,698 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.909e+02 2.646e+02 2.968e+02 3.640e+02 9.346e+02, threshold=5.937e+02, percent-clipped=2.0 2023-05-16 18:18:04,715 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=232476.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 18:18:13,582 INFO [finetune.py:992] (0/2) Epoch 11, batch 10150, loss[loss=0.189, simple_loss=0.2781, pruned_loss=0.0499, over 12081.00 frames. ], tot_loss[loss=0.1702, simple_loss=0.2585, pruned_loss=0.04096, over 2352164.45 frames. ], batch size: 42, lr: 3.87e-03, grad_scale: 16.0 2023-05-16 18:18:32,884 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.6836, 4.3542, 4.4504, 4.5427, 4.4206, 4.5833, 4.4809, 2.5609], device='cuda:0'), covar=tensor([0.0086, 0.0071, 0.0088, 0.0065, 0.0047, 0.0099, 0.0078, 0.0745], device='cuda:0'), in_proj_covar=tensor([0.0070, 0.0080, 0.0083, 0.0075, 0.0061, 0.0092, 0.0083, 0.0098], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 18:18:34,280 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=232518.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 18:18:47,836 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=232537.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 18:18:49,010 INFO [finetune.py:992] (0/2) Epoch 11, batch 10200, loss[loss=0.1791, simple_loss=0.2679, pruned_loss=0.04513, over 11999.00 frames. ], tot_loss[loss=0.17, simple_loss=0.2585, pruned_loss=0.04076, over 2355118.61 frames. ], batch size: 40, lr: 3.87e-03, grad_scale: 16.0 2023-05-16 18:19:08,217 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.941e+02 2.823e+02 3.310e+02 4.105e+02 7.764e+02, threshold=6.621e+02, percent-clipped=2.0 2023-05-16 18:19:12,614 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.1089, 5.0538, 4.9554, 5.0104, 4.6665, 5.0502, 5.0684, 5.2128], device='cuda:0'), covar=tensor([0.0222, 0.0139, 0.0194, 0.0349, 0.0717, 0.0319, 0.0159, 0.0185], device='cuda:0'), in_proj_covar=tensor([0.0193, 0.0195, 0.0188, 0.0247, 0.0243, 0.0215, 0.0176, 0.0222], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0005, 0.0004, 0.0004, 0.0003, 0.0005], device='cuda:0') 2023-05-16 18:19:18,457 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=232579.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 18:19:25,339 INFO [finetune.py:992] (0/2) Epoch 11, batch 10250, loss[loss=0.1612, simple_loss=0.2543, pruned_loss=0.03404, over 12307.00 frames. ], tot_loss[loss=0.17, simple_loss=0.2587, pruned_loss=0.04065, over 2351559.72 frames. ], batch size: 34, lr: 3.87e-03, grad_scale: 16.0 2023-05-16 18:19:35,632 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=232603.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 18:20:01,075 INFO [finetune.py:992] (0/2) Epoch 11, batch 10300, loss[loss=0.1471, simple_loss=0.231, pruned_loss=0.03156, over 12018.00 frames. ], tot_loss[loss=0.1692, simple_loss=0.2578, pruned_loss=0.04034, over 2358575.99 frames. ], batch size: 28, lr: 3.87e-03, grad_scale: 16.0 2023-05-16 18:20:04,888 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 2023-05-16 18:20:19,849 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.043e+02 2.566e+02 2.999e+02 3.659e+02 5.204e+02, threshold=5.997e+02, percent-clipped=0.0 2023-05-16 18:20:21,352 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.0191, 4.6930, 4.7448, 4.9224, 4.6975, 4.9340, 4.9036, 2.5596], device='cuda:0'), covar=tensor([0.0079, 0.0072, 0.0105, 0.0063, 0.0056, 0.0097, 0.0092, 0.0823], device='cuda:0'), in_proj_covar=tensor([0.0070, 0.0079, 0.0082, 0.0074, 0.0061, 0.0092, 0.0083, 0.0098], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 18:20:35,895 INFO [finetune.py:992] (0/2) Epoch 11, batch 10350, loss[loss=0.1508, simple_loss=0.2347, pruned_loss=0.03351, over 12269.00 frames. ], tot_loss[loss=0.169, simple_loss=0.2574, pruned_loss=0.04023, over 2368124.03 frames. ], batch size: 32, lr: 3.87e-03, grad_scale: 16.0 2023-05-16 18:20:44,597 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.5261, 3.6235, 3.3022, 3.1831, 2.9512, 2.7280, 3.6747, 2.2502], device='cuda:0'), covar=tensor([0.0428, 0.0120, 0.0178, 0.0226, 0.0430, 0.0402, 0.0132, 0.0555], device='cuda:0'), in_proj_covar=tensor([0.0193, 0.0165, 0.0165, 0.0188, 0.0206, 0.0202, 0.0171, 0.0203], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 18:21:11,095 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=232737.0, num_to_drop=1, layers_to_drop={1} 2023-05-16 18:21:12,398 INFO [finetune.py:992] (0/2) Epoch 11, batch 10400, loss[loss=0.1489, simple_loss=0.2328, pruned_loss=0.03255, over 11987.00 frames. ], tot_loss[loss=0.1698, simple_loss=0.2587, pruned_loss=0.04045, over 2361491.27 frames. ], batch size: 28, lr: 3.87e-03, grad_scale: 16.0 2023-05-16 18:21:13,413 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.4855, 5.1110, 5.5029, 4.8844, 5.1188, 4.8530, 5.5324, 5.1159], device='cuda:0'), covar=tensor([0.0253, 0.0319, 0.0221, 0.0216, 0.0360, 0.0312, 0.0190, 0.0235], device='cuda:0'), in_proj_covar=tensor([0.0258, 0.0262, 0.0287, 0.0257, 0.0258, 0.0261, 0.0233, 0.0216], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 18:21:20,111 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.95 vs. limit=2.0 2023-05-16 18:21:31,591 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.024e+02 2.598e+02 3.079e+02 3.556e+02 7.043e+02, threshold=6.158e+02, percent-clipped=2.0 2023-05-16 18:21:42,366 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.4777, 4.2499, 4.2570, 4.6194, 3.1663, 4.0244, 2.6736, 4.2827], device='cuda:0'), covar=tensor([0.1519, 0.0672, 0.0841, 0.0671, 0.1111, 0.0586, 0.1794, 0.1235], device='cuda:0'), in_proj_covar=tensor([0.0229, 0.0268, 0.0295, 0.0359, 0.0240, 0.0243, 0.0258, 0.0380], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 18:21:42,954 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.8717, 5.5445, 5.1493, 5.2062, 5.6967, 5.0806, 5.1051, 5.1181], device='cuda:0'), covar=tensor([0.1290, 0.0965, 0.1120, 0.1732, 0.0873, 0.1992, 0.1942, 0.1143], device='cuda:0'), in_proj_covar=tensor([0.0361, 0.0499, 0.0394, 0.0445, 0.0463, 0.0443, 0.0399, 0.0392], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 18:21:47,780 INFO [finetune.py:992] (0/2) Epoch 11, batch 10450, loss[loss=0.1696, simple_loss=0.2626, pruned_loss=0.03831, over 12152.00 frames. ], tot_loss[loss=0.1691, simple_loss=0.2584, pruned_loss=0.03988, over 2369684.49 frames. ], batch size: 34, lr: 3.87e-03, grad_scale: 16.0 2023-05-16 18:21:54,373 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.8830, 3.4678, 5.2667, 2.6927, 3.0074, 3.9464, 3.4659, 4.0703], device='cuda:0'), covar=tensor([0.0426, 0.1058, 0.0310, 0.1155, 0.1764, 0.1398, 0.1233, 0.0940], device='cuda:0'), in_proj_covar=tensor([0.0236, 0.0233, 0.0251, 0.0182, 0.0237, 0.0294, 0.0222, 0.0263], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-16 18:22:18,718 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=232832.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 18:22:23,620 INFO [finetune.py:992] (0/2) Epoch 11, batch 10500, loss[loss=0.1588, simple_loss=0.2491, pruned_loss=0.03425, over 12328.00 frames. ], tot_loss[loss=0.1689, simple_loss=0.258, pruned_loss=0.03987, over 2363955.81 frames. ], batch size: 31, lr: 3.87e-03, grad_scale: 16.0 2023-05-16 18:22:43,735 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.093e+02 2.625e+02 3.024e+02 3.562e+02 9.074e+02, threshold=6.047e+02, percent-clipped=2.0 2023-05-16 18:22:49,477 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=232874.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 18:22:59,803 INFO [finetune.py:992] (0/2) Epoch 11, batch 10550, loss[loss=0.1367, simple_loss=0.2257, pruned_loss=0.02381, over 12272.00 frames. ], tot_loss[loss=0.1684, simple_loss=0.2575, pruned_loss=0.0396, over 2367443.62 frames. ], batch size: 37, lr: 3.87e-03, grad_scale: 16.0 2023-05-16 18:23:04,410 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 2023-05-16 18:23:09,833 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([6.1403, 6.0873, 5.8716, 5.3769, 5.1430, 5.9842, 5.6328, 5.3769], device='cuda:0'), covar=tensor([0.0602, 0.0907, 0.0569, 0.1505, 0.0703, 0.0710, 0.1490, 0.1070], device='cuda:0'), in_proj_covar=tensor([0.0623, 0.0557, 0.0519, 0.0635, 0.0422, 0.0717, 0.0781, 0.0574], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0004, 0.0004, 0.0003], device='cuda:0') 2023-05-16 18:23:09,858 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=232903.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 18:23:32,797 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.15 vs. limit=2.0 2023-05-16 18:23:35,253 INFO [finetune.py:992] (0/2) Epoch 11, batch 10600, loss[loss=0.1641, simple_loss=0.2555, pruned_loss=0.03636, over 12286.00 frames. ], tot_loss[loss=0.1678, simple_loss=0.2572, pruned_loss=0.03921, over 2378744.85 frames. ], batch size: 33, lr: 3.87e-03, grad_scale: 16.0 2023-05-16 18:23:43,742 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=232951.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 18:23:51,747 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.5066, 3.5298, 3.2621, 3.0980, 2.9189, 2.6708, 3.6382, 2.2801], device='cuda:0'), covar=tensor([0.0396, 0.0209, 0.0195, 0.0227, 0.0392, 0.0388, 0.0133, 0.0471], device='cuda:0'), in_proj_covar=tensor([0.0193, 0.0166, 0.0166, 0.0189, 0.0206, 0.0203, 0.0172, 0.0203], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 18:23:54,292 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.679e+02 2.546e+02 3.052e+02 3.532e+02 7.559e+02, threshold=6.103e+02, percent-clipped=1.0 2023-05-16 18:24:10,518 INFO [finetune.py:992] (0/2) Epoch 11, batch 10650, loss[loss=0.1805, simple_loss=0.2666, pruned_loss=0.04722, over 12120.00 frames. ], tot_loss[loss=0.1682, simple_loss=0.2573, pruned_loss=0.03962, over 2372225.91 frames. ], batch size: 38, lr: 3.87e-03, grad_scale: 16.0 2023-05-16 18:24:11,354 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=232990.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 18:24:32,801 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=233018.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 18:24:34,247 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.6468, 2.7372, 3.7957, 4.6064, 4.0644, 4.4749, 3.9565, 3.2242], device='cuda:0'), covar=tensor([0.0036, 0.0381, 0.0129, 0.0035, 0.0113, 0.0085, 0.0120, 0.0352], device='cuda:0'), in_proj_covar=tensor([0.0085, 0.0122, 0.0104, 0.0078, 0.0102, 0.0116, 0.0097, 0.0136], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 18:24:46,279 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=233037.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 18:24:47,546 INFO [finetune.py:992] (0/2) Epoch 11, batch 10700, loss[loss=0.1816, simple_loss=0.2752, pruned_loss=0.04399, over 12066.00 frames. ], tot_loss[loss=0.1681, simple_loss=0.2572, pruned_loss=0.03951, over 2375950.37 frames. ], batch size: 42, lr: 3.87e-03, grad_scale: 16.0 2023-05-16 18:24:56,435 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=233051.0, num_to_drop=1, layers_to_drop={3} 2023-05-16 18:25:07,316 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.769e+02 2.764e+02 3.107e+02 3.687e+02 8.150e+02, threshold=6.215e+02, percent-clipped=3.0 2023-05-16 18:25:14,531 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.9954, 5.9754, 5.7301, 5.3009, 5.0533, 5.8646, 5.4892, 5.2134], device='cuda:0'), covar=tensor([0.0860, 0.1007, 0.0635, 0.1551, 0.0657, 0.0737, 0.1748, 0.1158], device='cuda:0'), in_proj_covar=tensor([0.0620, 0.0554, 0.0515, 0.0631, 0.0420, 0.0715, 0.0778, 0.0572], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0004, 0.0004, 0.0003], device='cuda:0') 2023-05-16 18:25:16,033 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=233079.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 18:25:16,701 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.3623, 5.1644, 5.2925, 5.3086, 4.9378, 5.0300, 4.7256, 5.2355], device='cuda:0'), covar=tensor([0.0684, 0.0621, 0.0769, 0.0583, 0.1961, 0.1357, 0.0570, 0.1150], device='cuda:0'), in_proj_covar=tensor([0.0550, 0.0712, 0.0616, 0.0629, 0.0860, 0.0755, 0.0558, 0.0486], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0005, 0.0004, 0.0004, 0.0005, 0.0005, 0.0004, 0.0003], device='cuda:0') 2023-05-16 18:25:20,163 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=233085.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 18:25:22,991 INFO [finetune.py:992] (0/2) Epoch 11, batch 10750, loss[loss=0.1661, simple_loss=0.2601, pruned_loss=0.0361, over 12153.00 frames. ], tot_loss[loss=0.1692, simple_loss=0.258, pruned_loss=0.04017, over 2366954.44 frames. ], batch size: 39, lr: 3.87e-03, grad_scale: 8.0 2023-05-16 18:25:53,515 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=233132.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 18:25:58,823 INFO [finetune.py:992] (0/2) Epoch 11, batch 10800, loss[loss=0.1921, simple_loss=0.2859, pruned_loss=0.0491, over 12050.00 frames. ], tot_loss[loss=0.1697, simple_loss=0.2581, pruned_loss=0.04066, over 2363794.03 frames. ], batch size: 40, lr: 3.87e-03, grad_scale: 8.0 2023-05-16 18:26:18,143 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.1337, 2.0253, 2.3507, 2.2287, 2.2411, 2.3872, 1.9612, 2.3196], device='cuda:0'), covar=tensor([0.0124, 0.0266, 0.0170, 0.0178, 0.0119, 0.0169, 0.0252, 0.0133], device='cuda:0'), in_proj_covar=tensor([0.0178, 0.0206, 0.0191, 0.0189, 0.0220, 0.0169, 0.0198, 0.0189], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 18:26:19,299 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.949e+02 2.655e+02 3.240e+02 3.813e+02 6.450e+02, threshold=6.480e+02, percent-clipped=1.0 2023-05-16 18:26:22,626 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=4.52 vs. limit=5.0 2023-05-16 18:26:24,330 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=233174.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 18:26:27,889 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.0650, 4.7460, 5.0737, 4.4181, 4.7445, 4.3933, 5.0246, 4.7274], device='cuda:0'), covar=tensor([0.0329, 0.0403, 0.0339, 0.0294, 0.0391, 0.0381, 0.0349, 0.0379], device='cuda:0'), in_proj_covar=tensor([0.0258, 0.0261, 0.0287, 0.0257, 0.0259, 0.0261, 0.0234, 0.0215], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 18:26:28,472 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=233180.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 18:26:34,608 INFO [finetune.py:992] (0/2) Epoch 11, batch 10850, loss[loss=0.1601, simple_loss=0.2461, pruned_loss=0.0371, over 12096.00 frames. ], tot_loss[loss=0.1693, simple_loss=0.2575, pruned_loss=0.04054, over 2368179.14 frames. ], batch size: 33, lr: 3.86e-03, grad_scale: 8.0 2023-05-16 18:27:00,370 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=233222.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 18:27:12,429 INFO [finetune.py:992] (0/2) Epoch 11, batch 10900, loss[loss=0.1782, simple_loss=0.2762, pruned_loss=0.04008, over 11691.00 frames. ], tot_loss[loss=0.169, simple_loss=0.2576, pruned_loss=0.04017, over 2370556.51 frames. ], batch size: 48, lr: 3.86e-03, grad_scale: 8.0 2023-05-16 18:27:26,891 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=233259.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 18:27:32,450 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.889e+02 2.720e+02 3.240e+02 3.874e+02 5.388e+02, threshold=6.479e+02, percent-clipped=0.0 2023-05-16 18:27:48,681 INFO [finetune.py:992] (0/2) Epoch 11, batch 10950, loss[loss=0.1934, simple_loss=0.2845, pruned_loss=0.05111, over 11672.00 frames. ], tot_loss[loss=0.1699, simple_loss=0.2587, pruned_loss=0.04053, over 2363830.99 frames. ], batch size: 48, lr: 3.86e-03, grad_scale: 8.0 2023-05-16 18:28:11,403 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=233320.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 18:28:24,716 INFO [finetune.py:992] (0/2) Epoch 11, batch 11000, loss[loss=0.1753, simple_loss=0.2725, pruned_loss=0.039, over 12035.00 frames. ], tot_loss[loss=0.1709, simple_loss=0.2599, pruned_loss=0.04097, over 2364152.18 frames. ], batch size: 40, lr: 3.86e-03, grad_scale: 8.0 2023-05-16 18:28:29,803 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=233346.0, num_to_drop=1, layers_to_drop={2} 2023-05-16 18:28:44,641 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.727e+02 2.841e+02 3.437e+02 4.261e+02 8.985e+02, threshold=6.874e+02, percent-clipped=4.0 2023-05-16 18:28:49,555 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=233374.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 18:28:50,308 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.9183, 4.9131, 4.7987, 4.8196, 4.1656, 4.9975, 4.9982, 5.0849], device='cuda:0'), covar=tensor([0.0283, 0.0155, 0.0200, 0.0336, 0.1020, 0.0307, 0.0167, 0.0229], device='cuda:0'), in_proj_covar=tensor([0.0192, 0.0195, 0.0188, 0.0248, 0.0242, 0.0217, 0.0176, 0.0223], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0005, 0.0004, 0.0004, 0.0003, 0.0005], device='cuda:0') 2023-05-16 18:28:59,923 INFO [finetune.py:992] (0/2) Epoch 11, batch 11050, loss[loss=0.1614, simple_loss=0.2553, pruned_loss=0.03374, over 12085.00 frames. ], tot_loss[loss=0.1742, simple_loss=0.2631, pruned_loss=0.04269, over 2331546.44 frames. ], batch size: 32, lr: 3.86e-03, grad_scale: 8.0 2023-05-16 18:29:36,232 INFO [finetune.py:992] (0/2) Epoch 11, batch 11100, loss[loss=0.219, simple_loss=0.2916, pruned_loss=0.07317, over 8090.00 frames. ], tot_loss[loss=0.1788, simple_loss=0.2669, pruned_loss=0.04531, over 2273369.38 frames. ], batch size: 98, lr: 3.86e-03, grad_scale: 8.0 2023-05-16 18:29:55,778 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.847e+02 3.180e+02 3.861e+02 4.760e+02 8.899e+02, threshold=7.722e+02, percent-clipped=6.0 2023-05-16 18:30:11,651 INFO [finetune.py:992] (0/2) Epoch 11, batch 11150, loss[loss=0.1816, simple_loss=0.2721, pruned_loss=0.04553, over 11449.00 frames. ], tot_loss[loss=0.1841, simple_loss=0.2716, pruned_loss=0.04829, over 2220467.08 frames. ], batch size: 25, lr: 3.86e-03, grad_scale: 8.0 2023-05-16 18:30:47,988 INFO [finetune.py:992] (0/2) Epoch 11, batch 11200, loss[loss=0.2291, simple_loss=0.3213, pruned_loss=0.06848, over 12060.00 frames. ], tot_loss[loss=0.1919, simple_loss=0.2786, pruned_loss=0.05262, over 2158606.85 frames. ], batch size: 42, lr: 3.86e-03, grad_scale: 8.0 2023-05-16 18:31:07,937 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.326e+02 3.488e+02 4.042e+02 5.103e+02 1.073e+03, threshold=8.084e+02, percent-clipped=2.0 2023-05-16 18:31:23,956 INFO [finetune.py:992] (0/2) Epoch 11, batch 11250, loss[loss=0.227, simple_loss=0.3187, pruned_loss=0.06768, over 11116.00 frames. ], tot_loss[loss=0.2012, simple_loss=0.2866, pruned_loss=0.05788, over 2098308.17 frames. ], batch size: 55, lr: 3.86e-03, grad_scale: 8.0 2023-05-16 18:31:35,759 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=5.01 vs. limit=5.0 2023-05-16 18:31:42,080 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=233615.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 18:31:58,985 INFO [finetune.py:992] (0/2) Epoch 11, batch 11300, loss[loss=0.2603, simple_loss=0.3308, pruned_loss=0.09491, over 7119.00 frames. ], tot_loss[loss=0.2083, simple_loss=0.2927, pruned_loss=0.062, over 2035476.95 frames. ], batch size: 99, lr: 3.86e-03, grad_scale: 8.0 2023-05-16 18:31:59,166 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=233639.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 18:32:03,963 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=233646.0, num_to_drop=1, layers_to_drop={1} 2023-05-16 18:32:04,317 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.61 vs. limit=2.0 2023-05-16 18:32:07,545 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.4064, 3.1146, 2.9376, 3.4801, 2.6023, 3.1521, 2.5175, 2.7652], device='cuda:0'), covar=tensor([0.1535, 0.0806, 0.0770, 0.0520, 0.1036, 0.0760, 0.1577, 0.0586], device='cuda:0'), in_proj_covar=tensor([0.0221, 0.0258, 0.0286, 0.0345, 0.0232, 0.0236, 0.0251, 0.0367], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 18:32:19,034 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.452e+02 3.573e+02 4.203e+02 4.687e+02 1.009e+03, threshold=8.407e+02, percent-clipped=1.0 2023-05-16 18:32:24,027 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=233674.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 18:32:26,750 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.5508, 3.8150, 3.4773, 3.3467, 3.1244, 2.9168, 3.6834, 2.2982], device='cuda:0'), covar=tensor([0.0389, 0.0158, 0.0151, 0.0184, 0.0332, 0.0348, 0.0141, 0.0521], device='cuda:0'), in_proj_covar=tensor([0.0192, 0.0162, 0.0163, 0.0185, 0.0204, 0.0201, 0.0171, 0.0201], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 18:32:33,996 INFO [finetune.py:992] (0/2) Epoch 11, batch 11350, loss[loss=0.3072, simple_loss=0.3623, pruned_loss=0.1261, over 6677.00 frames. ], tot_loss[loss=0.2129, simple_loss=0.2967, pruned_loss=0.06458, over 1990055.82 frames. ], batch size: 98, lr: 3.86e-03, grad_scale: 8.0 2023-05-16 18:32:37,456 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=233694.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 18:32:41,736 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=233700.0, num_to_drop=1, layers_to_drop={3} 2023-05-16 18:32:48,939 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=233710.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 18:32:53,049 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.6073, 4.5392, 4.5981, 4.6279, 4.3618, 4.4353, 4.2688, 4.5668], device='cuda:0'), covar=tensor([0.0843, 0.0594, 0.0887, 0.0653, 0.1705, 0.1261, 0.0592, 0.0869], device='cuda:0'), in_proj_covar=tensor([0.0529, 0.0679, 0.0588, 0.0596, 0.0814, 0.0721, 0.0536, 0.0466], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0005, 0.0004, 0.0003, 0.0003], device='cuda:0') 2023-05-16 18:32:57,144 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=233722.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 18:33:00,043 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.3145, 2.9069, 3.6128, 2.2448, 2.6185, 3.0781, 2.8216, 3.1911], device='cuda:0'), covar=tensor([0.0519, 0.1018, 0.0284, 0.1276, 0.1642, 0.1312, 0.1157, 0.0934], device='cuda:0'), in_proj_covar=tensor([0.0231, 0.0229, 0.0242, 0.0177, 0.0232, 0.0287, 0.0217, 0.0255], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:0') 2023-05-16 18:33:09,359 INFO [finetune.py:992] (0/2) Epoch 11, batch 11400, loss[loss=0.2717, simple_loss=0.3414, pruned_loss=0.1009, over 7041.00 frames. ], tot_loss[loss=0.2178, simple_loss=0.3008, pruned_loss=0.0674, over 1934334.39 frames. ], batch size: 98, lr: 3.86e-03, grad_scale: 8.0 2023-05-16 18:33:27,627 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.7710, 3.5219, 3.6000, 3.7610, 3.4434, 3.7884, 3.7809, 3.8478], device='cuda:0'), covar=tensor([0.0173, 0.0160, 0.0166, 0.0272, 0.0510, 0.0263, 0.0153, 0.0211], device='cuda:0'), in_proj_covar=tensor([0.0179, 0.0182, 0.0176, 0.0230, 0.0227, 0.0202, 0.0165, 0.0210], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:0') 2023-05-16 18:33:28,734 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.455e+02 3.541e+02 4.081e+02 5.202e+02 1.297e+03, threshold=8.162e+02, percent-clipped=1.0 2023-05-16 18:33:31,625 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=233771.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 18:33:43,588 INFO [finetune.py:992] (0/2) Epoch 11, batch 11450, loss[loss=0.1814, simple_loss=0.2731, pruned_loss=0.04487, over 12111.00 frames. ], tot_loss[loss=0.2229, simple_loss=0.3045, pruned_loss=0.07066, over 1882392.54 frames. ], batch size: 33, lr: 3.86e-03, grad_scale: 8.0 2023-05-16 18:34:19,043 INFO [finetune.py:992] (0/2) Epoch 11, batch 11500, loss[loss=0.2193, simple_loss=0.3025, pruned_loss=0.06806, over 12012.00 frames. ], tot_loss[loss=0.2263, simple_loss=0.3073, pruned_loss=0.07264, over 1853604.57 frames. ], batch size: 40, lr: 3.86e-03, grad_scale: 8.0 2023-05-16 18:34:38,557 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.233e+02 3.519e+02 4.029e+02 4.678e+02 7.673e+02, threshold=8.058e+02, percent-clipped=0.0 2023-05-16 18:34:54,111 INFO [finetune.py:992] (0/2) Epoch 11, batch 11550, loss[loss=0.2322, simple_loss=0.3025, pruned_loss=0.08094, over 6285.00 frames. ], tot_loss[loss=0.2276, simple_loss=0.3081, pruned_loss=0.07354, over 1828118.98 frames. ], batch size: 99, lr: 3.86e-03, grad_scale: 8.0 2023-05-16 18:35:07,004 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.7163, 3.0461, 2.3630, 2.1880, 2.7869, 2.2475, 3.0401, 2.5588], device='cuda:0'), covar=tensor([0.0607, 0.0530, 0.0926, 0.1512, 0.0251, 0.1217, 0.0427, 0.0768], device='cuda:0'), in_proj_covar=tensor([0.0184, 0.0245, 0.0171, 0.0195, 0.0137, 0.0178, 0.0191, 0.0170], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 18:35:08,444 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.2723, 3.2668, 3.1626, 2.9576, 2.8222, 2.6147, 3.0558, 1.9899], device='cuda:0'), covar=tensor([0.0425, 0.0121, 0.0128, 0.0189, 0.0288, 0.0278, 0.0195, 0.0517], device='cuda:0'), in_proj_covar=tensor([0.0189, 0.0158, 0.0161, 0.0182, 0.0201, 0.0197, 0.0168, 0.0198], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 18:35:13,103 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=233915.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 18:35:23,147 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.53 vs. limit=2.0 2023-05-16 18:35:28,831 INFO [finetune.py:992] (0/2) Epoch 11, batch 11600, loss[loss=0.266, simple_loss=0.3328, pruned_loss=0.09961, over 7552.00 frames. ], tot_loss[loss=0.2295, simple_loss=0.3094, pruned_loss=0.07483, over 1813920.12 frames. ], batch size: 100, lr: 3.86e-03, grad_scale: 8.0 2023-05-16 18:35:37,642 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.8928, 4.8141, 4.8894, 4.8966, 4.5776, 4.6709, 4.5000, 4.8075], device='cuda:0'), covar=tensor([0.0734, 0.0532, 0.0802, 0.0562, 0.1781, 0.1149, 0.0569, 0.1037], device='cuda:0'), in_proj_covar=tensor([0.0512, 0.0660, 0.0575, 0.0581, 0.0791, 0.0701, 0.0522, 0.0455], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0005, 0.0004, 0.0003, 0.0003], device='cuda:0') 2023-05-16 18:35:45,761 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=233963.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 18:35:48,495 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.535e+02 3.403e+02 3.814e+02 4.898e+02 8.762e+02, threshold=7.629e+02, percent-clipped=1.0 2023-05-16 18:36:05,419 INFO [finetune.py:992] (0/2) Epoch 11, batch 11650, loss[loss=0.2664, simple_loss=0.3386, pruned_loss=0.09715, over 7038.00 frames. ], tot_loss[loss=0.2294, simple_loss=0.3089, pruned_loss=0.07495, over 1811870.66 frames. ], batch size: 101, lr: 3.86e-03, grad_scale: 8.0 2023-05-16 18:36:09,908 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=233995.0, num_to_drop=1, layers_to_drop={2} 2023-05-16 18:36:11,575 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.93 vs. limit=2.0 2023-05-16 18:36:13,475 INFO [checkpoint.py:75] (0/2) Saving checkpoint to pruned_transducer_stateless7/exp_giga_finetune/checkpoint-134000.pt 2023-05-16 18:36:33,973 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=234025.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 18:36:43,277 INFO [finetune.py:992] (0/2) Epoch 11, batch 11700, loss[loss=0.1986, simple_loss=0.2874, pruned_loss=0.05492, over 11260.00 frames. ], tot_loss[loss=0.2309, simple_loss=0.3093, pruned_loss=0.07625, over 1778493.75 frames. ], batch size: 55, lr: 3.86e-03, grad_scale: 8.0 2023-05-16 18:36:56,630 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=4.65 vs. limit=5.0 2023-05-16 18:37:02,232 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=234066.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 18:37:02,775 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.425e+02 3.573e+02 4.120e+02 4.948e+02 7.827e+02, threshold=8.240e+02, percent-clipped=1.0 2023-05-16 18:37:11,027 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([1.9887, 2.2370, 2.2840, 2.2272, 2.0754, 1.9781, 2.2261, 1.6519], device='cuda:0'), covar=tensor([0.0306, 0.0167, 0.0204, 0.0190, 0.0300, 0.0240, 0.0164, 0.0377], device='cuda:0'), in_proj_covar=tensor([0.0187, 0.0156, 0.0158, 0.0180, 0.0199, 0.0194, 0.0166, 0.0196], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 18:37:16,511 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=234086.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 18:37:18,366 INFO [finetune.py:992] (0/2) Epoch 11, batch 11750, loss[loss=0.2281, simple_loss=0.2997, pruned_loss=0.07825, over 7080.00 frames. ], tot_loss[loss=0.2316, simple_loss=0.3094, pruned_loss=0.07686, over 1754972.44 frames. ], batch size: 97, lr: 3.86e-03, grad_scale: 8.0 2023-05-16 18:37:46,236 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.5432, 5.5348, 5.3382, 4.8960, 4.8022, 5.5024, 5.2052, 4.9930], device='cuda:0'), covar=tensor([0.0659, 0.0852, 0.0574, 0.1413, 0.0939, 0.0641, 0.1278, 0.0918], device='cuda:0'), in_proj_covar=tensor([0.0583, 0.0523, 0.0485, 0.0597, 0.0395, 0.0666, 0.0718, 0.0537], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002], device='cuda:0') 2023-05-16 18:37:53,762 INFO [finetune.py:992] (0/2) Epoch 11, batch 11800, loss[loss=0.2225, simple_loss=0.3178, pruned_loss=0.0636, over 10246.00 frames. ], tot_loss[loss=0.2339, simple_loss=0.3115, pruned_loss=0.07817, over 1738844.77 frames. ], batch size: 68, lr: 3.86e-03, grad_scale: 8.0 2023-05-16 18:38:11,412 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.6866, 5.2579, 4.9028, 5.0052, 5.3518, 4.7807, 4.9551, 4.8608], device='cuda:0'), covar=tensor([0.1304, 0.0900, 0.0983, 0.1605, 0.0800, 0.2079, 0.1557, 0.1109], device='cuda:0'), in_proj_covar=tensor([0.0342, 0.0476, 0.0378, 0.0423, 0.0442, 0.0417, 0.0377, 0.0369], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002], device='cuda:0') 2023-05-16 18:38:13,322 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.460e+02 3.496e+02 4.148e+02 4.897e+02 9.284e+02, threshold=8.296e+02, percent-clipped=3.0 2023-05-16 18:38:28,016 INFO [finetune.py:992] (0/2) Epoch 11, batch 11850, loss[loss=0.2347, simple_loss=0.3042, pruned_loss=0.08256, over 6819.00 frames. ], tot_loss[loss=0.2369, simple_loss=0.3141, pruned_loss=0.07988, over 1703415.17 frames. ], batch size: 99, lr: 3.86e-03, grad_scale: 8.0 2023-05-16 18:38:35,206 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.85 vs. limit=2.0 2023-05-16 18:38:38,008 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=234202.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 18:38:58,128 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=234231.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 18:39:03,469 INFO [finetune.py:992] (0/2) Epoch 11, batch 11900, loss[loss=0.2608, simple_loss=0.3296, pruned_loss=0.09602, over 6781.00 frames. ], tot_loss[loss=0.2355, simple_loss=0.3135, pruned_loss=0.0787, over 1706161.83 frames. ], batch size: 99, lr: 3.85e-03, grad_scale: 8.0 2023-05-16 18:39:12,668 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.5414, 4.5133, 4.4101, 3.9952, 4.0856, 4.5256, 4.2838, 4.0922], device='cuda:0'), covar=tensor([0.0787, 0.1019, 0.0696, 0.1456, 0.1841, 0.0792, 0.1498, 0.1017], device='cuda:0'), in_proj_covar=tensor([0.0584, 0.0525, 0.0487, 0.0598, 0.0396, 0.0665, 0.0719, 0.0537], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002], device='cuda:0') 2023-05-16 18:39:20,812 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=234263.0, num_to_drop=1, layers_to_drop={2} 2023-05-16 18:39:21,552 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.4203, 3.1880, 3.1139, 3.3792, 2.7208, 3.1794, 2.5760, 2.8290], device='cuda:0'), covar=tensor([0.1519, 0.0784, 0.0943, 0.0587, 0.1028, 0.0737, 0.1665, 0.0801], device='cuda:0'), in_proj_covar=tensor([0.0223, 0.0261, 0.0287, 0.0343, 0.0232, 0.0238, 0.0254, 0.0366], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 18:39:23,311 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.247e+02 3.250e+02 3.654e+02 4.294e+02 1.197e+03, threshold=7.308e+02, percent-clipped=1.0 2023-05-16 18:39:38,866 INFO [finetune.py:992] (0/2) Epoch 11, batch 11950, loss[loss=0.1935, simple_loss=0.2847, pruned_loss=0.05117, over 10378.00 frames. ], tot_loss[loss=0.2315, simple_loss=0.3106, pruned_loss=0.07621, over 1691846.54 frames. ], batch size: 68, lr: 3.85e-03, grad_scale: 8.0 2023-05-16 18:39:41,102 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=234292.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 18:39:43,187 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=234295.0, num_to_drop=1, layers_to_drop={1} 2023-05-16 18:40:13,539 INFO [finetune.py:992] (0/2) Epoch 11, batch 12000, loss[loss=0.1851, simple_loss=0.2794, pruned_loss=0.04536, over 10483.00 frames. ], tot_loss[loss=0.2254, simple_loss=0.3057, pruned_loss=0.07255, over 1689568.24 frames. ], batch size: 69, lr: 3.85e-03, grad_scale: 8.0 2023-05-16 18:40:13,540 INFO [finetune.py:1017] (0/2) Computing validation loss 2023-05-16 18:40:31,662 INFO [finetune.py:1026] (0/2) Epoch 11, validation: loss=0.2893, simple_loss=0.3643, pruned_loss=0.1072, over 1020973.00 frames. 2023-05-16 18:40:31,663 INFO [finetune.py:1027] (0/2) Maximum memory allocated so far is 12827MB 2023-05-16 18:40:34,463 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=234343.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 18:40:44,039 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 2023-05-16 18:40:48,510 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.2729, 5.1846, 5.2845, 5.2782, 4.9271, 5.0219, 4.8252, 5.1813], device='cuda:0'), covar=tensor([0.0754, 0.0536, 0.0774, 0.0583, 0.1693, 0.1078, 0.0578, 0.1016], device='cuda:0'), in_proj_covar=tensor([0.0495, 0.0638, 0.0559, 0.0564, 0.0763, 0.0675, 0.0505, 0.0440], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0003], device='cuda:0') 2023-05-16 18:40:50,645 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=234366.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 18:40:51,167 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.340e+02 2.882e+02 3.344e+02 3.924e+02 1.313e+03, threshold=6.687e+02, percent-clipped=3.0 2023-05-16 18:41:00,815 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=234381.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 18:41:07,056 INFO [finetune.py:992] (0/2) Epoch 11, batch 12050, loss[loss=0.2265, simple_loss=0.297, pruned_loss=0.07801, over 6894.00 frames. ], tot_loss[loss=0.2202, simple_loss=0.3013, pruned_loss=0.0695, over 1696152.62 frames. ], batch size: 101, lr: 3.85e-03, grad_scale: 8.0 2023-05-16 18:41:23,663 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=234414.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 18:41:39,565 INFO [finetune.py:992] (0/2) Epoch 11, batch 12100, loss[loss=0.2345, simple_loss=0.3054, pruned_loss=0.08179, over 6990.00 frames. ], tot_loss[loss=0.2185, simple_loss=0.3005, pruned_loss=0.06825, over 1705699.22 frames. ], batch size: 98, lr: 3.85e-03, grad_scale: 8.0 2023-05-16 18:41:58,105 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.758e+02 2.997e+02 3.482e+02 4.109e+02 7.850e+02, threshold=6.964e+02, percent-clipped=2.0 2023-05-16 18:42:10,334 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.8150, 3.6758, 3.7962, 3.6003, 3.7146, 3.5800, 3.7843, 3.5237], device='cuda:0'), covar=tensor([0.0387, 0.0356, 0.0317, 0.0244, 0.0354, 0.0315, 0.0348, 0.0986], device='cuda:0'), in_proj_covar=tensor([0.0236, 0.0240, 0.0260, 0.0237, 0.0238, 0.0237, 0.0213, 0.0197], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 18:42:12,187 INFO [finetune.py:992] (0/2) Epoch 11, batch 12150, loss[loss=0.2096, simple_loss=0.2982, pruned_loss=0.06046, over 7339.00 frames. ], tot_loss[loss=0.2188, simple_loss=0.3006, pruned_loss=0.0685, over 1695098.95 frames. ], batch size: 98, lr: 3.85e-03, grad_scale: 8.0 2023-05-16 18:42:34,486 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.36 vs. limit=2.0 2023-05-16 18:42:44,102 INFO [finetune.py:992] (0/2) Epoch 11, batch 12200, loss[loss=0.2476, simple_loss=0.3112, pruned_loss=0.09197, over 7040.00 frames. ], tot_loss[loss=0.2193, simple_loss=0.3008, pruned_loss=0.06886, over 1683633.89 frames. ], batch size: 98, lr: 3.85e-03, grad_scale: 8.0 2023-05-16 18:42:49,243 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.7697, 3.0972, 2.4244, 2.1575, 2.7865, 2.2918, 3.0412, 2.6192], device='cuda:0'), covar=tensor([0.0607, 0.0450, 0.0978, 0.1520, 0.0266, 0.1207, 0.0456, 0.0732], device='cuda:0'), in_proj_covar=tensor([0.0178, 0.0231, 0.0165, 0.0188, 0.0130, 0.0172, 0.0181, 0.0164], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 18:42:55,924 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=234558.0, num_to_drop=1, layers_to_drop={1} 2023-05-16 18:43:01,278 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.292e+02 3.263e+02 3.711e+02 4.394e+02 2.112e+03, threshold=7.422e+02, percent-clipped=3.0 2023-05-16 18:43:06,523 INFO [checkpoint.py:75] (0/2) Saving checkpoint to pruned_transducer_stateless7/exp_giga_finetune/epoch-11.pt 2023-05-16 18:43:28,050 INFO [finetune.py:992] (0/2) Epoch 12, batch 0, loss[loss=0.1892, simple_loss=0.2816, pruned_loss=0.04842, over 12349.00 frames. ], tot_loss[loss=0.1892, simple_loss=0.2816, pruned_loss=0.04842, over 12349.00 frames. ], batch size: 35, lr: 3.85e-03, grad_scale: 8.0 2023-05-16 18:43:28,051 INFO [finetune.py:1017] (0/2) Computing validation loss 2023-05-16 18:43:33,248 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.1447, 4.9498, 4.9493, 4.9832, 4.7816, 5.0062, 5.0824, 2.6372], device='cuda:0'), covar=tensor([0.0075, 0.0067, 0.0096, 0.0068, 0.0053, 0.0130, 0.0084, 0.1027], device='cuda:0'), in_proj_covar=tensor([0.0066, 0.0075, 0.0079, 0.0070, 0.0058, 0.0088, 0.0078, 0.0095], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-16 18:43:35,676 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.1253, 5.0552, 4.9799, 4.4152, 4.6821, 5.1246, 4.7761, 4.7505], device='cuda:0'), covar=tensor([0.0681, 0.0953, 0.0526, 0.1718, 0.0625, 0.0531, 0.1140, 0.0686], device='cuda:0'), in_proj_covar=tensor([0.0568, 0.0510, 0.0474, 0.0579, 0.0384, 0.0647, 0.0695, 0.0520], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002], device='cuda:0') 2023-05-16 18:43:45,303 INFO [finetune.py:1026] (0/2) Epoch 12, validation: loss=0.2833, simple_loss=0.36, pruned_loss=0.1033, over 1020973.00 frames. 2023-05-16 18:43:45,304 INFO [finetune.py:1027] (0/2) Maximum memory allocated so far is 12827MB 2023-05-16 18:43:54,312 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=234587.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 18:44:21,533 INFO [finetune.py:992] (0/2) Epoch 12, batch 50, loss[loss=0.1952, simple_loss=0.2822, pruned_loss=0.05411, over 10402.00 frames. ], tot_loss[loss=0.1838, simple_loss=0.2731, pruned_loss=0.04725, over 531665.43 frames. ], batch size: 68, lr: 3.85e-03, grad_scale: 8.0 2023-05-16 18:44:42,458 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.5423, 4.7644, 4.2058, 5.0202, 4.6241, 2.9535, 4.3019, 3.0641], device='cuda:0'), covar=tensor([0.0707, 0.0722, 0.1348, 0.0604, 0.1033, 0.1787, 0.1037, 0.3604], device='cuda:0'), in_proj_covar=tensor([0.0288, 0.0357, 0.0339, 0.0283, 0.0346, 0.0257, 0.0327, 0.0348], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-16 18:44:52,090 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.782e+02 2.975e+02 3.374e+02 4.225e+02 3.070e+03, threshold=6.748e+02, percent-clipped=4.0 2023-05-16 18:44:53,772 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=234669.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 18:44:57,223 INFO [finetune.py:992] (0/2) Epoch 12, batch 100, loss[loss=0.1864, simple_loss=0.2755, pruned_loss=0.04871, over 12043.00 frames. ], tot_loss[loss=0.179, simple_loss=0.2687, pruned_loss=0.04461, over 940316.88 frames. ], batch size: 37, lr: 3.85e-03, grad_scale: 8.0 2023-05-16 18:45:03,126 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=234681.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 18:45:17,839 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.46 vs. limit=2.0 2023-05-16 18:45:21,082 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.0311, 2.3584, 3.5625, 2.9443, 3.4081, 3.1045, 2.3765, 3.4444], device='cuda:0'), covar=tensor([0.0145, 0.0422, 0.0180, 0.0307, 0.0139, 0.0226, 0.0401, 0.0140], device='cuda:0'), in_proj_covar=tensor([0.0164, 0.0194, 0.0174, 0.0176, 0.0200, 0.0155, 0.0184, 0.0173], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 18:45:33,378 INFO [finetune.py:992] (0/2) Epoch 12, batch 150, loss[loss=0.1833, simple_loss=0.278, pruned_loss=0.04432, over 12370.00 frames. ], tot_loss[loss=0.1778, simple_loss=0.2679, pruned_loss=0.04382, over 1255915.63 frames. ], batch size: 38, lr: 3.85e-03, grad_scale: 8.0 2023-05-16 18:45:34,891 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=234726.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 18:45:37,660 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=234729.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 18:45:38,499 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=234730.0, num_to_drop=1, layers_to_drop={3} 2023-05-16 18:46:04,844 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.817e+02 2.565e+02 3.090e+02 3.582e+02 7.594e+02, threshold=6.180e+02, percent-clipped=1.0 2023-05-16 18:46:09,957 INFO [finetune.py:992] (0/2) Epoch 12, batch 200, loss[loss=0.1768, simple_loss=0.2669, pruned_loss=0.04339, over 12042.00 frames. ], tot_loss[loss=0.1752, simple_loss=0.2653, pruned_loss=0.04255, over 1506467.35 frames. ], batch size: 31, lr: 3.85e-03, grad_scale: 8.0 2023-05-16 18:46:16,002 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=234782.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 18:46:19,550 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=234787.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 18:46:20,239 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.3182, 2.6914, 3.6250, 4.3403, 3.7323, 4.3097, 3.6198, 3.0034], device='cuda:0'), covar=tensor([0.0042, 0.0351, 0.0130, 0.0034, 0.0116, 0.0065, 0.0146, 0.0357], device='cuda:0'), in_proj_covar=tensor([0.0082, 0.0121, 0.0101, 0.0075, 0.0099, 0.0113, 0.0093, 0.0134], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 18:46:30,381 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.5631, 3.6477, 3.2553, 3.2579, 2.9397, 2.7022, 3.7146, 2.2723], device='cuda:0'), covar=tensor([0.0364, 0.0144, 0.0209, 0.0195, 0.0411, 0.0367, 0.0122, 0.0498], device='cuda:0'), in_proj_covar=tensor([0.0181, 0.0151, 0.0153, 0.0175, 0.0193, 0.0188, 0.0159, 0.0192], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 18:46:37,081 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2023-05-16 18:46:46,154 INFO [finetune.py:992] (0/2) Epoch 12, batch 250, loss[loss=0.1597, simple_loss=0.2562, pruned_loss=0.03159, over 12340.00 frames. ], tot_loss[loss=0.1731, simple_loss=0.2629, pruned_loss=0.04172, over 1712549.78 frames. ], batch size: 31, lr: 3.85e-03, grad_scale: 4.0 2023-05-16 18:47:00,886 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=234843.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 18:47:11,370 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=234858.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 18:47:18,376 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.028e+02 2.787e+02 3.242e+02 3.922e+02 7.095e+02, threshold=6.485e+02, percent-clipped=2.0 2023-05-16 18:47:18,623 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.1482, 4.4640, 2.7378, 2.5739, 3.8460, 2.6558, 3.9693, 3.0419], device='cuda:0'), covar=tensor([0.0855, 0.0628, 0.1322, 0.1648, 0.0342, 0.1376, 0.0444, 0.0966], device='cuda:0'), in_proj_covar=tensor([0.0185, 0.0242, 0.0171, 0.0195, 0.0136, 0.0179, 0.0189, 0.0171], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 18:47:22,654 INFO [finetune.py:992] (0/2) Epoch 12, batch 300, loss[loss=0.1592, simple_loss=0.2424, pruned_loss=0.03801, over 12337.00 frames. ], tot_loss[loss=0.1722, simple_loss=0.2618, pruned_loss=0.04132, over 1862934.07 frames. ], batch size: 31, lr: 3.85e-03, grad_scale: 4.0 2023-05-16 18:47:28,538 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=234882.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 18:47:32,597 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=234887.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 18:47:46,265 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=234906.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 18:47:58,888 INFO [finetune.py:992] (0/2) Epoch 12, batch 350, loss[loss=0.1831, simple_loss=0.2841, pruned_loss=0.04101, over 12279.00 frames. ], tot_loss[loss=0.1717, simple_loss=0.2614, pruned_loss=0.04103, over 1970599.00 frames. ], batch size: 37, lr: 3.85e-03, grad_scale: 4.0 2023-05-16 18:48:06,657 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=234935.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 18:48:12,489 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=234943.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 18:48:22,951 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=234958.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 18:48:24,348 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=234960.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 18:48:29,994 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.948e+02 2.564e+02 3.005e+02 3.635e+02 9.262e+02, threshold=6.010e+02, percent-clipped=1.0 2023-05-16 18:48:34,366 INFO [finetune.py:992] (0/2) Epoch 12, batch 400, loss[loss=0.1981, simple_loss=0.2847, pruned_loss=0.05572, over 10348.00 frames. ], tot_loss[loss=0.1706, simple_loss=0.2599, pruned_loss=0.0406, over 2065262.27 frames. ], batch size: 68, lr: 3.85e-03, grad_scale: 8.0 2023-05-16 18:49:05,295 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.63 vs. limit=2.0 2023-05-16 18:49:07,290 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=235019.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 18:49:08,659 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=235021.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 18:49:10,654 INFO [finetune.py:992] (0/2) Epoch 12, batch 450, loss[loss=0.1855, simple_loss=0.2772, pruned_loss=0.04686, over 12148.00 frames. ], tot_loss[loss=0.1702, simple_loss=0.2594, pruned_loss=0.0405, over 2137879.98 frames. ], batch size: 36, lr: 3.85e-03, grad_scale: 8.0 2023-05-16 18:49:11,447 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=235025.0, num_to_drop=1, layers_to_drop={3} 2023-05-16 18:49:42,216 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.002e+02 2.777e+02 3.203e+02 3.888e+02 1.136e+03, threshold=6.405e+02, percent-clipped=5.0 2023-05-16 18:49:46,479 INFO [finetune.py:992] (0/2) Epoch 12, batch 500, loss[loss=0.1427, simple_loss=0.2264, pruned_loss=0.02953, over 12143.00 frames. ], tot_loss[loss=0.1714, simple_loss=0.2605, pruned_loss=0.04116, over 2193242.34 frames. ], batch size: 30, lr: 3.85e-03, grad_scale: 8.0 2023-05-16 18:49:46,607 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.1725, 4.8667, 5.1633, 4.5390, 4.8886, 4.4943, 5.1534, 4.8996], device='cuda:0'), covar=tensor([0.0360, 0.0417, 0.0375, 0.0321, 0.0423, 0.0430, 0.0335, 0.0367], device='cuda:0'), in_proj_covar=tensor([0.0246, 0.0249, 0.0272, 0.0248, 0.0247, 0.0249, 0.0224, 0.0206], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 18:49:52,067 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=235082.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 18:49:59,360 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.5181, 2.8800, 3.7407, 4.5495, 3.7907, 4.5138, 3.7401, 3.0219], device='cuda:0'), covar=tensor([0.0034, 0.0346, 0.0128, 0.0032, 0.0124, 0.0059, 0.0130, 0.0406], device='cuda:0'), in_proj_covar=tensor([0.0085, 0.0124, 0.0104, 0.0077, 0.0102, 0.0117, 0.0096, 0.0138], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 18:50:13,216 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.8837, 5.7016, 5.2866, 5.1259, 5.8538, 5.1442, 5.2514, 5.1362], device='cuda:0'), covar=tensor([0.1514, 0.1108, 0.1194, 0.2306, 0.1008, 0.2297, 0.2197, 0.1291], device='cuda:0'), in_proj_covar=tensor([0.0349, 0.0482, 0.0382, 0.0430, 0.0449, 0.0422, 0.0384, 0.0372], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 18:50:21,687 INFO [finetune.py:992] (0/2) Epoch 12, batch 550, loss[loss=0.1753, simple_loss=0.2679, pruned_loss=0.0414, over 12157.00 frames. ], tot_loss[loss=0.1709, simple_loss=0.26, pruned_loss=0.04093, over 2237139.24 frames. ], batch size: 39, lr: 3.84e-03, grad_scale: 8.0 2023-05-16 18:50:32,239 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=235138.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 18:50:53,594 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.722e+02 2.532e+02 3.050e+02 3.463e+02 5.793e+02, threshold=6.100e+02, percent-clipped=0.0 2023-05-16 18:50:55,279 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.2668, 2.8335, 4.7457, 2.3905, 2.5500, 3.6762, 2.8404, 3.7052], device='cuda:0'), covar=tensor([0.0562, 0.1515, 0.0393, 0.1360, 0.2125, 0.1436, 0.1572, 0.1175], device='cuda:0'), in_proj_covar=tensor([0.0232, 0.0232, 0.0240, 0.0180, 0.0235, 0.0289, 0.0220, 0.0256], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:0') 2023-05-16 18:50:57,772 INFO [finetune.py:992] (0/2) Epoch 12, batch 600, loss[loss=0.1499, simple_loss=0.2357, pruned_loss=0.0321, over 12126.00 frames. ], tot_loss[loss=0.1697, simple_loss=0.259, pruned_loss=0.04019, over 2272015.43 frames. ], batch size: 30, lr: 3.84e-03, grad_scale: 8.0 2023-05-16 18:51:07,900 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.9097, 4.8165, 4.7150, 4.7181, 4.4690, 5.0119, 4.8669, 5.1607], device='cuda:0'), covar=tensor([0.0197, 0.0156, 0.0206, 0.0369, 0.0751, 0.0280, 0.0166, 0.0203], device='cuda:0'), in_proj_covar=tensor([0.0184, 0.0186, 0.0182, 0.0238, 0.0234, 0.0208, 0.0171, 0.0216], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0005, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:0') 2023-05-16 18:51:26,216 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.52 vs. limit=2.0 2023-05-16 18:51:34,099 INFO [finetune.py:992] (0/2) Epoch 12, batch 650, loss[loss=0.1491, simple_loss=0.236, pruned_loss=0.03115, over 12341.00 frames. ], tot_loss[loss=0.1691, simple_loss=0.2583, pruned_loss=0.03993, over 2294112.47 frames. ], batch size: 31, lr: 3.84e-03, grad_scale: 8.0 2023-05-16 18:51:44,325 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=235238.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 18:51:56,523 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=235255.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 18:52:05,688 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.702e+02 2.751e+02 3.353e+02 3.897e+02 6.243e+02, threshold=6.706e+02, percent-clipped=1.0 2023-05-16 18:52:10,053 INFO [finetune.py:992] (0/2) Epoch 12, batch 700, loss[loss=0.1953, simple_loss=0.2899, pruned_loss=0.05034, over 12354.00 frames. ], tot_loss[loss=0.1687, simple_loss=0.2581, pruned_loss=0.03969, over 2320099.33 frames. ], batch size: 36, lr: 3.84e-03, grad_scale: 8.0 2023-05-16 18:52:38,854 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=235314.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 18:52:40,246 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=235316.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 18:52:40,394 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=235316.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 18:52:45,965 INFO [finetune.py:992] (0/2) Epoch 12, batch 750, loss[loss=0.1529, simple_loss=0.2375, pruned_loss=0.03415, over 12174.00 frames. ], tot_loss[loss=0.168, simple_loss=0.2571, pruned_loss=0.03945, over 2330783.39 frames. ], batch size: 29, lr: 3.84e-03, grad_scale: 8.0 2023-05-16 18:52:46,753 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=235325.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 18:53:06,674 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.1921, 2.6159, 3.7276, 3.1873, 3.5967, 3.2541, 2.5704, 3.5805], device='cuda:0'), covar=tensor([0.0135, 0.0346, 0.0155, 0.0248, 0.0136, 0.0171, 0.0367, 0.0157], device='cuda:0'), in_proj_covar=tensor([0.0172, 0.0200, 0.0182, 0.0183, 0.0210, 0.0161, 0.0191, 0.0181], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 18:53:12,192 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=235360.0, num_to_drop=1, layers_to_drop={1} 2023-05-16 18:53:17,625 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.734e+02 2.741e+02 3.316e+02 4.028e+02 1.402e+03, threshold=6.632e+02, percent-clipped=1.0 2023-05-16 18:53:21,208 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=235373.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 18:53:21,858 INFO [finetune.py:992] (0/2) Epoch 12, batch 800, loss[loss=0.1539, simple_loss=0.2342, pruned_loss=0.03675, over 12323.00 frames. ], tot_loss[loss=0.1687, simple_loss=0.2577, pruned_loss=0.03981, over 2342289.76 frames. ], batch size: 30, lr: 3.84e-03, grad_scale: 8.0 2023-05-16 18:53:25,686 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 2023-05-16 18:53:27,660 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=235382.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 18:53:55,503 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=235421.0, num_to_drop=1, layers_to_drop={3} 2023-05-16 18:53:57,303 INFO [finetune.py:992] (0/2) Epoch 12, batch 850, loss[loss=0.149, simple_loss=0.2301, pruned_loss=0.03394, over 12132.00 frames. ], tot_loss[loss=0.1689, simple_loss=0.2577, pruned_loss=0.04004, over 2350876.13 frames. ], batch size: 30, lr: 3.84e-03, grad_scale: 8.0 2023-05-16 18:54:01,643 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=235430.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 18:54:06,370 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.88 vs. limit=2.0 2023-05-16 18:54:07,396 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=235438.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 18:54:18,241 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2023-05-16 18:54:20,170 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.3406, 4.8798, 5.2853, 4.6371, 4.9660, 4.6440, 5.3084, 4.9130], device='cuda:0'), covar=tensor([0.0255, 0.0399, 0.0275, 0.0258, 0.0319, 0.0346, 0.0216, 0.0309], device='cuda:0'), in_proj_covar=tensor([0.0247, 0.0252, 0.0275, 0.0250, 0.0249, 0.0251, 0.0226, 0.0209], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 18:54:29,372 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.919e+02 2.777e+02 3.296e+02 3.961e+02 7.490e+02, threshold=6.592e+02, percent-clipped=2.0 2023-05-16 18:54:33,577 INFO [finetune.py:992] (0/2) Epoch 12, batch 900, loss[loss=0.1729, simple_loss=0.2545, pruned_loss=0.04562, over 12261.00 frames. ], tot_loss[loss=0.169, simple_loss=0.2579, pruned_loss=0.03999, over 2364882.43 frames. ], batch size: 32, lr: 3.84e-03, grad_scale: 8.0 2023-05-16 18:54:33,833 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.0787, 2.5555, 3.6884, 3.1558, 3.4632, 3.2087, 2.6253, 3.5380], device='cuda:0'), covar=tensor([0.0157, 0.0361, 0.0144, 0.0269, 0.0160, 0.0211, 0.0339, 0.0138], device='cuda:0'), in_proj_covar=tensor([0.0171, 0.0199, 0.0181, 0.0182, 0.0208, 0.0160, 0.0190, 0.0180], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 18:54:42,374 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=235486.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 18:55:10,327 INFO [finetune.py:992] (0/2) Epoch 12, batch 950, loss[loss=0.152, simple_loss=0.2483, pruned_loss=0.02786, over 12344.00 frames. ], tot_loss[loss=0.1684, simple_loss=0.2575, pruned_loss=0.03971, over 2368555.08 frames. ], batch size: 36, lr: 3.84e-03, grad_scale: 8.0 2023-05-16 18:55:20,547 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=235538.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 18:55:23,487 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.1718, 3.7248, 3.9438, 4.2608, 2.8108, 3.7173, 2.4016, 3.8705], device='cuda:0'), covar=tensor([0.1803, 0.0950, 0.1021, 0.0838, 0.1360, 0.0768, 0.2068, 0.1168], device='cuda:0'), in_proj_covar=tensor([0.0229, 0.0265, 0.0294, 0.0350, 0.0236, 0.0242, 0.0258, 0.0372], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 18:55:41,852 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.922e+02 2.824e+02 3.256e+02 3.898e+02 8.712e+02, threshold=6.511e+02, percent-clipped=2.0 2023-05-16 18:55:46,189 INFO [finetune.py:992] (0/2) Epoch 12, batch 1000, loss[loss=0.1575, simple_loss=0.2555, pruned_loss=0.02975, over 12025.00 frames. ], tot_loss[loss=0.1682, simple_loss=0.2573, pruned_loss=0.03956, over 2373572.29 frames. ], batch size: 31, lr: 3.84e-03, grad_scale: 8.0 2023-05-16 18:55:54,783 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=235586.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 18:56:13,638 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=235611.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 18:56:15,551 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=235614.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 18:56:17,012 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=235616.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 18:56:22,600 INFO [finetune.py:992] (0/2) Epoch 12, batch 1050, loss[loss=0.1461, simple_loss=0.2266, pruned_loss=0.03283, over 12003.00 frames. ], tot_loss[loss=0.1678, simple_loss=0.2567, pruned_loss=0.03944, over 2374636.23 frames. ], batch size: 28, lr: 3.84e-03, grad_scale: 8.0 2023-05-16 18:56:27,103 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.3169, 3.8205, 3.9766, 4.3609, 2.9302, 3.9222, 2.6804, 4.0973], device='cuda:0'), covar=tensor([0.1686, 0.0896, 0.1024, 0.0814, 0.1288, 0.0687, 0.1887, 0.1208], device='cuda:0'), in_proj_covar=tensor([0.0229, 0.0267, 0.0295, 0.0352, 0.0238, 0.0243, 0.0259, 0.0375], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 18:56:32,993 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.5712, 2.6669, 4.3189, 4.5011, 2.8171, 2.6192, 2.7690, 2.0778], device='cuda:0'), covar=tensor([0.1738, 0.3207, 0.0544, 0.0468, 0.1383, 0.2451, 0.2984, 0.4147], device='cuda:0'), in_proj_covar=tensor([0.0291, 0.0381, 0.0272, 0.0294, 0.0268, 0.0301, 0.0378, 0.0373], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-16 18:56:50,311 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=235662.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 18:56:51,705 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=235664.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 18:56:54,507 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.777e+02 2.852e+02 3.251e+02 3.804e+02 7.432e+02, threshold=6.502e+02, percent-clipped=1.0 2023-05-16 18:56:58,806 INFO [finetune.py:992] (0/2) Epoch 12, batch 1100, loss[loss=0.1664, simple_loss=0.2588, pruned_loss=0.03699, over 12163.00 frames. ], tot_loss[loss=0.1674, simple_loss=0.2565, pruned_loss=0.03919, over 2379069.08 frames. ], batch size: 36, lr: 3.84e-03, grad_scale: 8.0 2023-05-16 18:56:59,334 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.91 vs. limit=2.0 2023-05-16 18:57:02,650 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=235679.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 18:57:20,193 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.2904, 4.9325, 5.2752, 5.2360, 4.3850, 4.5849, 4.6680, 5.0241], device='cuda:0'), covar=tensor([0.0845, 0.1215, 0.0880, 0.0728, 0.3709, 0.2135, 0.0720, 0.1645], device='cuda:0'), in_proj_covar=tensor([0.0527, 0.0684, 0.0594, 0.0597, 0.0816, 0.0721, 0.0530, 0.0469], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0005, 0.0004, 0.0003, 0.0003], device='cuda:0') 2023-05-16 18:57:28,576 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=235716.0, num_to_drop=1, layers_to_drop={2} 2023-05-16 18:57:33,980 INFO [finetune.py:992] (0/2) Epoch 12, batch 1150, loss[loss=0.2029, simple_loss=0.2835, pruned_loss=0.06116, over 12366.00 frames. ], tot_loss[loss=0.1679, simple_loss=0.2572, pruned_loss=0.03935, over 2383277.05 frames. ], batch size: 35, lr: 3.84e-03, grad_scale: 8.0 2023-05-16 18:57:37,748 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.1089, 4.6706, 5.0395, 4.4018, 4.7195, 4.4303, 5.0828, 4.8158], device='cuda:0'), covar=tensor([0.0260, 0.0393, 0.0301, 0.0282, 0.0392, 0.0371, 0.0231, 0.0350], device='cuda:0'), in_proj_covar=tensor([0.0251, 0.0256, 0.0279, 0.0254, 0.0253, 0.0257, 0.0230, 0.0212], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 18:57:46,323 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=235740.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 18:58:05,944 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.757e+02 2.666e+02 3.094e+02 3.777e+02 6.961e+02, threshold=6.188e+02, percent-clipped=4.0 2023-05-16 18:58:10,223 INFO [finetune.py:992] (0/2) Epoch 12, batch 1200, loss[loss=0.1414, simple_loss=0.223, pruned_loss=0.02984, over 11993.00 frames. ], tot_loss[loss=0.1678, simple_loss=0.2569, pruned_loss=0.03933, over 2384345.81 frames. ], batch size: 28, lr: 3.84e-03, grad_scale: 8.0 2023-05-16 18:58:22,481 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=235791.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 18:58:29,694 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.0225, 4.3938, 3.9248, 4.7601, 4.4387, 2.7518, 4.1577, 2.9025], device='cuda:0'), covar=tensor([0.0913, 0.0917, 0.1580, 0.0540, 0.1159, 0.1787, 0.1018, 0.3538], device='cuda:0'), in_proj_covar=tensor([0.0300, 0.0374, 0.0353, 0.0300, 0.0363, 0.0267, 0.0341, 0.0364], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 18:58:32,634 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.36 vs. limit=2.0 2023-05-16 18:58:39,095 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-05-16 18:58:46,424 INFO [finetune.py:992] (0/2) Epoch 12, batch 1250, loss[loss=0.1936, simple_loss=0.2836, pruned_loss=0.05186, over 12290.00 frames. ], tot_loss[loss=0.1687, simple_loss=0.2578, pruned_loss=0.03977, over 2379224.83 frames. ], batch size: 37, lr: 3.84e-03, grad_scale: 8.0 2023-05-16 18:58:53,734 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.8021, 2.7603, 3.3133, 4.6858, 2.5316, 4.5806, 4.6780, 4.8617], device='cuda:0'), covar=tensor([0.0076, 0.1123, 0.0462, 0.0106, 0.1276, 0.0205, 0.0125, 0.0061], device='cuda:0'), in_proj_covar=tensor([0.0120, 0.0201, 0.0182, 0.0116, 0.0188, 0.0176, 0.0175, 0.0119], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 18:59:06,678 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=235852.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 18:59:17,929 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.771e+02 2.652e+02 3.056e+02 3.662e+02 6.029e+02, threshold=6.111e+02, percent-clipped=0.0 2023-05-16 18:59:22,117 INFO [finetune.py:992] (0/2) Epoch 12, batch 1300, loss[loss=0.1709, simple_loss=0.2663, pruned_loss=0.03772, over 11109.00 frames. ], tot_loss[loss=0.1687, simple_loss=0.258, pruned_loss=0.03972, over 2384069.80 frames. ], batch size: 55, lr: 3.84e-03, grad_scale: 8.0 2023-05-16 18:59:49,262 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=235911.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 18:59:58,263 INFO [finetune.py:992] (0/2) Epoch 12, batch 1350, loss[loss=0.2541, simple_loss=0.318, pruned_loss=0.09506, over 8313.00 frames. ], tot_loss[loss=0.1692, simple_loss=0.2583, pruned_loss=0.03999, over 2375147.54 frames. ], batch size: 97, lr: 3.84e-03, grad_scale: 8.0 2023-05-16 19:00:24,239 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=235959.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 19:00:30,495 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.591e+02 2.594e+02 3.047e+02 3.701e+02 8.074e+02, threshold=6.093e+02, percent-clipped=3.0 2023-05-16 19:00:34,775 INFO [finetune.py:992] (0/2) Epoch 12, batch 1400, loss[loss=0.1644, simple_loss=0.256, pruned_loss=0.03637, over 12257.00 frames. ], tot_loss[loss=0.1686, simple_loss=0.2577, pruned_loss=0.03976, over 2377384.99 frames. ], batch size: 32, lr: 3.84e-03, grad_scale: 8.0 2023-05-16 19:00:39,610 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=4.76 vs. limit=5.0 2023-05-16 19:00:53,609 INFO [checkpoint.py:75] (0/2) Saving checkpoint to pruned_transducer_stateless7/exp_giga_finetune/checkpoint-136000.pt 2023-05-16 19:01:03,271 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=4.41 vs. limit=5.0 2023-05-16 19:01:08,003 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=236016.0, num_to_drop=1, layers_to_drop={0} 2023-05-16 19:01:09,384 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=236018.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 19:01:13,401 INFO [finetune.py:992] (0/2) Epoch 12, batch 1450, loss[loss=0.2018, simple_loss=0.287, pruned_loss=0.05825, over 12363.00 frames. ], tot_loss[loss=0.1679, simple_loss=0.2571, pruned_loss=0.03931, over 2383238.27 frames. ], batch size: 38, lr: 3.84e-03, grad_scale: 8.0 2023-05-16 19:01:19,198 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.2073, 6.0657, 5.5730, 5.6847, 6.1334, 5.3461, 5.5333, 5.6240], device='cuda:0'), covar=tensor([0.1501, 0.0911, 0.0955, 0.1787, 0.0852, 0.2174, 0.1849, 0.1230], device='cuda:0'), in_proj_covar=tensor([0.0354, 0.0488, 0.0385, 0.0437, 0.0452, 0.0424, 0.0391, 0.0378], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 19:01:20,390 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.73 vs. limit=2.0 2023-05-16 19:01:22,030 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=236035.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 19:01:43,071 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=236064.0, num_to_drop=1, layers_to_drop={0} 2023-05-16 19:01:45,846 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.817e+02 2.819e+02 3.335e+02 3.913e+02 5.807e+02, threshold=6.670e+02, percent-clipped=0.0 2023-05-16 19:01:48,996 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.91 vs. limit=2.0 2023-05-16 19:01:50,045 INFO [finetune.py:992] (0/2) Epoch 12, batch 1500, loss[loss=0.1599, simple_loss=0.2526, pruned_loss=0.03363, over 12358.00 frames. ], tot_loss[loss=0.1678, simple_loss=0.257, pruned_loss=0.03925, over 2387649.33 frames. ], batch size: 35, lr: 3.84e-03, grad_scale: 8.0 2023-05-16 19:01:51,131 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.56 vs. limit=2.0 2023-05-16 19:01:53,753 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=236079.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 19:01:58,175 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.33 vs. limit=2.0 2023-05-16 19:02:26,072 INFO [finetune.py:992] (0/2) Epoch 12, batch 1550, loss[loss=0.1506, simple_loss=0.2268, pruned_loss=0.03716, over 11989.00 frames. ], tot_loss[loss=0.1685, simple_loss=0.2578, pruned_loss=0.03963, over 2377365.95 frames. ], batch size: 28, lr: 3.84e-03, grad_scale: 8.0 2023-05-16 19:02:42,368 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=236147.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 19:02:57,378 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.882e+02 2.682e+02 3.126e+02 4.093e+02 1.267e+03, threshold=6.252e+02, percent-clipped=5.0 2023-05-16 19:03:01,721 INFO [finetune.py:992] (0/2) Epoch 12, batch 1600, loss[loss=0.1664, simple_loss=0.2604, pruned_loss=0.03624, over 12312.00 frames. ], tot_loss[loss=0.1679, simple_loss=0.2569, pruned_loss=0.0394, over 2381875.50 frames. ], batch size: 34, lr: 3.84e-03, grad_scale: 8.0 2023-05-16 19:03:13,292 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 2023-05-16 19:03:38,604 INFO [finetune.py:992] (0/2) Epoch 12, batch 1650, loss[loss=0.2038, simple_loss=0.2907, pruned_loss=0.05846, over 12113.00 frames. ], tot_loss[loss=0.1684, simple_loss=0.2572, pruned_loss=0.03979, over 2381575.04 frames. ], batch size: 38, lr: 3.83e-03, grad_scale: 4.0 2023-05-16 19:04:08,284 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.9409, 3.7642, 3.9098, 3.6156, 3.8235, 3.6773, 3.9270, 3.5399], device='cuda:0'), covar=tensor([0.0324, 0.0385, 0.0357, 0.0277, 0.0320, 0.0294, 0.0293, 0.1137], device='cuda:0'), in_proj_covar=tensor([0.0254, 0.0260, 0.0284, 0.0257, 0.0256, 0.0259, 0.0232, 0.0213], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 19:04:10,981 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.929e+02 2.699e+02 3.140e+02 3.673e+02 8.053e+02, threshold=6.280e+02, percent-clipped=3.0 2023-05-16 19:04:14,570 INFO [finetune.py:992] (0/2) Epoch 12, batch 1700, loss[loss=0.1601, simple_loss=0.2552, pruned_loss=0.0325, over 12300.00 frames. ], tot_loss[loss=0.1671, simple_loss=0.256, pruned_loss=0.03913, over 2377602.28 frames. ], batch size: 34, lr: 3.83e-03, grad_scale: 4.0 2023-05-16 19:04:50,312 INFO [finetune.py:992] (0/2) Epoch 12, batch 1750, loss[loss=0.1953, simple_loss=0.2912, pruned_loss=0.04973, over 12107.00 frames. ], tot_loss[loss=0.1664, simple_loss=0.2553, pruned_loss=0.03879, over 2384251.03 frames. ], batch size: 39, lr: 3.83e-03, grad_scale: 4.0 2023-05-16 19:04:58,225 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=236335.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 19:05:03,377 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.38 vs. limit=2.0 2023-05-16 19:05:17,903 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=2.46 vs. limit=5.0 2023-05-16 19:05:22,115 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.88 vs. limit=2.0 2023-05-16 19:05:22,393 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.127e+02 2.700e+02 3.222e+02 3.974e+02 5.690e+02, threshold=6.445e+02, percent-clipped=0.0 2023-05-16 19:05:24,307 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.94 vs. limit=2.0 2023-05-16 19:05:26,034 INFO [finetune.py:992] (0/2) Epoch 12, batch 1800, loss[loss=0.177, simple_loss=0.2663, pruned_loss=0.04382, over 12002.00 frames. ], tot_loss[loss=0.1681, simple_loss=0.257, pruned_loss=0.03962, over 2383976.17 frames. ], batch size: 40, lr: 3.83e-03, grad_scale: 4.0 2023-05-16 19:05:26,107 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=236374.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 19:05:33,121 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=236383.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 19:06:02,432 INFO [finetune.py:992] (0/2) Epoch 12, batch 1850, loss[loss=0.1637, simple_loss=0.2504, pruned_loss=0.03857, over 12333.00 frames. ], tot_loss[loss=0.1681, simple_loss=0.257, pruned_loss=0.03962, over 2386913.41 frames. ], batch size: 31, lr: 3.83e-03, grad_scale: 4.0 2023-05-16 19:06:06,895 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.0983, 2.5682, 3.6939, 3.1010, 3.4698, 3.2805, 2.5406, 3.5266], device='cuda:0'), covar=tensor([0.0147, 0.0370, 0.0144, 0.0256, 0.0154, 0.0159, 0.0376, 0.0144], device='cuda:0'), in_proj_covar=tensor([0.0176, 0.0203, 0.0186, 0.0186, 0.0214, 0.0163, 0.0194, 0.0185], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 19:06:19,018 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=236447.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 19:06:34,314 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.861e+02 2.812e+02 3.208e+02 3.818e+02 6.942e+02, threshold=6.416e+02, percent-clipped=0.0 2023-05-16 19:06:37,921 INFO [finetune.py:992] (0/2) Epoch 12, batch 1900, loss[loss=0.1632, simple_loss=0.2564, pruned_loss=0.03498, over 12101.00 frames. ], tot_loss[loss=0.1684, simple_loss=0.2574, pruned_loss=0.0397, over 2389140.57 frames. ], batch size: 39, lr: 3.83e-03, grad_scale: 4.0 2023-05-16 19:06:44,533 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.0050, 4.6352, 4.7158, 4.8939, 4.7534, 4.8240, 4.8311, 2.4000], device='cuda:0'), covar=tensor([0.0106, 0.0070, 0.0092, 0.0066, 0.0045, 0.0106, 0.0072, 0.0901], device='cuda:0'), in_proj_covar=tensor([0.0069, 0.0078, 0.0081, 0.0073, 0.0059, 0.0092, 0.0081, 0.0097], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-16 19:06:53,175 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-05-16 19:06:53,658 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=236495.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 19:07:15,024 INFO [finetune.py:992] (0/2) Epoch 12, batch 1950, loss[loss=0.2111, simple_loss=0.2968, pruned_loss=0.06274, over 7998.00 frames. ], tot_loss[loss=0.1683, simple_loss=0.2572, pruned_loss=0.03974, over 2379515.37 frames. ], batch size: 98, lr: 3.83e-03, grad_scale: 4.0 2023-05-16 19:07:18,635 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.2551, 2.6663, 3.8008, 3.1602, 3.6405, 3.3141, 2.6657, 3.6680], device='cuda:0'), covar=tensor([0.0143, 0.0350, 0.0141, 0.0266, 0.0135, 0.0183, 0.0365, 0.0130], device='cuda:0'), in_proj_covar=tensor([0.0177, 0.0204, 0.0187, 0.0188, 0.0216, 0.0165, 0.0195, 0.0187], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 19:07:30,472 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.37 vs. limit=2.0 2023-05-16 19:07:46,999 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.010e+02 2.574e+02 3.108e+02 3.850e+02 7.109e+02, threshold=6.217e+02, percent-clipped=4.0 2023-05-16 19:07:50,484 INFO [finetune.py:992] (0/2) Epoch 12, batch 2000, loss[loss=0.177, simple_loss=0.264, pruned_loss=0.04496, over 11592.00 frames. ], tot_loss[loss=0.1676, simple_loss=0.2564, pruned_loss=0.0394, over 2379032.09 frames. ], batch size: 48, lr: 3.83e-03, grad_scale: 8.0 2023-05-16 19:07:58,724 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.62 vs. limit=2.0 2023-05-16 19:08:20,035 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.92 vs. limit=2.0 2023-05-16 19:08:26,758 INFO [finetune.py:992] (0/2) Epoch 12, batch 2050, loss[loss=0.213, simple_loss=0.289, pruned_loss=0.06849, over 8201.00 frames. ], tot_loss[loss=0.1684, simple_loss=0.2571, pruned_loss=0.0398, over 2375611.54 frames. ], batch size: 98, lr: 3.83e-03, grad_scale: 8.0 2023-05-16 19:08:39,687 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.2223, 4.5631, 2.8528, 2.7636, 3.8391, 2.4892, 3.9694, 3.2383], device='cuda:0'), covar=tensor([0.0827, 0.0566, 0.1116, 0.1359, 0.0320, 0.1361, 0.0473, 0.0837], device='cuda:0'), in_proj_covar=tensor([0.0190, 0.0254, 0.0176, 0.0200, 0.0141, 0.0183, 0.0197, 0.0176], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 19:08:59,857 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.871e+02 2.605e+02 3.083e+02 3.688e+02 8.642e+02, threshold=6.167e+02, percent-clipped=2.0 2023-05-16 19:09:03,387 INFO [finetune.py:992] (0/2) Epoch 12, batch 2100, loss[loss=0.1981, simple_loss=0.2792, pruned_loss=0.05851, over 12349.00 frames. ], tot_loss[loss=0.1683, simple_loss=0.2572, pruned_loss=0.03973, over 2379921.57 frames. ], batch size: 35, lr: 3.83e-03, grad_scale: 8.0 2023-05-16 19:09:03,546 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=236674.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 19:09:12,009 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.3259, 6.1349, 5.5796, 5.6581, 6.2101, 5.5349, 5.6854, 5.6403], device='cuda:0'), covar=tensor([0.1392, 0.0941, 0.1318, 0.1949, 0.0902, 0.2221, 0.1823, 0.1120], device='cuda:0'), in_proj_covar=tensor([0.0356, 0.0490, 0.0388, 0.0437, 0.0455, 0.0430, 0.0390, 0.0379], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 19:09:26,007 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=3.52 vs. limit=5.0 2023-05-16 19:09:37,534 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=236722.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 19:09:39,002 INFO [finetune.py:992] (0/2) Epoch 12, batch 2150, loss[loss=0.142, simple_loss=0.2304, pruned_loss=0.02683, over 12297.00 frames. ], tot_loss[loss=0.1676, simple_loss=0.2563, pruned_loss=0.03943, over 2379398.09 frames. ], batch size: 28, lr: 3.83e-03, grad_scale: 8.0 2023-05-16 19:09:45,640 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.5068, 5.0823, 5.4118, 4.7513, 5.0568, 4.8835, 5.4735, 5.1050], device='cuda:0'), covar=tensor([0.0267, 0.0348, 0.0301, 0.0292, 0.0388, 0.0337, 0.0246, 0.0257], device='cuda:0'), in_proj_covar=tensor([0.0255, 0.0261, 0.0285, 0.0259, 0.0259, 0.0261, 0.0233, 0.0213], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 19:09:47,996 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.64 vs. limit=2.0 2023-05-16 19:10:11,904 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.100e+02 2.781e+02 3.230e+02 3.738e+02 9.151e+02, threshold=6.461e+02, percent-clipped=3.0 2023-05-16 19:10:15,465 INFO [finetune.py:992] (0/2) Epoch 12, batch 2200, loss[loss=0.2305, simple_loss=0.3101, pruned_loss=0.07541, over 8235.00 frames. ], tot_loss[loss=0.1674, simple_loss=0.2559, pruned_loss=0.03948, over 2369378.04 frames. ], batch size: 98, lr: 3.83e-03, grad_scale: 4.0 2023-05-16 19:10:52,385 INFO [finetune.py:992] (0/2) Epoch 12, batch 2250, loss[loss=0.1756, simple_loss=0.2698, pruned_loss=0.0407, over 11194.00 frames. ], tot_loss[loss=0.1676, simple_loss=0.2565, pruned_loss=0.03933, over 2369329.59 frames. ], batch size: 55, lr: 3.83e-03, grad_scale: 4.0 2023-05-16 19:11:21,921 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.9966, 4.6471, 4.8179, 4.8387, 4.7238, 4.8323, 4.7552, 2.8357], device='cuda:0'), covar=tensor([0.0107, 0.0083, 0.0087, 0.0071, 0.0046, 0.0119, 0.0119, 0.0735], device='cuda:0'), in_proj_covar=tensor([0.0068, 0.0077, 0.0080, 0.0073, 0.0059, 0.0091, 0.0081, 0.0096], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-16 19:11:25,239 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.674e+02 2.624e+02 2.991e+02 3.746e+02 7.102e+02, threshold=5.982e+02, percent-clipped=2.0 2023-05-16 19:11:28,185 INFO [finetune.py:992] (0/2) Epoch 12, batch 2300, loss[loss=0.1612, simple_loss=0.2491, pruned_loss=0.03669, over 11329.00 frames. ], tot_loss[loss=0.1675, simple_loss=0.2563, pruned_loss=0.03935, over 2362173.63 frames. ], batch size: 55, lr: 3.83e-03, grad_scale: 4.0 2023-05-16 19:11:41,333 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=236892.0, num_to_drop=1, layers_to_drop={0} 2023-05-16 19:12:04,928 INFO [finetune.py:992] (0/2) Epoch 12, batch 2350, loss[loss=0.1644, simple_loss=0.2579, pruned_loss=0.03546, over 12207.00 frames. ], tot_loss[loss=0.1668, simple_loss=0.2556, pruned_loss=0.03902, over 2363136.50 frames. ], batch size: 35, lr: 3.83e-03, grad_scale: 4.0 2023-05-16 19:12:17,107 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.1374, 2.3569, 3.7574, 3.0656, 3.5266, 3.1852, 2.4923, 3.5639], device='cuda:0'), covar=tensor([0.0139, 0.0403, 0.0117, 0.0236, 0.0121, 0.0209, 0.0396, 0.0116], device='cuda:0'), in_proj_covar=tensor([0.0176, 0.0203, 0.0187, 0.0187, 0.0215, 0.0164, 0.0195, 0.0186], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 19:12:26,107 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=236953.0, num_to_drop=1, layers_to_drop={0} 2023-05-16 19:12:30,454 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.3308, 4.7879, 2.9847, 2.7187, 4.0537, 2.7824, 4.0343, 3.3914], device='cuda:0'), covar=tensor([0.0693, 0.0408, 0.1080, 0.1443, 0.0266, 0.1162, 0.0470, 0.0710], device='cuda:0'), in_proj_covar=tensor([0.0192, 0.0257, 0.0178, 0.0202, 0.0143, 0.0186, 0.0200, 0.0178], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 19:12:38,064 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.716e+02 2.618e+02 2.916e+02 3.600e+02 7.295e+02, threshold=5.831e+02, percent-clipped=3.0 2023-05-16 19:12:40,929 INFO [finetune.py:992] (0/2) Epoch 12, batch 2400, loss[loss=0.1765, simple_loss=0.2767, pruned_loss=0.03818, over 10434.00 frames. ], tot_loss[loss=0.1677, simple_loss=0.2568, pruned_loss=0.03931, over 2365960.45 frames. ], batch size: 68, lr: 3.83e-03, grad_scale: 8.0 2023-05-16 19:13:14,749 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.7914, 4.6751, 4.6330, 4.7050, 4.3497, 4.8593, 4.7570, 4.9772], device='cuda:0'), covar=tensor([0.0229, 0.0168, 0.0222, 0.0333, 0.0801, 0.0349, 0.0183, 0.0202], device='cuda:0'), in_proj_covar=tensor([0.0194, 0.0197, 0.0190, 0.0251, 0.0244, 0.0218, 0.0178, 0.0227], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0005, 0.0004, 0.0004, 0.0003, 0.0005], device='cuda:0') 2023-05-16 19:13:16,056 INFO [finetune.py:992] (0/2) Epoch 12, batch 2450, loss[loss=0.146, simple_loss=0.2337, pruned_loss=0.02909, over 12184.00 frames. ], tot_loss[loss=0.1682, simple_loss=0.2574, pruned_loss=0.03953, over 2367137.32 frames. ], batch size: 29, lr: 3.83e-03, grad_scale: 4.0 2023-05-16 19:13:49,920 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.729e+02 2.894e+02 3.316e+02 3.902e+02 6.635e+02, threshold=6.633e+02, percent-clipped=2.0 2023-05-16 19:13:52,169 INFO [finetune.py:992] (0/2) Epoch 12, batch 2500, loss[loss=0.1578, simple_loss=0.2499, pruned_loss=0.03282, over 12192.00 frames. ], tot_loss[loss=0.1677, simple_loss=0.2567, pruned_loss=0.03938, over 2375822.79 frames. ], batch size: 35, lr: 3.83e-03, grad_scale: 4.0 2023-05-16 19:14:08,183 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.62 vs. limit=2.0 2023-05-16 19:14:18,702 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.6431, 3.5430, 3.1959, 3.2159, 2.9254, 2.8186, 3.6401, 2.3042], device='cuda:0'), covar=tensor([0.0371, 0.0166, 0.0235, 0.0199, 0.0390, 0.0359, 0.0134, 0.0483], device='cuda:0'), in_proj_covar=tensor([0.0194, 0.0161, 0.0163, 0.0187, 0.0204, 0.0200, 0.0171, 0.0204], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 19:14:28,245 INFO [finetune.py:992] (0/2) Epoch 12, batch 2550, loss[loss=0.1772, simple_loss=0.2511, pruned_loss=0.05159, over 11801.00 frames. ], tot_loss[loss=0.1673, simple_loss=0.2568, pruned_loss=0.03892, over 2381606.93 frames. ], batch size: 26, lr: 3.83e-03, grad_scale: 4.0 2023-05-16 19:15:01,498 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.786e+02 2.724e+02 3.089e+02 3.882e+02 6.578e+02, threshold=6.177e+02, percent-clipped=0.0 2023-05-16 19:15:03,641 INFO [finetune.py:992] (0/2) Epoch 12, batch 2600, loss[loss=0.1719, simple_loss=0.2698, pruned_loss=0.037, over 12271.00 frames. ], tot_loss[loss=0.1679, simple_loss=0.2571, pruned_loss=0.03933, over 2384641.15 frames. ], batch size: 37, lr: 3.83e-03, grad_scale: 4.0 2023-05-16 19:15:39,929 INFO [finetune.py:992] (0/2) Epoch 12, batch 2650, loss[loss=0.146, simple_loss=0.228, pruned_loss=0.03195, over 11984.00 frames. ], tot_loss[loss=0.1682, simple_loss=0.2575, pruned_loss=0.03941, over 2389486.92 frames. ], batch size: 28, lr: 3.83e-03, grad_scale: 4.0 2023-05-16 19:15:57,482 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=237248.0, num_to_drop=1, layers_to_drop={3} 2023-05-16 19:16:13,839 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.585e+02 2.563e+02 3.095e+02 3.616e+02 7.529e+02, threshold=6.190e+02, percent-clipped=2.0 2023-05-16 19:16:15,964 INFO [finetune.py:992] (0/2) Epoch 12, batch 2700, loss[loss=0.163, simple_loss=0.2479, pruned_loss=0.03903, over 12081.00 frames. ], tot_loss[loss=0.1686, simple_loss=0.2579, pruned_loss=0.03971, over 2385320.44 frames. ], batch size: 32, lr: 3.83e-03, grad_scale: 4.0 2023-05-16 19:16:43,327 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.83 vs. limit=2.0 2023-05-16 19:16:51,227 INFO [finetune.py:992] (0/2) Epoch 12, batch 2750, loss[loss=0.1574, simple_loss=0.2528, pruned_loss=0.03103, over 12115.00 frames. ], tot_loss[loss=0.1691, simple_loss=0.2584, pruned_loss=0.03993, over 2382845.62 frames. ], batch size: 33, lr: 3.82e-03, grad_scale: 4.0 2023-05-16 19:17:12,234 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=4.48 vs. limit=5.0 2023-05-16 19:17:25,095 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.967e+02 2.699e+02 3.189e+02 3.843e+02 1.550e+03, threshold=6.378e+02, percent-clipped=4.0 2023-05-16 19:17:27,226 INFO [finetune.py:992] (0/2) Epoch 12, batch 2800, loss[loss=0.1489, simple_loss=0.2317, pruned_loss=0.03302, over 12185.00 frames. ], tot_loss[loss=0.169, simple_loss=0.2579, pruned_loss=0.04011, over 2371778.90 frames. ], batch size: 31, lr: 3.82e-03, grad_scale: 8.0 2023-05-16 19:17:27,840 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 2023-05-16 19:18:03,963 INFO [finetune.py:992] (0/2) Epoch 12, batch 2850, loss[loss=0.1447, simple_loss=0.233, pruned_loss=0.02815, over 12408.00 frames. ], tot_loss[loss=0.1687, simple_loss=0.2576, pruned_loss=0.03987, over 2377198.71 frames. ], batch size: 32, lr: 3.82e-03, grad_scale: 8.0 2023-05-16 19:18:37,459 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.825e+02 2.552e+02 3.036e+02 3.726e+02 9.226e+02, threshold=6.072e+02, percent-clipped=2.0 2023-05-16 19:18:39,516 INFO [finetune.py:992] (0/2) Epoch 12, batch 2900, loss[loss=0.1907, simple_loss=0.2765, pruned_loss=0.05246, over 12380.00 frames. ], tot_loss[loss=0.1676, simple_loss=0.2563, pruned_loss=0.03943, over 2375509.43 frames. ], batch size: 38, lr: 3.82e-03, grad_scale: 8.0 2023-05-16 19:18:52,502 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=237491.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 19:19:14,921 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=4.88 vs. limit=5.0 2023-05-16 19:19:15,815 INFO [finetune.py:992] (0/2) Epoch 12, batch 2950, loss[loss=0.1834, simple_loss=0.2666, pruned_loss=0.05009, over 12041.00 frames. ], tot_loss[loss=0.1674, simple_loss=0.2566, pruned_loss=0.0391, over 2376038.20 frames. ], batch size: 40, lr: 3.82e-03, grad_scale: 8.0 2023-05-16 19:19:33,687 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=237548.0, num_to_drop=1, layers_to_drop={1} 2023-05-16 19:19:36,632 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=237552.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 19:19:44,499 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=237563.0, num_to_drop=1, layers_to_drop={1} 2023-05-16 19:19:49,810 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.905e+02 2.638e+02 3.009e+02 3.582e+02 2.121e+03, threshold=6.018e+02, percent-clipped=2.0 2023-05-16 19:19:51,955 INFO [finetune.py:992] (0/2) Epoch 12, batch 3000, loss[loss=0.1579, simple_loss=0.2453, pruned_loss=0.03526, over 12137.00 frames. ], tot_loss[loss=0.1671, simple_loss=0.2564, pruned_loss=0.03895, over 2369864.68 frames. ], batch size: 30, lr: 3.82e-03, grad_scale: 8.0 2023-05-16 19:19:51,956 INFO [finetune.py:1017] (0/2) Computing validation loss 2023-05-16 19:20:10,386 INFO [finetune.py:1026] (0/2) Epoch 12, validation: loss=0.3162, simple_loss=0.3931, pruned_loss=0.1196, over 1020973.00 frames. 2023-05-16 19:20:10,386 INFO [finetune.py:1027] (0/2) Maximum memory allocated so far is 12827MB 2023-05-16 19:20:26,404 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=237596.0, num_to_drop=1, layers_to_drop={0} 2023-05-16 19:20:47,071 INFO [finetune.py:992] (0/2) Epoch 12, batch 3050, loss[loss=0.1639, simple_loss=0.2617, pruned_loss=0.03304, over 12095.00 frames. ], tot_loss[loss=0.1679, simple_loss=0.2569, pruned_loss=0.03939, over 2376495.57 frames. ], batch size: 33, lr: 3.82e-03, grad_scale: 8.0 2023-05-16 19:20:47,296 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=237624.0, num_to_drop=1, layers_to_drop={0} 2023-05-16 19:21:03,845 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=3.06 vs. limit=5.0 2023-05-16 19:21:20,225 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.993e+02 2.634e+02 3.161e+02 3.725e+02 7.420e+02, threshold=6.322e+02, percent-clipped=3.0 2023-05-16 19:21:22,467 INFO [finetune.py:992] (0/2) Epoch 12, batch 3100, loss[loss=0.1406, simple_loss=0.232, pruned_loss=0.02464, over 12176.00 frames. ], tot_loss[loss=0.1674, simple_loss=0.2562, pruned_loss=0.03933, over 2376753.71 frames. ], batch size: 31, lr: 3.82e-03, grad_scale: 8.0 2023-05-16 19:21:58,087 INFO [finetune.py:992] (0/2) Epoch 12, batch 3150, loss[loss=0.1764, simple_loss=0.2705, pruned_loss=0.04115, over 12283.00 frames. ], tot_loss[loss=0.167, simple_loss=0.256, pruned_loss=0.03902, over 2372835.70 frames. ], batch size: 37, lr: 3.82e-03, grad_scale: 8.0 2023-05-16 19:22:32,920 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.886e+02 2.636e+02 3.028e+02 3.612e+02 1.044e+03, threshold=6.055e+02, percent-clipped=1.0 2023-05-16 19:22:35,068 INFO [finetune.py:992] (0/2) Epoch 12, batch 3200, loss[loss=0.1494, simple_loss=0.2343, pruned_loss=0.03227, over 12124.00 frames. ], tot_loss[loss=0.1666, simple_loss=0.2555, pruned_loss=0.03888, over 2377495.70 frames. ], batch size: 30, lr: 3.82e-03, grad_scale: 8.0 2023-05-16 19:22:46,700 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=237790.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 19:22:50,321 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.2555, 4.1014, 4.1511, 4.4285, 2.9711, 4.0386, 2.4764, 4.2389], device='cuda:0'), covar=tensor([0.1557, 0.0657, 0.0859, 0.0669, 0.1181, 0.0577, 0.1867, 0.1047], device='cuda:0'), in_proj_covar=tensor([0.0229, 0.0263, 0.0296, 0.0353, 0.0237, 0.0242, 0.0260, 0.0377], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 19:23:08,483 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=3.42 vs. limit=5.0 2023-05-16 19:23:10,847 INFO [finetune.py:992] (0/2) Epoch 12, batch 3250, loss[loss=0.1722, simple_loss=0.2578, pruned_loss=0.04325, over 12342.00 frames. ], tot_loss[loss=0.1667, simple_loss=0.2556, pruned_loss=0.03894, over 2379534.12 frames. ], batch size: 31, lr: 3.82e-03, grad_scale: 8.0 2023-05-16 19:23:26,792 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=237847.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 19:23:29,723 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=237851.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 19:23:43,410 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.760e+02 2.882e+02 3.204e+02 3.746e+02 1.309e+03, threshold=6.408e+02, percent-clipped=2.0 2023-05-16 19:23:45,418 INFO [finetune.py:992] (0/2) Epoch 12, batch 3300, loss[loss=0.1693, simple_loss=0.2645, pruned_loss=0.03708, over 12344.00 frames. ], tot_loss[loss=0.1684, simple_loss=0.2573, pruned_loss=0.03974, over 2369604.77 frames. ], batch size: 36, lr: 3.82e-03, grad_scale: 8.0 2023-05-16 19:24:13,316 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=237912.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 19:24:18,998 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=237919.0, num_to_drop=1, layers_to_drop={3} 2023-05-16 19:24:22,408 INFO [finetune.py:992] (0/2) Epoch 12, batch 3350, loss[loss=0.1401, simple_loss=0.2199, pruned_loss=0.0302, over 11804.00 frames. ], tot_loss[loss=0.1682, simple_loss=0.2573, pruned_loss=0.03962, over 2362666.38 frames. ], batch size: 26, lr: 3.82e-03, grad_scale: 8.0 2023-05-16 19:24:55,995 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.936e+02 2.823e+02 3.216e+02 3.693e+02 6.256e+02, threshold=6.432e+02, percent-clipped=0.0 2023-05-16 19:24:57,701 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=237973.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 19:24:58,224 INFO [finetune.py:992] (0/2) Epoch 12, batch 3400, loss[loss=0.1498, simple_loss=0.2305, pruned_loss=0.03457, over 11795.00 frames. ], tot_loss[loss=0.1672, simple_loss=0.256, pruned_loss=0.03915, over 2363876.99 frames. ], batch size: 26, lr: 3.82e-03, grad_scale: 8.0 2023-05-16 19:25:16,937 INFO [checkpoint.py:75] (0/2) Saving checkpoint to pruned_transducer_stateless7/exp_giga_finetune/checkpoint-138000.pt 2023-05-16 19:25:37,059 INFO [finetune.py:992] (0/2) Epoch 12, batch 3450, loss[loss=0.1657, simple_loss=0.2601, pruned_loss=0.03569, over 12062.00 frames. ], tot_loss[loss=0.1664, simple_loss=0.2553, pruned_loss=0.03871, over 2373918.10 frames. ], batch size: 37, lr: 3.82e-03, grad_scale: 8.0 2023-05-16 19:26:06,993 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=238064.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 19:26:11,657 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.027e+02 2.658e+02 2.911e+02 3.575e+02 8.675e+02, threshold=5.823e+02, percent-clipped=1.0 2023-05-16 19:26:13,821 INFO [finetune.py:992] (0/2) Epoch 12, batch 3500, loss[loss=0.1881, simple_loss=0.2784, pruned_loss=0.04888, over 12193.00 frames. ], tot_loss[loss=0.1666, simple_loss=0.2557, pruned_loss=0.03877, over 2377449.41 frames. ], batch size: 35, lr: 3.82e-03, grad_scale: 8.0 2023-05-16 19:26:14,015 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=238074.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 19:26:49,059 INFO [finetune.py:992] (0/2) Epoch 12, batch 3550, loss[loss=0.1639, simple_loss=0.2638, pruned_loss=0.03206, over 12276.00 frames. ], tot_loss[loss=0.1659, simple_loss=0.255, pruned_loss=0.03841, over 2381258.87 frames. ], batch size: 37, lr: 3.82e-03, grad_scale: 8.0 2023-05-16 19:26:49,998 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=238125.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 19:26:50,722 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=238126.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 19:26:57,180 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=238135.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 19:27:04,911 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=238146.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 19:27:05,736 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=238147.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 19:27:22,611 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.834e+02 2.776e+02 3.155e+02 3.758e+02 5.495e+02, threshold=6.310e+02, percent-clipped=0.0 2023-05-16 19:27:22,848 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.7226, 3.7003, 3.2891, 3.2474, 2.9572, 2.9273, 3.6959, 2.4487], device='cuda:0'), covar=tensor([0.0366, 0.0140, 0.0207, 0.0187, 0.0422, 0.0359, 0.0142, 0.0454], device='cuda:0'), in_proj_covar=tensor([0.0197, 0.0163, 0.0167, 0.0190, 0.0208, 0.0202, 0.0175, 0.0208], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 19:27:25,399 INFO [finetune.py:992] (0/2) Epoch 12, batch 3600, loss[loss=0.1781, simple_loss=0.2676, pruned_loss=0.04431, over 12191.00 frames. ], tot_loss[loss=0.1672, simple_loss=0.2558, pruned_loss=0.03929, over 2375292.21 frames. ], batch size: 35, lr: 3.82e-03, grad_scale: 8.0 2023-05-16 19:27:29,941 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.1124, 2.5502, 3.7213, 3.2142, 3.6161, 3.3367, 2.6711, 3.6431], device='cuda:0'), covar=tensor([0.0175, 0.0399, 0.0153, 0.0292, 0.0132, 0.0199, 0.0380, 0.0117], device='cuda:0'), in_proj_covar=tensor([0.0181, 0.0208, 0.0191, 0.0190, 0.0220, 0.0168, 0.0199, 0.0190], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 19:27:34,790 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=238187.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 19:27:40,288 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=238195.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 19:27:46,291 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.3016, 6.1439, 5.6915, 5.7438, 6.2088, 5.6338, 5.6519, 5.6803], device='cuda:0'), covar=tensor([0.1450, 0.0914, 0.1154, 0.2001, 0.0868, 0.1936, 0.1816, 0.1156], device='cuda:0'), in_proj_covar=tensor([0.0359, 0.0496, 0.0392, 0.0442, 0.0462, 0.0436, 0.0398, 0.0385], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 19:27:58,688 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=238219.0, num_to_drop=1, layers_to_drop={0} 2023-05-16 19:28:02,057 INFO [finetune.py:992] (0/2) Epoch 12, batch 3650, loss[loss=0.195, simple_loss=0.2843, pruned_loss=0.05283, over 12185.00 frames. ], tot_loss[loss=0.1674, simple_loss=0.2561, pruned_loss=0.03935, over 2367611.15 frames. ], batch size: 35, lr: 3.82e-03, grad_scale: 8.0 2023-05-16 19:28:08,622 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.5208, 5.5049, 5.3295, 4.9314, 4.9444, 5.4511, 5.0502, 4.9680], device='cuda:0'), covar=tensor([0.0797, 0.1010, 0.0722, 0.1665, 0.0807, 0.0905, 0.1749, 0.1052], device='cuda:0'), in_proj_covar=tensor([0.0621, 0.0561, 0.0522, 0.0636, 0.0415, 0.0716, 0.0788, 0.0576], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0004, 0.0004, 0.0003], device='cuda:0') 2023-05-16 19:28:13,801 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.5415, 2.9216, 4.6885, 4.8482, 2.8896, 2.6069, 2.9422, 2.1171], device='cuda:0'), covar=tensor([0.1702, 0.2905, 0.0444, 0.0383, 0.1271, 0.2332, 0.2724, 0.4233], device='cuda:0'), in_proj_covar=tensor([0.0294, 0.0387, 0.0273, 0.0300, 0.0272, 0.0302, 0.0382, 0.0373], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-16 19:28:32,680 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=238267.0, num_to_drop=1, layers_to_drop={0} 2023-05-16 19:28:33,372 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=238268.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 19:28:35,467 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.929e+02 2.557e+02 3.166e+02 3.787e+02 8.400e+02, threshold=6.332e+02, percent-clipped=1.0 2023-05-16 19:28:37,671 INFO [finetune.py:992] (0/2) Epoch 12, batch 3700, loss[loss=0.1731, simple_loss=0.2622, pruned_loss=0.042, over 12006.00 frames. ], tot_loss[loss=0.1678, simple_loss=0.2565, pruned_loss=0.03954, over 2373478.80 frames. ], batch size: 40, lr: 3.82e-03, grad_scale: 8.0 2023-05-16 19:29:13,317 INFO [finetune.py:992] (0/2) Epoch 12, batch 3750, loss[loss=0.18, simple_loss=0.2604, pruned_loss=0.0498, over 8446.00 frames. ], tot_loss[loss=0.1695, simple_loss=0.2583, pruned_loss=0.04035, over 2367342.96 frames. ], batch size: 98, lr: 3.82e-03, grad_scale: 8.0 2023-05-16 19:29:35,644 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.5125, 3.5730, 3.1900, 3.1209, 2.8272, 2.7696, 3.5690, 2.2502], device='cuda:0'), covar=tensor([0.0394, 0.0121, 0.0188, 0.0209, 0.0434, 0.0349, 0.0116, 0.0504], device='cuda:0'), in_proj_covar=tensor([0.0198, 0.0163, 0.0167, 0.0192, 0.0208, 0.0203, 0.0176, 0.0209], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 19:29:47,378 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.922e+02 2.858e+02 3.226e+02 3.875e+02 2.077e+03, threshold=6.452e+02, percent-clipped=4.0 2023-05-16 19:29:49,472 INFO [finetune.py:992] (0/2) Epoch 12, batch 3800, loss[loss=0.1908, simple_loss=0.2777, pruned_loss=0.05199, over 12358.00 frames. ], tot_loss[loss=0.1687, simple_loss=0.2577, pruned_loss=0.03984, over 2378247.57 frames. ], batch size: 35, lr: 3.82e-03, grad_scale: 8.0 2023-05-16 19:30:22,750 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=238420.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 19:30:22,827 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=238420.0, num_to_drop=1, layers_to_drop={0} 2023-05-16 19:30:25,412 INFO [finetune.py:992] (0/2) Epoch 12, batch 3850, loss[loss=0.1686, simple_loss=0.2551, pruned_loss=0.04104, over 12121.00 frames. ], tot_loss[loss=0.1691, simple_loss=0.2581, pruned_loss=0.0401, over 2377997.62 frames. ], batch size: 30, lr: 3.81e-03, grad_scale: 8.0 2023-05-16 19:30:27,134 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.4456, 4.8512, 3.0561, 2.8133, 4.1484, 2.5926, 4.1599, 3.3809], device='cuda:0'), covar=tensor([0.0704, 0.0503, 0.1184, 0.1480, 0.0292, 0.1491, 0.0423, 0.0829], device='cuda:0'), in_proj_covar=tensor([0.0191, 0.0256, 0.0178, 0.0200, 0.0142, 0.0183, 0.0199, 0.0177], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 19:30:29,873 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=238430.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 19:30:41,149 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=238446.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 19:30:58,812 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.912e+02 2.697e+02 3.146e+02 3.990e+02 1.679e+03, threshold=6.292e+02, percent-clipped=3.0 2023-05-16 19:31:00,938 INFO [finetune.py:992] (0/2) Epoch 12, batch 3900, loss[loss=0.1822, simple_loss=0.279, pruned_loss=0.04268, over 12359.00 frames. ], tot_loss[loss=0.1701, simple_loss=0.2592, pruned_loss=0.04052, over 2371195.41 frames. ], batch size: 36, lr: 3.81e-03, grad_scale: 8.0 2023-05-16 19:31:03,429 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.7787, 2.9681, 4.4552, 4.5570, 2.9043, 2.7104, 3.0919, 2.1743], device='cuda:0'), covar=tensor([0.1547, 0.2915, 0.0538, 0.0490, 0.1338, 0.2314, 0.2523, 0.4012], device='cuda:0'), in_proj_covar=tensor([0.0296, 0.0389, 0.0275, 0.0301, 0.0273, 0.0304, 0.0383, 0.0375], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-16 19:31:06,184 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=238481.0, num_to_drop=1, layers_to_drop={3} 2023-05-16 19:31:06,782 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=238482.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 19:31:16,249 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=238494.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 19:31:23,477 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.8766, 4.7886, 4.7778, 4.8105, 4.3651, 4.9061, 4.8338, 5.1071], device='cuda:0'), covar=tensor([0.0251, 0.0171, 0.0202, 0.0352, 0.0841, 0.0378, 0.0173, 0.0196], device='cuda:0'), in_proj_covar=tensor([0.0195, 0.0198, 0.0191, 0.0249, 0.0246, 0.0219, 0.0178, 0.0228], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0005, 0.0004, 0.0004, 0.0003, 0.0005], device='cuda:0') 2023-05-16 19:31:37,283 INFO [finetune.py:992] (0/2) Epoch 12, batch 3950, loss[loss=0.2107, simple_loss=0.289, pruned_loss=0.0662, over 7968.00 frames. ], tot_loss[loss=0.1695, simple_loss=0.2585, pruned_loss=0.04023, over 2373119.89 frames. ], batch size: 98, lr: 3.81e-03, grad_scale: 8.0 2023-05-16 19:31:52,555 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 2023-05-16 19:32:02,284 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.1863, 4.7407, 5.1211, 4.5027, 4.7898, 4.6001, 5.1822, 4.8164], device='cuda:0'), covar=tensor([0.0236, 0.0400, 0.0290, 0.0264, 0.0319, 0.0334, 0.0225, 0.0334], device='cuda:0'), in_proj_covar=tensor([0.0257, 0.0262, 0.0287, 0.0259, 0.0260, 0.0259, 0.0233, 0.0215], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 19:32:08,746 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=238568.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 19:32:10,601 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.816e+02 2.647e+02 3.036e+02 3.542e+02 7.006e+02, threshold=6.071e+02, percent-clipped=2.0 2023-05-16 19:32:12,714 INFO [finetune.py:992] (0/2) Epoch 12, batch 4000, loss[loss=0.1376, simple_loss=0.2148, pruned_loss=0.0302, over 11988.00 frames. ], tot_loss[loss=0.1687, simple_loss=0.2574, pruned_loss=0.03996, over 2380637.21 frames. ], batch size: 28, lr: 3.81e-03, grad_scale: 8.0 2023-05-16 19:32:24,323 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.14 vs. limit=2.0 2023-05-16 19:32:43,456 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=238616.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 19:32:49,125 INFO [finetune.py:992] (0/2) Epoch 12, batch 4050, loss[loss=0.1654, simple_loss=0.2514, pruned_loss=0.03967, over 10447.00 frames. ], tot_loss[loss=0.1685, simple_loss=0.2575, pruned_loss=0.03975, over 2375938.95 frames. ], batch size: 68, lr: 3.81e-03, grad_scale: 8.0 2023-05-16 19:33:23,036 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.815e+02 2.680e+02 3.070e+02 3.796e+02 8.784e+02, threshold=6.139e+02, percent-clipped=3.0 2023-05-16 19:33:25,215 INFO [finetune.py:992] (0/2) Epoch 12, batch 4100, loss[loss=0.1669, simple_loss=0.2492, pruned_loss=0.04231, over 12261.00 frames. ], tot_loss[loss=0.1686, simple_loss=0.2573, pruned_loss=0.0399, over 2371324.40 frames. ], batch size: 32, lr: 3.81e-03, grad_scale: 8.0 2023-05-16 19:33:30,350 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=238681.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 19:33:40,478 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.4617, 4.8517, 3.1568, 2.6275, 4.1415, 2.6056, 4.1516, 3.3885], device='cuda:0'), covar=tensor([0.0664, 0.0477, 0.0964, 0.1632, 0.0292, 0.1384, 0.0454, 0.0761], device='cuda:0'), in_proj_covar=tensor([0.0190, 0.0256, 0.0176, 0.0199, 0.0141, 0.0181, 0.0198, 0.0176], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 19:33:58,318 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=238720.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 19:34:01,076 INFO [finetune.py:992] (0/2) Epoch 12, batch 4150, loss[loss=0.149, simple_loss=0.2233, pruned_loss=0.03741, over 12003.00 frames. ], tot_loss[loss=0.1673, simple_loss=0.2561, pruned_loss=0.03922, over 2377831.85 frames. ], batch size: 28, lr: 3.81e-03, grad_scale: 8.0 2023-05-16 19:34:05,518 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=238730.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 19:34:12,837 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([6.2345, 6.1795, 5.9575, 5.4273, 5.3239, 6.0738, 5.7050, 5.4515], device='cuda:0'), covar=tensor([0.0654, 0.0954, 0.0652, 0.1486, 0.0601, 0.0678, 0.1523, 0.0945], device='cuda:0'), in_proj_covar=tensor([0.0621, 0.0560, 0.0523, 0.0637, 0.0417, 0.0715, 0.0789, 0.0578], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0004, 0.0004, 0.0003], device='cuda:0') 2023-05-16 19:34:14,327 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=238742.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 19:34:32,162 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.1441, 4.8814, 5.1149, 4.4541, 4.7591, 4.5132, 5.1213, 4.8619], device='cuda:0'), covar=tensor([0.0347, 0.0393, 0.0405, 0.0323, 0.0421, 0.0420, 0.0353, 0.0371], device='cuda:0'), in_proj_covar=tensor([0.0256, 0.0260, 0.0286, 0.0258, 0.0258, 0.0258, 0.0232, 0.0214], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 19:34:33,454 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=238768.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 19:34:35,370 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.751e+02 2.514e+02 2.973e+02 3.546e+02 5.463e+02, threshold=5.946e+02, percent-clipped=0.0 2023-05-16 19:34:37,461 INFO [finetune.py:992] (0/2) Epoch 12, batch 4200, loss[loss=0.1461, simple_loss=0.2294, pruned_loss=0.03145, over 12009.00 frames. ], tot_loss[loss=0.1669, simple_loss=0.2557, pruned_loss=0.03902, over 2379186.23 frames. ], batch size: 28, lr: 3.81e-03, grad_scale: 8.0 2023-05-16 19:34:38,973 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=238776.0, num_to_drop=1, layers_to_drop={0} 2023-05-16 19:34:40,410 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=238778.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 19:34:43,351 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=238782.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 19:35:13,785 INFO [finetune.py:992] (0/2) Epoch 12, batch 4250, loss[loss=0.1761, simple_loss=0.2697, pruned_loss=0.04127, over 12006.00 frames. ], tot_loss[loss=0.1674, simple_loss=0.2563, pruned_loss=0.03919, over 2383894.45 frames. ], batch size: 40, lr: 3.81e-03, grad_scale: 8.0 2023-05-16 19:35:18,048 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=238830.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 19:35:21,126 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 2023-05-16 19:35:34,495 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.5744, 4.4738, 4.5401, 4.6193, 4.2973, 4.3550, 4.1798, 4.5420], device='cuda:0'), covar=tensor([0.0898, 0.0655, 0.1024, 0.0678, 0.2030, 0.1404, 0.0596, 0.0957], device='cuda:0'), in_proj_covar=tensor([0.0553, 0.0714, 0.0619, 0.0630, 0.0867, 0.0757, 0.0557, 0.0486], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0005, 0.0004, 0.0004, 0.0005, 0.0005, 0.0003, 0.0003], device='cuda:0') 2023-05-16 19:35:47,020 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.988e+02 2.596e+02 3.163e+02 3.758e+02 5.764e+02, threshold=6.325e+02, percent-clipped=0.0 2023-05-16 19:35:47,461 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.79 vs. limit=2.0 2023-05-16 19:35:49,159 INFO [finetune.py:992] (0/2) Epoch 12, batch 4300, loss[loss=0.1503, simple_loss=0.2393, pruned_loss=0.0306, over 11997.00 frames. ], tot_loss[loss=0.1682, simple_loss=0.2576, pruned_loss=0.03943, over 2380673.64 frames. ], batch size: 28, lr: 3.81e-03, grad_scale: 8.0 2023-05-16 19:36:00,629 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.5384, 4.8323, 4.2058, 5.1673, 4.6973, 2.8054, 4.4082, 3.1757], device='cuda:0'), covar=tensor([0.0737, 0.0713, 0.1398, 0.0436, 0.1102, 0.1838, 0.1081, 0.3253], device='cuda:0'), in_proj_covar=tensor([0.0305, 0.0380, 0.0362, 0.0310, 0.0368, 0.0272, 0.0346, 0.0369], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 19:36:24,940 INFO [finetune.py:992] (0/2) Epoch 12, batch 4350, loss[loss=0.1778, simple_loss=0.2684, pruned_loss=0.04355, over 12043.00 frames. ], tot_loss[loss=0.1678, simple_loss=0.2571, pruned_loss=0.03927, over 2373823.15 frames. ], batch size: 37, lr: 3.81e-03, grad_scale: 8.0 2023-05-16 19:36:33,472 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=238936.0, num_to_drop=1, layers_to_drop={1} 2023-05-16 19:36:58,690 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.500e+02 2.722e+02 3.105e+02 3.870e+02 7.844e+02, threshold=6.211e+02, percent-clipped=1.0 2023-05-16 19:37:00,885 INFO [finetune.py:992] (0/2) Epoch 12, batch 4400, loss[loss=0.1844, simple_loss=0.2809, pruned_loss=0.04389, over 12148.00 frames. ], tot_loss[loss=0.1679, simple_loss=0.2571, pruned_loss=0.03935, over 2369181.99 frames. ], batch size: 36, lr: 3.81e-03, grad_scale: 8.0 2023-05-16 19:37:17,340 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=238997.0, num_to_drop=1, layers_to_drop={2} 2023-05-16 19:37:20,677 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.88 vs. limit=2.0 2023-05-16 19:37:36,408 INFO [finetune.py:992] (0/2) Epoch 12, batch 4450, loss[loss=0.154, simple_loss=0.2357, pruned_loss=0.03611, over 12022.00 frames. ], tot_loss[loss=0.1688, simple_loss=0.2579, pruned_loss=0.03987, over 2372893.94 frames. ], batch size: 31, lr: 3.81e-03, grad_scale: 16.0 2023-05-16 19:37:45,629 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=239037.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 19:38:10,405 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.932e+02 2.792e+02 3.466e+02 4.003e+02 6.008e+02, threshold=6.932e+02, percent-clipped=0.0 2023-05-16 19:38:12,508 INFO [finetune.py:992] (0/2) Epoch 12, batch 4500, loss[loss=0.2242, simple_loss=0.2952, pruned_loss=0.07661, over 7496.00 frames. ], tot_loss[loss=0.1689, simple_loss=0.2579, pruned_loss=0.04001, over 2363467.02 frames. ], batch size: 98, lr: 3.81e-03, grad_scale: 16.0 2023-05-16 19:38:14,007 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=239076.0, num_to_drop=1, layers_to_drop={0} 2023-05-16 19:38:48,043 INFO [finetune.py:992] (0/2) Epoch 12, batch 4550, loss[loss=0.1848, simple_loss=0.276, pruned_loss=0.04684, over 12120.00 frames. ], tot_loss[loss=0.1687, simple_loss=0.2576, pruned_loss=0.03991, over 2367855.45 frames. ], batch size: 39, lr: 3.81e-03, grad_scale: 16.0 2023-05-16 19:38:48,104 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=239124.0, num_to_drop=1, layers_to_drop={1} 2023-05-16 19:38:54,432 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.0316, 4.8497, 4.9613, 5.0284, 4.6801, 4.7634, 4.5056, 4.9664], device='cuda:0'), covar=tensor([0.0697, 0.0674, 0.0857, 0.0522, 0.1771, 0.1305, 0.0539, 0.0891], device='cuda:0'), in_proj_covar=tensor([0.0548, 0.0708, 0.0616, 0.0627, 0.0859, 0.0755, 0.0554, 0.0482], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0005, 0.0004, 0.0004, 0.0005, 0.0005, 0.0003, 0.0003], device='cuda:0') 2023-05-16 19:39:21,052 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.926e+02 2.710e+02 3.133e+02 3.780e+02 8.254e+02, threshold=6.265e+02, percent-clipped=2.0 2023-05-16 19:39:23,230 INFO [finetune.py:992] (0/2) Epoch 12, batch 4600, loss[loss=0.1738, simple_loss=0.264, pruned_loss=0.04181, over 12280.00 frames. ], tot_loss[loss=0.1683, simple_loss=0.2574, pruned_loss=0.03962, over 2374877.11 frames. ], batch size: 37, lr: 3.81e-03, grad_scale: 16.0 2023-05-16 19:39:30,299 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.3630, 5.0842, 5.2994, 4.6591, 5.0446, 4.7846, 5.3019, 4.9779], device='cuda:0'), covar=tensor([0.0267, 0.0332, 0.0386, 0.0284, 0.0352, 0.0318, 0.0332, 0.0281], device='cuda:0'), in_proj_covar=tensor([0.0260, 0.0263, 0.0290, 0.0263, 0.0261, 0.0263, 0.0237, 0.0217], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 19:39:59,402 INFO [finetune.py:992] (0/2) Epoch 12, batch 4650, loss[loss=0.1525, simple_loss=0.2421, pruned_loss=0.03151, over 12085.00 frames. ], tot_loss[loss=0.1682, simple_loss=0.257, pruned_loss=0.03966, over 2375952.00 frames. ], batch size: 32, lr: 3.81e-03, grad_scale: 16.0 2023-05-16 19:40:33,917 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.878e+02 2.619e+02 3.215e+02 4.293e+02 1.000e+03, threshold=6.429e+02, percent-clipped=5.0 2023-05-16 19:40:35,239 INFO [finetune.py:992] (0/2) Epoch 12, batch 4700, loss[loss=0.208, simple_loss=0.292, pruned_loss=0.06204, over 10316.00 frames. ], tot_loss[loss=0.1678, simple_loss=0.2568, pruned_loss=0.03943, over 2373213.25 frames. ], batch size: 68, lr: 3.81e-03, grad_scale: 8.0 2023-05-16 19:40:48,416 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=239292.0, num_to_drop=1, layers_to_drop={1} 2023-05-16 19:40:54,206 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.5090, 5.0650, 5.4544, 4.7961, 5.1080, 4.9101, 5.5012, 5.0971], device='cuda:0'), covar=tensor([0.0232, 0.0399, 0.0289, 0.0257, 0.0358, 0.0314, 0.0204, 0.0263], device='cuda:0'), in_proj_covar=tensor([0.0261, 0.0264, 0.0291, 0.0263, 0.0261, 0.0264, 0.0238, 0.0217], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 19:41:11,120 INFO [finetune.py:992] (0/2) Epoch 12, batch 4750, loss[loss=0.1889, simple_loss=0.2848, pruned_loss=0.04645, over 11362.00 frames. ], tot_loss[loss=0.1674, simple_loss=0.2563, pruned_loss=0.03927, over 2372484.94 frames. ], batch size: 55, lr: 3.81e-03, grad_scale: 8.0 2023-05-16 19:41:21,146 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=239337.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 19:41:45,256 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=2.51 vs. limit=5.0 2023-05-16 19:41:46,345 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.819e+02 2.662e+02 3.103e+02 3.697e+02 5.990e+02, threshold=6.207e+02, percent-clipped=0.0 2023-05-16 19:41:47,710 INFO [finetune.py:992] (0/2) Epoch 12, batch 4800, loss[loss=0.1997, simple_loss=0.2883, pruned_loss=0.05557, over 12126.00 frames. ], tot_loss[loss=0.1682, simple_loss=0.2572, pruned_loss=0.03957, over 2368568.30 frames. ], batch size: 39, lr: 3.81e-03, grad_scale: 8.0 2023-05-16 19:41:55,609 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=239385.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 19:42:03,069 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.92 vs. limit=2.0 2023-05-16 19:42:09,652 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.31 vs. limit=2.0 2023-05-16 19:42:23,428 INFO [finetune.py:992] (0/2) Epoch 12, batch 4850, loss[loss=0.1733, simple_loss=0.2586, pruned_loss=0.04401, over 12352.00 frames. ], tot_loss[loss=0.1678, simple_loss=0.2568, pruned_loss=0.03938, over 2377520.61 frames. ], batch size: 30, lr: 3.81e-03, grad_scale: 8.0 2023-05-16 19:42:57,300 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.806e+02 2.566e+02 2.984e+02 3.406e+02 6.973e+02, threshold=5.968e+02, percent-clipped=1.0 2023-05-16 19:42:58,752 INFO [finetune.py:992] (0/2) Epoch 12, batch 4900, loss[loss=0.1344, simple_loss=0.22, pruned_loss=0.02434, over 12343.00 frames. ], tot_loss[loss=0.1677, simple_loss=0.2567, pruned_loss=0.03933, over 2379123.79 frames. ], batch size: 30, lr: 3.81e-03, grad_scale: 8.0 2023-05-16 19:43:02,399 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.3332, 3.1589, 4.6802, 2.3651, 2.7169, 3.5769, 3.1549, 3.7627], device='cuda:0'), covar=tensor([0.0478, 0.1138, 0.0272, 0.1259, 0.1810, 0.1455, 0.1322, 0.1074], device='cuda:0'), in_proj_covar=tensor([0.0238, 0.0240, 0.0253, 0.0186, 0.0241, 0.0299, 0.0225, 0.0266], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-16 19:43:08,532 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.3504, 4.7651, 2.9658, 2.7051, 4.0688, 2.6050, 4.0118, 3.2846], device='cuda:0'), covar=tensor([0.0722, 0.0607, 0.1132, 0.1492, 0.0252, 0.1318, 0.0467, 0.0790], device='cuda:0'), in_proj_covar=tensor([0.0189, 0.0256, 0.0176, 0.0199, 0.0141, 0.0181, 0.0197, 0.0176], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 19:43:18,492 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.4829, 4.7663, 3.0532, 2.6973, 4.1023, 2.5007, 4.0029, 3.2146], device='cuda:0'), covar=tensor([0.0612, 0.0557, 0.1085, 0.1486, 0.0251, 0.1428, 0.0490, 0.0814], device='cuda:0'), in_proj_covar=tensor([0.0188, 0.0254, 0.0175, 0.0198, 0.0140, 0.0180, 0.0197, 0.0175], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 19:43:35,465 INFO [finetune.py:992] (0/2) Epoch 12, batch 4950, loss[loss=0.157, simple_loss=0.2393, pruned_loss=0.03734, over 12364.00 frames. ], tot_loss[loss=0.1677, simple_loss=0.2566, pruned_loss=0.03939, over 2371816.54 frames. ], batch size: 30, lr: 3.80e-03, grad_scale: 8.0 2023-05-16 19:43:49,919 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.3252, 4.1736, 4.2034, 4.5969, 3.0351, 4.1208, 2.5900, 4.3044], device='cuda:0'), covar=tensor([0.1476, 0.0679, 0.0892, 0.0710, 0.1151, 0.0547, 0.1783, 0.1198], device='cuda:0'), in_proj_covar=tensor([0.0228, 0.0265, 0.0296, 0.0355, 0.0238, 0.0242, 0.0260, 0.0375], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 19:44:09,767 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.945e+02 2.810e+02 3.291e+02 4.221e+02 1.405e+03, threshold=6.582e+02, percent-clipped=5.0 2023-05-16 19:44:09,996 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.0301, 2.3830, 2.3373, 2.2883, 2.1687, 2.0277, 2.3858, 1.7033], device='cuda:0'), covar=tensor([0.0364, 0.0198, 0.0239, 0.0201, 0.0402, 0.0275, 0.0190, 0.0444], device='cuda:0'), in_proj_covar=tensor([0.0196, 0.0161, 0.0167, 0.0191, 0.0206, 0.0201, 0.0174, 0.0208], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 19:44:11,188 INFO [finetune.py:992] (0/2) Epoch 12, batch 5000, loss[loss=0.174, simple_loss=0.2515, pruned_loss=0.04826, over 12139.00 frames. ], tot_loss[loss=0.1678, simple_loss=0.2563, pruned_loss=0.03963, over 2362288.68 frames. ], batch size: 30, lr: 3.80e-03, grad_scale: 8.0 2023-05-16 19:44:24,418 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=239592.0, num_to_drop=1, layers_to_drop={0} 2023-05-16 19:44:46,575 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=239622.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 19:44:47,843 INFO [finetune.py:992] (0/2) Epoch 12, batch 5050, loss[loss=0.171, simple_loss=0.259, pruned_loss=0.04147, over 12018.00 frames. ], tot_loss[loss=0.1677, simple_loss=0.2565, pruned_loss=0.03942, over 2373045.17 frames. ], batch size: 40, lr: 3.80e-03, grad_scale: 8.0 2023-05-16 19:44:59,519 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=239640.0, num_to_drop=1, layers_to_drop={0} 2023-05-16 19:44:59,666 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.1507, 3.9544, 4.1583, 4.4880, 2.9409, 3.8937, 2.4629, 4.1436], device='cuda:0'), covar=tensor([0.1603, 0.0827, 0.0876, 0.0590, 0.1147, 0.0669, 0.1995, 0.1263], device='cuda:0'), in_proj_covar=tensor([0.0229, 0.0266, 0.0298, 0.0357, 0.0239, 0.0244, 0.0262, 0.0377], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 19:45:22,689 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.608e+02 2.849e+02 3.125e+02 3.864e+02 7.121e+02, threshold=6.249e+02, percent-clipped=1.0 2023-05-16 19:45:24,109 INFO [finetune.py:992] (0/2) Epoch 12, batch 5100, loss[loss=0.1564, simple_loss=0.258, pruned_loss=0.02741, over 12157.00 frames. ], tot_loss[loss=0.1676, simple_loss=0.2561, pruned_loss=0.03949, over 2366100.67 frames. ], batch size: 34, lr: 3.80e-03, grad_scale: 8.0 2023-05-16 19:45:30,562 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=239683.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 19:45:49,968 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.4264, 4.9688, 5.3708, 4.7045, 5.0345, 4.7830, 5.4307, 5.0510], device='cuda:0'), covar=tensor([0.0251, 0.0361, 0.0266, 0.0279, 0.0367, 0.0369, 0.0200, 0.0282], device='cuda:0'), in_proj_covar=tensor([0.0259, 0.0263, 0.0289, 0.0262, 0.0260, 0.0262, 0.0236, 0.0217], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 19:45:52,288 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.7502, 4.0626, 3.6052, 4.1611, 3.8774, 2.4265, 3.4847, 2.9121], device='cuda:0'), covar=tensor([0.0917, 0.0914, 0.1529, 0.0864, 0.1285, 0.1918, 0.1526, 0.3224], device='cuda:0'), in_proj_covar=tensor([0.0305, 0.0380, 0.0360, 0.0310, 0.0369, 0.0272, 0.0346, 0.0369], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 19:45:59,778 INFO [finetune.py:992] (0/2) Epoch 12, batch 5150, loss[loss=0.15, simple_loss=0.2314, pruned_loss=0.03434, over 12346.00 frames. ], tot_loss[loss=0.1666, simple_loss=0.2549, pruned_loss=0.03921, over 2368933.85 frames. ], batch size: 30, lr: 3.80e-03, grad_scale: 8.0 2023-05-16 19:46:14,405 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=3.43 vs. limit=5.0 2023-05-16 19:46:34,242 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.015e+02 2.573e+02 3.136e+02 3.793e+02 5.974e+02, threshold=6.273e+02, percent-clipped=0.0 2023-05-16 19:46:35,697 INFO [finetune.py:992] (0/2) Epoch 12, batch 5200, loss[loss=0.1881, simple_loss=0.2789, pruned_loss=0.04866, over 12124.00 frames. ], tot_loss[loss=0.1677, simple_loss=0.2563, pruned_loss=0.03959, over 2369281.66 frames. ], batch size: 39, lr: 3.80e-03, grad_scale: 8.0 2023-05-16 19:46:56,815 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.3824, 2.5070, 3.0387, 4.2234, 2.1633, 4.2859, 4.3596, 4.4284], device='cuda:0'), covar=tensor([0.0139, 0.1275, 0.0566, 0.0196, 0.1504, 0.0259, 0.0141, 0.0105], device='cuda:0'), in_proj_covar=tensor([0.0126, 0.0206, 0.0186, 0.0121, 0.0194, 0.0182, 0.0179, 0.0120], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 19:47:11,115 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([1.8948, 4.1183, 4.2454, 4.2874, 2.7360, 3.9566, 2.6972, 4.1280], device='cuda:0'), covar=tensor([0.1713, 0.0644, 0.0633, 0.0435, 0.1194, 0.0631, 0.1733, 0.1067], device='cuda:0'), in_proj_covar=tensor([0.0228, 0.0266, 0.0297, 0.0356, 0.0238, 0.0242, 0.0260, 0.0375], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 19:47:12,306 INFO [finetune.py:992] (0/2) Epoch 12, batch 5250, loss[loss=0.158, simple_loss=0.244, pruned_loss=0.03603, over 11400.00 frames. ], tot_loss[loss=0.1675, simple_loss=0.2562, pruned_loss=0.03945, over 2366189.84 frames. ], batch size: 25, lr: 3.80e-03, grad_scale: 8.0 2023-05-16 19:47:21,323 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.90 vs. limit=2.0 2023-05-16 19:47:46,313 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.935e+02 2.888e+02 3.226e+02 3.799e+02 6.835e+02, threshold=6.451e+02, percent-clipped=1.0 2023-05-16 19:47:47,713 INFO [finetune.py:992] (0/2) Epoch 12, batch 5300, loss[loss=0.1826, simple_loss=0.2713, pruned_loss=0.04693, over 12031.00 frames. ], tot_loss[loss=0.1675, simple_loss=0.2562, pruned_loss=0.03941, over 2365474.02 frames. ], batch size: 40, lr: 3.80e-03, grad_scale: 8.0 2023-05-16 19:48:05,449 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.9434, 3.9053, 3.9126, 4.0221, 3.7744, 3.8046, 3.6977, 3.9572], device='cuda:0'), covar=tensor([0.1240, 0.0727, 0.1466, 0.0694, 0.1847, 0.1456, 0.0626, 0.0905], device='cuda:0'), in_proj_covar=tensor([0.0547, 0.0706, 0.0615, 0.0629, 0.0859, 0.0754, 0.0556, 0.0482], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0005, 0.0004, 0.0004, 0.0005, 0.0005, 0.0003, 0.0003], device='cuda:0') 2023-05-16 19:48:14,783 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.4219, 2.3879, 3.7191, 4.3523, 3.9095, 4.3415, 3.7740, 2.8485], device='cuda:0'), covar=tensor([0.0043, 0.0413, 0.0158, 0.0049, 0.0130, 0.0084, 0.0137, 0.0435], device='cuda:0'), in_proj_covar=tensor([0.0088, 0.0124, 0.0107, 0.0078, 0.0103, 0.0116, 0.0097, 0.0138], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 19:48:24,062 INFO [finetune.py:992] (0/2) Epoch 12, batch 5350, loss[loss=0.2259, simple_loss=0.3057, pruned_loss=0.0731, over 7565.00 frames. ], tot_loss[loss=0.1685, simple_loss=0.2575, pruned_loss=0.0398, over 2353005.25 frames. ], batch size: 98, lr: 3.80e-03, grad_scale: 8.0 2023-05-16 19:48:41,593 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.8003, 2.8601, 4.6619, 4.7582, 2.8664, 2.7210, 2.9896, 2.2142], device='cuda:0'), covar=tensor([0.1500, 0.2913, 0.0409, 0.0391, 0.1300, 0.2274, 0.2637, 0.4005], device='cuda:0'), in_proj_covar=tensor([0.0293, 0.0386, 0.0273, 0.0299, 0.0270, 0.0302, 0.0379, 0.0371], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-16 19:48:59,111 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.896e+02 2.651e+02 3.240e+02 3.822e+02 6.055e+02, threshold=6.479e+02, percent-clipped=0.0 2023-05-16 19:48:59,623 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.38 vs. limit=2.0 2023-05-16 19:49:00,575 INFO [finetune.py:992] (0/2) Epoch 12, batch 5400, loss[loss=0.1519, simple_loss=0.2299, pruned_loss=0.03696, over 11876.00 frames. ], tot_loss[loss=0.1672, simple_loss=0.256, pruned_loss=0.0392, over 2359407.78 frames. ], batch size: 26, lr: 3.80e-03, grad_scale: 8.0 2023-05-16 19:49:03,473 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=239978.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 19:49:10,954 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2023-05-16 19:49:13,088 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=4.65 vs. limit=5.0 2023-05-16 19:49:19,281 INFO [checkpoint.py:75] (0/2) Saving checkpoint to pruned_transducer_stateless7/exp_giga_finetune/checkpoint-140000.pt 2023-05-16 19:49:39,036 INFO [finetune.py:992] (0/2) Epoch 12, batch 5450, loss[loss=0.1786, simple_loss=0.2773, pruned_loss=0.03997, over 12158.00 frames. ], tot_loss[loss=0.1682, simple_loss=0.2571, pruned_loss=0.03958, over 2366844.73 frames. ], batch size: 36, lr: 3.80e-03, grad_scale: 8.0 2023-05-16 19:49:48,348 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.7759, 4.3879, 4.6485, 4.7386, 4.4922, 4.7155, 4.5838, 2.4775], device='cuda:0'), covar=tensor([0.0096, 0.0081, 0.0076, 0.0060, 0.0050, 0.0093, 0.0065, 0.0805], device='cuda:0'), in_proj_covar=tensor([0.0070, 0.0079, 0.0081, 0.0074, 0.0060, 0.0093, 0.0082, 0.0098], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-16 19:50:13,740 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.772e+02 2.697e+02 3.215e+02 4.095e+02 7.076e+02, threshold=6.429e+02, percent-clipped=3.0 2023-05-16 19:50:15,182 INFO [finetune.py:992] (0/2) Epoch 12, batch 5500, loss[loss=0.151, simple_loss=0.2466, pruned_loss=0.02771, over 12294.00 frames. ], tot_loss[loss=0.1678, simple_loss=0.2571, pruned_loss=0.0393, over 2374065.21 frames. ], batch size: 33, lr: 3.80e-03, grad_scale: 8.0 2023-05-16 19:50:51,050 INFO [finetune.py:992] (0/2) Epoch 12, batch 5550, loss[loss=0.1415, simple_loss=0.2246, pruned_loss=0.02919, over 12183.00 frames. ], tot_loss[loss=0.1679, simple_loss=0.2569, pruned_loss=0.03947, over 2374560.30 frames. ], batch size: 29, lr: 3.80e-03, grad_scale: 8.0 2023-05-16 19:51:25,101 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.976e+02 2.669e+02 3.151e+02 3.800e+02 8.227e+02, threshold=6.301e+02, percent-clipped=3.0 2023-05-16 19:51:26,579 INFO [finetune.py:992] (0/2) Epoch 12, batch 5600, loss[loss=0.1581, simple_loss=0.2511, pruned_loss=0.0325, over 12098.00 frames. ], tot_loss[loss=0.1667, simple_loss=0.2558, pruned_loss=0.03878, over 2377308.47 frames. ], batch size: 33, lr: 3.80e-03, grad_scale: 8.0 2023-05-16 19:51:54,622 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.3013, 2.7852, 3.9095, 3.3399, 3.7243, 3.4415, 2.8853, 3.8032], device='cuda:0'), covar=tensor([0.0140, 0.0344, 0.0148, 0.0221, 0.0135, 0.0178, 0.0309, 0.0124], device='cuda:0'), in_proj_covar=tensor([0.0182, 0.0209, 0.0192, 0.0193, 0.0223, 0.0169, 0.0199, 0.0194], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 19:52:02,939 INFO [finetune.py:992] (0/2) Epoch 12, batch 5650, loss[loss=0.2071, simple_loss=0.3001, pruned_loss=0.0571, over 12103.00 frames. ], tot_loss[loss=0.1676, simple_loss=0.2569, pruned_loss=0.03916, over 2367392.60 frames. ], batch size: 38, lr: 3.80e-03, grad_scale: 8.0 2023-05-16 19:52:08,323 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.58 vs. limit=2.0 2023-05-16 19:52:12,456 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.33 vs. limit=2.0 2023-05-16 19:52:37,553 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.622e+02 2.686e+02 3.253e+02 3.697e+02 6.012e+02, threshold=6.505e+02, percent-clipped=0.0 2023-05-16 19:52:38,896 INFO [finetune.py:992] (0/2) Epoch 12, batch 5700, loss[loss=0.1451, simple_loss=0.227, pruned_loss=0.03161, over 12348.00 frames. ], tot_loss[loss=0.1678, simple_loss=0.2569, pruned_loss=0.03934, over 2371433.35 frames. ], batch size: 30, lr: 3.80e-03, grad_scale: 8.0 2023-05-16 19:52:41,943 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=240278.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 19:53:07,445 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.0292, 4.9431, 4.8688, 4.8705, 4.5360, 5.0507, 5.0593, 5.2591], device='cuda:0'), covar=tensor([0.0254, 0.0168, 0.0192, 0.0363, 0.0805, 0.0305, 0.0145, 0.0176], device='cuda:0'), in_proj_covar=tensor([0.0197, 0.0201, 0.0193, 0.0251, 0.0249, 0.0219, 0.0180, 0.0229], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0005, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:0') 2023-05-16 19:53:14,238 INFO [finetune.py:992] (0/2) Epoch 12, batch 5750, loss[loss=0.1624, simple_loss=0.2526, pruned_loss=0.03612, over 11587.00 frames. ], tot_loss[loss=0.1674, simple_loss=0.2564, pruned_loss=0.03924, over 2373948.90 frames. ], batch size: 48, lr: 3.80e-03, grad_scale: 8.0 2023-05-16 19:53:15,758 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=240326.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 19:53:25,905 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=240340.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 19:53:36,058 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.1695, 4.5257, 2.7833, 2.4821, 3.8961, 2.4106, 3.9608, 2.9880], device='cuda:0'), covar=tensor([0.0850, 0.0482, 0.1135, 0.1624, 0.0334, 0.1433, 0.0517, 0.0934], device='cuda:0'), in_proj_covar=tensor([0.0191, 0.0256, 0.0177, 0.0200, 0.0141, 0.0181, 0.0200, 0.0177], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 19:53:49,053 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.047e+02 2.763e+02 3.303e+02 4.001e+02 5.915e+02, threshold=6.607e+02, percent-clipped=0.0 2023-05-16 19:53:50,503 INFO [finetune.py:992] (0/2) Epoch 12, batch 5800, loss[loss=0.1611, simple_loss=0.2541, pruned_loss=0.03402, over 11292.00 frames. ], tot_loss[loss=0.1675, simple_loss=0.2563, pruned_loss=0.03931, over 2377371.11 frames. ], batch size: 55, lr: 3.80e-03, grad_scale: 8.0 2023-05-16 19:53:54,953 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.8719, 3.6393, 3.6964, 3.8263, 3.5450, 3.9828, 3.9497, 4.0339], device='cuda:0'), covar=tensor([0.0269, 0.0243, 0.0233, 0.0542, 0.0656, 0.0379, 0.0193, 0.0260], device='cuda:0'), in_proj_covar=tensor([0.0197, 0.0201, 0.0193, 0.0251, 0.0248, 0.0219, 0.0179, 0.0229], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0005, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:0') 2023-05-16 19:54:10,870 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=240401.0, num_to_drop=1, layers_to_drop={0} 2023-05-16 19:54:19,345 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.4526, 5.2443, 5.3328, 5.4186, 5.0435, 5.1105, 4.8795, 5.3881], device='cuda:0'), covar=tensor([0.0576, 0.0588, 0.0757, 0.0528, 0.1794, 0.1269, 0.0581, 0.0840], device='cuda:0'), in_proj_covar=tensor([0.0549, 0.0711, 0.0619, 0.0632, 0.0862, 0.0758, 0.0559, 0.0487], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0005, 0.0004, 0.0004, 0.0005, 0.0005, 0.0003, 0.0003], device='cuda:0') 2023-05-16 19:54:25,986 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.93 vs. limit=2.0 2023-05-16 19:54:26,962 INFO [finetune.py:992] (0/2) Epoch 12, batch 5850, loss[loss=0.1773, simple_loss=0.2658, pruned_loss=0.04444, over 12128.00 frames. ], tot_loss[loss=0.1683, simple_loss=0.2568, pruned_loss=0.03987, over 2370308.95 frames. ], batch size: 38, lr: 3.80e-03, grad_scale: 8.0 2023-05-16 19:55:00,850 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.903e+02 2.627e+02 3.128e+02 3.726e+02 9.792e+02, threshold=6.257e+02, percent-clipped=3.0 2023-05-16 19:55:02,344 INFO [finetune.py:992] (0/2) Epoch 12, batch 5900, loss[loss=0.2426, simple_loss=0.3178, pruned_loss=0.08365, over 8060.00 frames. ], tot_loss[loss=0.1685, simple_loss=0.2572, pruned_loss=0.03988, over 2371482.74 frames. ], batch size: 99, lr: 3.80e-03, grad_scale: 8.0 2023-05-16 19:55:09,605 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.6321, 2.7742, 4.4786, 4.7861, 2.8927, 2.6280, 2.9917, 2.1401], device='cuda:0'), covar=tensor([0.1656, 0.3253, 0.0505, 0.0355, 0.1320, 0.2353, 0.2731, 0.4146], device='cuda:0'), in_proj_covar=tensor([0.0297, 0.0389, 0.0276, 0.0302, 0.0273, 0.0307, 0.0383, 0.0376], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-16 19:55:14,362 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.4051, 5.1202, 4.6541, 4.6975, 5.2661, 4.6318, 4.7322, 4.5441], device='cuda:0'), covar=tensor([0.1674, 0.1151, 0.1321, 0.1890, 0.1066, 0.2146, 0.1991, 0.1316], device='cuda:0'), in_proj_covar=tensor([0.0368, 0.0510, 0.0401, 0.0444, 0.0480, 0.0444, 0.0407, 0.0390], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 19:55:38,255 INFO [finetune.py:992] (0/2) Epoch 12, batch 5950, loss[loss=0.1491, simple_loss=0.2452, pruned_loss=0.02645, over 12352.00 frames. ], tot_loss[loss=0.1679, simple_loss=0.2569, pruned_loss=0.03939, over 2375986.37 frames. ], batch size: 35, lr: 3.80e-03, grad_scale: 8.0 2023-05-16 19:56:13,078 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.909e+02 2.622e+02 3.199e+02 3.651e+02 6.632e+02, threshold=6.397e+02, percent-clipped=1.0 2023-05-16 19:56:14,465 INFO [finetune.py:992] (0/2) Epoch 12, batch 6000, loss[loss=0.1677, simple_loss=0.2619, pruned_loss=0.03669, over 11791.00 frames. ], tot_loss[loss=0.1686, simple_loss=0.2571, pruned_loss=0.04006, over 2368647.27 frames. ], batch size: 44, lr: 3.80e-03, grad_scale: 8.0 2023-05-16 19:56:14,466 INFO [finetune.py:1017] (0/2) Computing validation loss 2023-05-16 19:56:32,704 INFO [finetune.py:1026] (0/2) Epoch 12, validation: loss=0.3152, simple_loss=0.3923, pruned_loss=0.1191, over 1020973.00 frames. 2023-05-16 19:56:32,705 INFO [finetune.py:1027] (0/2) Maximum memory allocated so far is 12827MB 2023-05-16 19:56:35,934 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.56 vs. limit=2.0 2023-05-16 19:56:39,931 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.5468, 5.1337, 5.5382, 4.8377, 5.1802, 4.9145, 5.5652, 5.1546], device='cuda:0'), covar=tensor([0.0229, 0.0315, 0.0225, 0.0214, 0.0308, 0.0287, 0.0194, 0.0237], device='cuda:0'), in_proj_covar=tensor([0.0256, 0.0261, 0.0287, 0.0260, 0.0256, 0.0260, 0.0236, 0.0215], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 19:57:09,270 INFO [finetune.py:992] (0/2) Epoch 12, batch 6050, loss[loss=0.1772, simple_loss=0.2703, pruned_loss=0.04205, over 12348.00 frames. ], tot_loss[loss=0.1683, simple_loss=0.2569, pruned_loss=0.03988, over 2373621.73 frames. ], batch size: 35, lr: 3.79e-03, grad_scale: 8.0 2023-05-16 19:57:43,808 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.650e+02 2.690e+02 3.291e+02 4.006e+02 1.196e+03, threshold=6.583e+02, percent-clipped=4.0 2023-05-16 19:57:45,257 INFO [finetune.py:992] (0/2) Epoch 12, batch 6100, loss[loss=0.165, simple_loss=0.2483, pruned_loss=0.04087, over 12336.00 frames. ], tot_loss[loss=0.1695, simple_loss=0.2579, pruned_loss=0.04058, over 2367444.10 frames. ], batch size: 31, lr: 3.79e-03, grad_scale: 8.0 2023-05-16 19:58:01,026 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=240696.0, num_to_drop=1, layers_to_drop={1} 2023-05-16 19:58:08,182 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.6534, 2.6007, 3.4320, 4.5681, 2.2904, 4.5406, 4.6420, 4.8158], device='cuda:0'), covar=tensor([0.0137, 0.1191, 0.0399, 0.0142, 0.1337, 0.0185, 0.0111, 0.0071], device='cuda:0'), in_proj_covar=tensor([0.0126, 0.0204, 0.0183, 0.0121, 0.0191, 0.0181, 0.0178, 0.0120], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 19:58:10,828 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.1348, 5.9838, 5.5831, 5.6239, 6.0892, 5.3047, 5.6096, 5.5181], device='cuda:0'), covar=tensor([0.1487, 0.1063, 0.1067, 0.1821, 0.0963, 0.2198, 0.1797, 0.1164], device='cuda:0'), in_proj_covar=tensor([0.0369, 0.0509, 0.0400, 0.0444, 0.0479, 0.0445, 0.0406, 0.0390], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 19:58:20,712 INFO [finetune.py:992] (0/2) Epoch 12, batch 6150, loss[loss=0.1784, simple_loss=0.2588, pruned_loss=0.04902, over 12333.00 frames. ], tot_loss[loss=0.1701, simple_loss=0.2587, pruned_loss=0.04075, over 2362956.07 frames. ], batch size: 31, lr: 3.79e-03, grad_scale: 8.0 2023-05-16 19:58:32,272 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=4.01 vs. limit=5.0 2023-05-16 19:58:54,921 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.179e+02 2.787e+02 3.331e+02 4.158e+02 7.198e+02, threshold=6.662e+02, percent-clipped=3.0 2023-05-16 19:58:56,393 INFO [finetune.py:992] (0/2) Epoch 12, batch 6200, loss[loss=0.1462, simple_loss=0.2303, pruned_loss=0.03102, over 12265.00 frames. ], tot_loss[loss=0.17, simple_loss=0.2588, pruned_loss=0.04057, over 2366816.24 frames. ], batch size: 32, lr: 3.79e-03, grad_scale: 8.0 2023-05-16 19:59:02,580 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 2023-05-16 19:59:32,518 INFO [finetune.py:992] (0/2) Epoch 12, batch 6250, loss[loss=0.1744, simple_loss=0.2654, pruned_loss=0.04171, over 12163.00 frames. ], tot_loss[loss=0.1695, simple_loss=0.2586, pruned_loss=0.04023, over 2373281.84 frames. ], batch size: 39, lr: 3.79e-03, grad_scale: 8.0 2023-05-16 19:59:53,713 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.41 vs. limit=2.0 2023-05-16 20:00:06,436 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.582e+02 2.812e+02 3.369e+02 3.939e+02 7.796e+02, threshold=6.738e+02, percent-clipped=1.0 2023-05-16 20:00:06,579 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([6.0972, 6.0642, 5.8681, 5.3253, 5.0672, 5.9927, 5.5875, 5.3876], device='cuda:0'), covar=tensor([0.0628, 0.0805, 0.0642, 0.1570, 0.0694, 0.0692, 0.1565, 0.1062], device='cuda:0'), in_proj_covar=tensor([0.0627, 0.0570, 0.0526, 0.0642, 0.0419, 0.0719, 0.0790, 0.0581], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0004, 0.0004, 0.0003], device='cuda:0') 2023-05-16 20:00:07,820 INFO [finetune.py:992] (0/2) Epoch 12, batch 6300, loss[loss=0.157, simple_loss=0.2498, pruned_loss=0.03205, over 12287.00 frames. ], tot_loss[loss=0.1695, simple_loss=0.2584, pruned_loss=0.04028, over 2364533.47 frames. ], batch size: 34, lr: 3.79e-03, grad_scale: 8.0 2023-05-16 20:00:43,809 INFO [finetune.py:992] (0/2) Epoch 12, batch 6350, loss[loss=0.1546, simple_loss=0.2496, pruned_loss=0.02979, over 12111.00 frames. ], tot_loss[loss=0.1691, simple_loss=0.2582, pruned_loss=0.03999, over 2369889.31 frames. ], batch size: 33, lr: 3.79e-03, grad_scale: 8.0 2023-05-16 20:00:50,332 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=240932.0, num_to_drop=1, layers_to_drop={1} 2023-05-16 20:01:07,655 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=3.40 vs. limit=5.0 2023-05-16 20:01:12,485 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.8053, 2.9299, 4.6066, 4.7875, 2.8687, 2.6976, 3.0278, 2.0838], device='cuda:0'), covar=tensor([0.1459, 0.2893, 0.0425, 0.0370, 0.1305, 0.2265, 0.2679, 0.4125], device='cuda:0'), in_proj_covar=tensor([0.0296, 0.0391, 0.0276, 0.0304, 0.0274, 0.0307, 0.0385, 0.0376], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-16 20:01:12,648 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 2023-05-16 20:01:18,684 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.842e+02 2.526e+02 2.951e+02 3.583e+02 3.182e+03, threshold=5.902e+02, percent-clipped=2.0 2023-05-16 20:01:20,171 INFO [finetune.py:992] (0/2) Epoch 12, batch 6400, loss[loss=0.1806, simple_loss=0.278, pruned_loss=0.04158, over 12045.00 frames. ], tot_loss[loss=0.1691, simple_loss=0.2584, pruned_loss=0.03995, over 2371408.00 frames. ], batch size: 40, lr: 3.79e-03, grad_scale: 8.0 2023-05-16 20:01:21,693 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.2500, 5.1455, 5.0158, 5.1503, 4.7775, 5.1679, 5.1861, 5.4390], device='cuda:0'), covar=tensor([0.0185, 0.0135, 0.0161, 0.0293, 0.0668, 0.0261, 0.0123, 0.0132], device='cuda:0'), in_proj_covar=tensor([0.0198, 0.0202, 0.0194, 0.0253, 0.0249, 0.0222, 0.0180, 0.0230], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0005, 0.0004, 0.0005, 0.0004, 0.0005], device='cuda:0') 2023-05-16 20:01:28,815 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=240986.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 20:01:33,827 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=240993.0, num_to_drop=1, layers_to_drop={3} 2023-05-16 20:01:35,779 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=240996.0, num_to_drop=1, layers_to_drop={2} 2023-05-16 20:01:47,607 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.3057, 2.8792, 2.8320, 2.8024, 2.5954, 2.4323, 2.8683, 1.9702], device='cuda:0'), covar=tensor([0.0376, 0.0170, 0.0204, 0.0190, 0.0375, 0.0298, 0.0164, 0.0479], device='cuda:0'), in_proj_covar=tensor([0.0195, 0.0162, 0.0164, 0.0190, 0.0205, 0.0201, 0.0173, 0.0207], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 20:01:56,575 INFO [finetune.py:992] (0/2) Epoch 12, batch 6450, loss[loss=0.2198, simple_loss=0.3062, pruned_loss=0.06665, over 8316.00 frames. ], tot_loss[loss=0.169, simple_loss=0.2582, pruned_loss=0.03988, over 2361800.29 frames. ], batch size: 100, lr: 3.79e-03, grad_scale: 8.0 2023-05-16 20:02:10,808 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=241044.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 20:02:13,096 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=241047.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 20:02:31,312 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.843e+02 2.681e+02 3.309e+02 3.959e+02 7.143e+02, threshold=6.619e+02, percent-clipped=4.0 2023-05-16 20:02:32,734 INFO [finetune.py:992] (0/2) Epoch 12, batch 6500, loss[loss=0.1975, simple_loss=0.2871, pruned_loss=0.05402, over 12011.00 frames. ], tot_loss[loss=0.1693, simple_loss=0.2585, pruned_loss=0.04001, over 2365607.64 frames. ], batch size: 40, lr: 3.79e-03, grad_scale: 8.0 2023-05-16 20:02:52,080 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.1343, 4.7665, 4.9006, 5.0402, 4.8203, 5.0187, 4.9020, 2.9336], device='cuda:0'), covar=tensor([0.0097, 0.0069, 0.0087, 0.0059, 0.0045, 0.0092, 0.0072, 0.0684], device='cuda:0'), in_proj_covar=tensor([0.0069, 0.0080, 0.0082, 0.0074, 0.0060, 0.0093, 0.0082, 0.0098], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-16 20:03:07,228 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.6523, 4.5208, 4.4742, 4.5757, 4.2327, 4.6364, 4.6630, 4.8132], device='cuda:0'), covar=tensor([0.0231, 0.0157, 0.0190, 0.0310, 0.0701, 0.0311, 0.0144, 0.0188], device='cuda:0'), in_proj_covar=tensor([0.0196, 0.0200, 0.0193, 0.0251, 0.0247, 0.0221, 0.0179, 0.0229], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0005, 0.0004, 0.0005, 0.0004, 0.0005], device='cuda:0') 2023-05-16 20:03:08,526 INFO [finetune.py:992] (0/2) Epoch 12, batch 6550, loss[loss=0.1928, simple_loss=0.2777, pruned_loss=0.05392, over 11601.00 frames. ], tot_loss[loss=0.1696, simple_loss=0.2588, pruned_loss=0.04022, over 2353438.13 frames. ], batch size: 48, lr: 3.79e-03, grad_scale: 8.0 2023-05-16 20:03:43,403 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.766e+02 2.717e+02 3.207e+02 3.960e+02 7.122e+02, threshold=6.413e+02, percent-clipped=1.0 2023-05-16 20:03:44,395 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 2023-05-16 20:03:44,732 INFO [finetune.py:992] (0/2) Epoch 12, batch 6600, loss[loss=0.1671, simple_loss=0.2617, pruned_loss=0.03627, over 12194.00 frames. ], tot_loss[loss=0.1703, simple_loss=0.2599, pruned_loss=0.04039, over 2350830.57 frames. ], batch size: 35, lr: 3.79e-03, grad_scale: 8.0 2023-05-16 20:04:07,710 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=4.71 vs. limit=5.0 2023-05-16 20:04:21,374 INFO [finetune.py:992] (0/2) Epoch 12, batch 6650, loss[loss=0.1862, simple_loss=0.2767, pruned_loss=0.04789, over 12280.00 frames. ], tot_loss[loss=0.17, simple_loss=0.2596, pruned_loss=0.04025, over 2363411.73 frames. ], batch size: 37, lr: 3.79e-03, grad_scale: 8.0 2023-05-16 20:04:55,414 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.147e+02 2.914e+02 3.361e+02 3.942e+02 8.822e+02, threshold=6.721e+02, percent-clipped=4.0 2023-05-16 20:04:56,895 INFO [finetune.py:992] (0/2) Epoch 12, batch 6700, loss[loss=0.1593, simple_loss=0.249, pruned_loss=0.03476, over 12277.00 frames. ], tot_loss[loss=0.1702, simple_loss=0.2599, pruned_loss=0.04025, over 2366269.84 frames. ], batch size: 33, lr: 3.79e-03, grad_scale: 16.0 2023-05-16 20:05:07,046 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=241288.0, num_to_drop=1, layers_to_drop={2} 2023-05-16 20:05:33,159 INFO [finetune.py:992] (0/2) Epoch 12, batch 6750, loss[loss=0.1734, simple_loss=0.2607, pruned_loss=0.04305, over 11580.00 frames. ], tot_loss[loss=0.17, simple_loss=0.2596, pruned_loss=0.04021, over 2368108.14 frames. ], batch size: 48, lr: 3.79e-03, grad_scale: 16.0 2023-05-16 20:05:45,912 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=241342.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 20:06:02,715 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.7851, 5.8119, 5.5586, 5.0189, 4.9700, 5.7293, 5.3060, 5.1292], device='cuda:0'), covar=tensor([0.0711, 0.0807, 0.0667, 0.1677, 0.0767, 0.0736, 0.1591, 0.1089], device='cuda:0'), in_proj_covar=tensor([0.0625, 0.0564, 0.0521, 0.0637, 0.0416, 0.0714, 0.0779, 0.0576], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-05-16 20:06:08,330 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.835e+02 2.614e+02 3.057e+02 3.687e+02 7.195e+02, threshold=6.114e+02, percent-clipped=1.0 2023-05-16 20:06:09,763 INFO [finetune.py:992] (0/2) Epoch 12, batch 6800, loss[loss=0.1748, simple_loss=0.2641, pruned_loss=0.04276, over 12011.00 frames. ], tot_loss[loss=0.1696, simple_loss=0.2592, pruned_loss=0.03995, over 2362044.02 frames. ], batch size: 40, lr: 3.79e-03, grad_scale: 16.0 2023-05-16 20:06:33,118 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=241406.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 20:06:45,789 INFO [finetune.py:992] (0/2) Epoch 12, batch 6850, loss[loss=0.1623, simple_loss=0.2452, pruned_loss=0.03965, over 12132.00 frames. ], tot_loss[loss=0.1692, simple_loss=0.2588, pruned_loss=0.03981, over 2373012.87 frames. ], batch size: 30, lr: 3.79e-03, grad_scale: 16.0 2023-05-16 20:06:48,899 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.3198, 2.5988, 3.1740, 4.2546, 2.2082, 4.1830, 4.1864, 4.4164], device='cuda:0'), covar=tensor([0.0132, 0.1104, 0.0461, 0.0138, 0.1282, 0.0309, 0.0214, 0.0084], device='cuda:0'), in_proj_covar=tensor([0.0126, 0.0203, 0.0184, 0.0120, 0.0190, 0.0181, 0.0179, 0.0120], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 20:06:50,411 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=3.06 vs. limit=5.0 2023-05-16 20:07:17,284 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=241467.0, num_to_drop=1, layers_to_drop={3} 2023-05-16 20:07:20,530 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.727e+02 2.775e+02 3.412e+02 3.852e+02 9.667e+02, threshold=6.824e+02, percent-clipped=3.0 2023-05-16 20:07:21,936 INFO [finetune.py:992] (0/2) Epoch 12, batch 6900, loss[loss=0.1728, simple_loss=0.2605, pruned_loss=0.04259, over 12134.00 frames. ], tot_loss[loss=0.1693, simple_loss=0.2586, pruned_loss=0.03997, over 2368027.02 frames. ], batch size: 30, lr: 3.79e-03, grad_scale: 16.0 2023-05-16 20:07:51,782 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.5612, 3.6779, 3.2819, 3.2449, 3.0545, 2.9406, 3.6883, 2.3660], device='cuda:0'), covar=tensor([0.0392, 0.0132, 0.0176, 0.0194, 0.0334, 0.0314, 0.0113, 0.0470], device='cuda:0'), in_proj_covar=tensor([0.0194, 0.0161, 0.0163, 0.0190, 0.0202, 0.0202, 0.0173, 0.0204], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 20:07:57,834 INFO [finetune.py:992] (0/2) Epoch 12, batch 6950, loss[loss=0.1632, simple_loss=0.2536, pruned_loss=0.03639, over 12107.00 frames. ], tot_loss[loss=0.1691, simple_loss=0.2584, pruned_loss=0.03992, over 2361062.24 frames. ], batch size: 33, lr: 3.79e-03, grad_scale: 16.0 2023-05-16 20:08:11,821 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=4.35 vs. limit=5.0 2023-05-16 20:08:31,997 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.054e+02 2.695e+02 3.162e+02 3.755e+02 7.672e+02, threshold=6.324e+02, percent-clipped=2.0 2023-05-16 20:08:33,379 INFO [finetune.py:992] (0/2) Epoch 12, batch 7000, loss[loss=0.1802, simple_loss=0.27, pruned_loss=0.04518, over 12354.00 frames. ], tot_loss[loss=0.1692, simple_loss=0.2585, pruned_loss=0.03991, over 2365643.20 frames. ], batch size: 35, lr: 3.79e-03, grad_scale: 16.0 2023-05-16 20:08:43,575 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=241588.0, num_to_drop=1, layers_to_drop={1} 2023-05-16 20:09:08,277 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=241621.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 20:09:10,212 INFO [finetune.py:992] (0/2) Epoch 12, batch 7050, loss[loss=0.1697, simple_loss=0.2573, pruned_loss=0.0411, over 12189.00 frames. ], tot_loss[loss=0.1687, simple_loss=0.258, pruned_loss=0.0397, over 2359130.97 frames. ], batch size: 35, lr: 3.79e-03, grad_scale: 16.0 2023-05-16 20:09:18,831 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=241636.0, num_to_drop=1, layers_to_drop={1} 2023-05-16 20:09:22,982 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=241642.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 20:09:45,058 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.927e+02 2.716e+02 3.212e+02 3.977e+02 7.981e+02, threshold=6.423e+02, percent-clipped=1.0 2023-05-16 20:09:46,257 INFO [finetune.py:992] (0/2) Epoch 12, batch 7100, loss[loss=0.1723, simple_loss=0.252, pruned_loss=0.04628, over 12077.00 frames. ], tot_loss[loss=0.1689, simple_loss=0.2583, pruned_loss=0.03977, over 2367538.85 frames. ], batch size: 32, lr: 3.79e-03, grad_scale: 16.0 2023-05-16 20:09:52,231 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=241682.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 20:09:54,493 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=2.02 vs. limit=2.0 2023-05-16 20:09:57,650 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=241690.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 20:10:21,792 INFO [finetune.py:992] (0/2) Epoch 12, batch 7150, loss[loss=0.1617, simple_loss=0.2582, pruned_loss=0.03263, over 12014.00 frames. ], tot_loss[loss=0.1682, simple_loss=0.2572, pruned_loss=0.03956, over 2368159.90 frames. ], batch size: 42, lr: 3.79e-03, grad_scale: 16.0 2023-05-16 20:10:49,600 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=241762.0, num_to_drop=1, layers_to_drop={3} 2023-05-16 20:10:56,605 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.932e+02 2.664e+02 3.124e+02 3.838e+02 7.600e+02, threshold=6.248e+02, percent-clipped=2.0 2023-05-16 20:10:57,991 INFO [finetune.py:992] (0/2) Epoch 12, batch 7200, loss[loss=0.1904, simple_loss=0.2762, pruned_loss=0.05234, over 12378.00 frames. ], tot_loss[loss=0.1673, simple_loss=0.2564, pruned_loss=0.03915, over 2382188.88 frames. ], batch size: 38, lr: 3.78e-03, grad_scale: 16.0 2023-05-16 20:11:10,556 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=2.48 vs. limit=5.0 2023-05-16 20:11:20,016 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=241803.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 20:11:34,461 INFO [finetune.py:992] (0/2) Epoch 12, batch 7250, loss[loss=0.1479, simple_loss=0.2303, pruned_loss=0.03277, over 12348.00 frames. ], tot_loss[loss=0.1671, simple_loss=0.2559, pruned_loss=0.03915, over 2382308.38 frames. ], batch size: 31, lr: 3.78e-03, grad_scale: 16.0 2023-05-16 20:11:39,590 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=241831.0, num_to_drop=1, layers_to_drop={0} 2023-05-16 20:11:45,390 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.13 vs. limit=2.0 2023-05-16 20:11:48,170 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.6030, 2.9001, 4.4723, 4.6114, 2.9477, 2.5976, 2.9747, 2.1267], device='cuda:0'), covar=tensor([0.1573, 0.2719, 0.0459, 0.0438, 0.1208, 0.2324, 0.2486, 0.3903], device='cuda:0'), in_proj_covar=tensor([0.0296, 0.0387, 0.0275, 0.0303, 0.0272, 0.0306, 0.0383, 0.0373], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-16 20:12:02,897 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=241864.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 20:12:08,390 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.908e+02 2.596e+02 3.129e+02 3.845e+02 1.160e+03, threshold=6.258e+02, percent-clipped=4.0 2023-05-16 20:12:09,826 INFO [finetune.py:992] (0/2) Epoch 12, batch 7300, loss[loss=0.1743, simple_loss=0.2732, pruned_loss=0.0377, over 11824.00 frames. ], tot_loss[loss=0.1672, simple_loss=0.2561, pruned_loss=0.03911, over 2383717.94 frames. ], batch size: 44, lr: 3.78e-03, grad_scale: 16.0 2023-05-16 20:12:18,617 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.3584, 1.9610, 3.8226, 4.2086, 3.7977, 4.1064, 3.8530, 2.9067], device='cuda:0'), covar=tensor([0.0051, 0.0648, 0.0125, 0.0058, 0.0144, 0.0102, 0.0146, 0.0421], device='cuda:0'), in_proj_covar=tensor([0.0089, 0.0125, 0.0107, 0.0080, 0.0104, 0.0117, 0.0099, 0.0140], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 20:12:23,580 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=241892.0, num_to_drop=1, layers_to_drop={0} 2023-05-16 20:12:46,194 INFO [finetune.py:992] (0/2) Epoch 12, batch 7350, loss[loss=0.1464, simple_loss=0.2305, pruned_loss=0.03115, over 12182.00 frames. ], tot_loss[loss=0.167, simple_loss=0.2559, pruned_loss=0.03911, over 2379377.89 frames. ], batch size: 31, lr: 3.78e-03, grad_scale: 16.0 2023-05-16 20:12:55,776 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.55 vs. limit=2.0 2023-05-16 20:13:20,760 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.762e+02 2.598e+02 2.904e+02 3.475e+02 6.796e+02, threshold=5.808e+02, percent-clipped=1.0 2023-05-16 20:13:22,266 INFO [finetune.py:992] (0/2) Epoch 12, batch 7400, loss[loss=0.1844, simple_loss=0.2722, pruned_loss=0.04835, over 12299.00 frames. ], tot_loss[loss=0.1671, simple_loss=0.2562, pruned_loss=0.03903, over 2378220.75 frames. ], batch size: 34, lr: 3.78e-03, grad_scale: 16.0 2023-05-16 20:13:24,545 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=241977.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 20:13:28,923 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.5294, 2.3856, 3.1803, 4.4132, 2.4734, 4.3693, 4.3764, 4.5971], device='cuda:0'), covar=tensor([0.0145, 0.1336, 0.0517, 0.0135, 0.1268, 0.0299, 0.0195, 0.0083], device='cuda:0'), in_proj_covar=tensor([0.0128, 0.0205, 0.0186, 0.0120, 0.0192, 0.0183, 0.0180, 0.0121], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 20:13:40,918 INFO [checkpoint.py:75] (0/2) Saving checkpoint to pruned_transducer_stateless7/exp_giga_finetune/checkpoint-142000.pt 2023-05-16 20:13:46,787 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.0728, 4.9436, 4.7971, 4.9392, 4.5932, 4.9902, 5.0721, 5.2401], device='cuda:0'), covar=tensor([0.0240, 0.0171, 0.0245, 0.0355, 0.0746, 0.0337, 0.0157, 0.0183], device='cuda:0'), in_proj_covar=tensor([0.0195, 0.0199, 0.0192, 0.0250, 0.0245, 0.0220, 0.0178, 0.0229], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0005, 0.0004, 0.0004, 0.0003, 0.0005], device='cuda:0') 2023-05-16 20:14:01,355 INFO [finetune.py:992] (0/2) Epoch 12, batch 7450, loss[loss=0.1754, simple_loss=0.2677, pruned_loss=0.04158, over 12353.00 frames. ], tot_loss[loss=0.1681, simple_loss=0.2573, pruned_loss=0.03947, over 2378309.91 frames. ], batch size: 36, lr: 3.78e-03, grad_scale: 16.0 2023-05-16 20:14:18,586 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.1542, 2.1152, 2.9027, 2.9647, 3.0312, 3.1400, 2.8873, 2.4184], device='cuda:0'), covar=tensor([0.0084, 0.0441, 0.0201, 0.0098, 0.0142, 0.0110, 0.0175, 0.0377], device='cuda:0'), in_proj_covar=tensor([0.0089, 0.0126, 0.0107, 0.0080, 0.0104, 0.0118, 0.0099, 0.0140], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 20:14:21,020 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-05-16 20:14:28,400 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=242062.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 20:14:35,408 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.965e+02 2.768e+02 3.267e+02 3.959e+02 8.297e+02, threshold=6.535e+02, percent-clipped=4.0 2023-05-16 20:14:36,873 INFO [finetune.py:992] (0/2) Epoch 12, batch 7500, loss[loss=0.158, simple_loss=0.2475, pruned_loss=0.03425, over 12070.00 frames. ], tot_loss[loss=0.169, simple_loss=0.2581, pruned_loss=0.03995, over 2379257.89 frames. ], batch size: 32, lr: 3.78e-03, grad_scale: 16.0 2023-05-16 20:14:41,993 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.4886, 3.5959, 3.2340, 3.1400, 2.7487, 2.5792, 3.5983, 2.1731], device='cuda:0'), covar=tensor([0.0423, 0.0174, 0.0213, 0.0235, 0.0490, 0.0454, 0.0129, 0.0575], device='cuda:0'), in_proj_covar=tensor([0.0196, 0.0163, 0.0164, 0.0192, 0.0205, 0.0204, 0.0175, 0.0208], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 20:15:03,154 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=242110.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 20:15:12,400 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.2204, 2.6209, 3.7325, 3.0312, 3.3867, 3.1592, 2.6678, 3.4950], device='cuda:0'), covar=tensor([0.0136, 0.0356, 0.0131, 0.0253, 0.0178, 0.0207, 0.0356, 0.0180], device='cuda:0'), in_proj_covar=tensor([0.0180, 0.0207, 0.0190, 0.0190, 0.0219, 0.0166, 0.0198, 0.0191], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 20:15:12,885 INFO [finetune.py:992] (0/2) Epoch 12, batch 7550, loss[loss=0.163, simple_loss=0.2553, pruned_loss=0.03531, over 12248.00 frames. ], tot_loss[loss=0.1695, simple_loss=0.2585, pruned_loss=0.04024, over 2372334.41 frames. ], batch size: 32, lr: 3.78e-03, grad_scale: 16.0 2023-05-16 20:15:33,476 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=242153.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 20:15:37,691 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=242159.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 20:15:37,772 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.3031, 4.8189, 5.2946, 4.5714, 4.9261, 4.6824, 5.2964, 4.9431], device='cuda:0'), covar=tensor([0.0285, 0.0441, 0.0292, 0.0305, 0.0417, 0.0363, 0.0228, 0.0322], device='cuda:0'), in_proj_covar=tensor([0.0260, 0.0264, 0.0289, 0.0264, 0.0261, 0.0264, 0.0239, 0.0217], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 20:15:47,474 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.046e+02 2.622e+02 3.220e+02 3.769e+02 1.536e+03, threshold=6.441e+02, percent-clipped=4.0 2023-05-16 20:15:48,918 INFO [finetune.py:992] (0/2) Epoch 12, batch 7600, loss[loss=0.1525, simple_loss=0.2446, pruned_loss=0.03021, over 12033.00 frames. ], tot_loss[loss=0.169, simple_loss=0.2578, pruned_loss=0.04011, over 2378089.55 frames. ], batch size: 40, lr: 3.78e-03, grad_scale: 16.0 2023-05-16 20:15:53,738 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.38 vs. limit=2.0 2023-05-16 20:15:58,281 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=242187.0, num_to_drop=1, layers_to_drop={3} 2023-05-16 20:16:15,281 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.3471, 4.2383, 4.2448, 4.6676, 3.0859, 3.9109, 2.7759, 4.2861], device='cuda:0'), covar=tensor([0.1463, 0.0603, 0.0844, 0.0681, 0.1102, 0.0610, 0.1673, 0.1234], device='cuda:0'), in_proj_covar=tensor([0.0229, 0.0265, 0.0297, 0.0358, 0.0240, 0.0243, 0.0260, 0.0379], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 20:16:18,243 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=242214.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 20:16:25,315 INFO [finetune.py:992] (0/2) Epoch 12, batch 7650, loss[loss=0.1502, simple_loss=0.231, pruned_loss=0.03475, over 12353.00 frames. ], tot_loss[loss=0.168, simple_loss=0.2567, pruned_loss=0.03964, over 2377555.85 frames. ], batch size: 30, lr: 3.78e-03, grad_scale: 16.0 2023-05-16 20:16:26,919 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=242226.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 20:16:35,083 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.87 vs. limit=2.0 2023-05-16 20:16:51,141 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.7807, 4.4978, 4.6438, 4.8032, 4.6314, 4.7881, 4.7275, 2.3395], device='cuda:0'), covar=tensor([0.0133, 0.0081, 0.0096, 0.0060, 0.0056, 0.0096, 0.0082, 0.0963], device='cuda:0'), in_proj_covar=tensor([0.0070, 0.0079, 0.0082, 0.0074, 0.0060, 0.0093, 0.0082, 0.0097], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-16 20:17:00,128 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.708e+02 2.730e+02 3.254e+02 3.873e+02 7.340e+02, threshold=6.508e+02, percent-clipped=1.0 2023-05-16 20:17:01,547 INFO [finetune.py:992] (0/2) Epoch 12, batch 7700, loss[loss=0.1723, simple_loss=0.2633, pruned_loss=0.04061, over 12342.00 frames. ], tot_loss[loss=0.1686, simple_loss=0.2575, pruned_loss=0.03986, over 2382580.35 frames. ], batch size: 36, lr: 3.78e-03, grad_scale: 16.0 2023-05-16 20:17:03,750 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=242277.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 20:17:10,144 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([6.0346, 6.0287, 5.7770, 5.2078, 5.1829, 5.9139, 5.5192, 5.3544], device='cuda:0'), covar=tensor([0.0737, 0.0906, 0.0645, 0.1646, 0.0638, 0.0716, 0.1568, 0.1032], device='cuda:0'), in_proj_covar=tensor([0.0618, 0.0555, 0.0514, 0.0632, 0.0410, 0.0705, 0.0775, 0.0566], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-05-16 20:17:10,991 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=242287.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 20:17:26,817 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=242309.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 20:17:32,315 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.4464, 6.0868, 5.6399, 5.6901, 6.1769, 5.5920, 5.6735, 5.6901], device='cuda:0'), covar=tensor([0.1183, 0.0855, 0.1141, 0.1534, 0.0906, 0.1784, 0.1493, 0.1039], device='cuda:0'), in_proj_covar=tensor([0.0353, 0.0489, 0.0390, 0.0434, 0.0464, 0.0430, 0.0390, 0.0380], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 20:17:37,999 INFO [finetune.py:992] (0/2) Epoch 12, batch 7750, loss[loss=0.1844, simple_loss=0.2775, pruned_loss=0.04562, over 12054.00 frames. ], tot_loss[loss=0.1689, simple_loss=0.2577, pruned_loss=0.04002, over 2378242.70 frames. ], batch size: 42, lr: 3.78e-03, grad_scale: 16.0 2023-05-16 20:17:38,755 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=242325.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 20:18:10,895 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=242370.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 20:18:12,116 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.835e+02 2.829e+02 3.213e+02 3.821e+02 6.496e+02, threshold=6.427e+02, percent-clipped=0.0 2023-05-16 20:18:13,535 INFO [finetune.py:992] (0/2) Epoch 12, batch 7800, loss[loss=0.1659, simple_loss=0.2636, pruned_loss=0.03408, over 12269.00 frames. ], tot_loss[loss=0.1698, simple_loss=0.2587, pruned_loss=0.04044, over 2375882.51 frames. ], batch size: 37, lr: 3.78e-03, grad_scale: 16.0 2023-05-16 20:18:41,901 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.0825, 4.7236, 4.8083, 4.9692, 4.7919, 4.9989, 4.9919, 2.5812], device='cuda:0'), covar=tensor([0.0085, 0.0070, 0.0091, 0.0061, 0.0045, 0.0089, 0.0059, 0.0833], device='cuda:0'), in_proj_covar=tensor([0.0069, 0.0079, 0.0081, 0.0074, 0.0060, 0.0093, 0.0081, 0.0097], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-16 20:18:49,451 INFO [finetune.py:992] (0/2) Epoch 12, batch 7850, loss[loss=0.1501, simple_loss=0.242, pruned_loss=0.02903, over 12344.00 frames. ], tot_loss[loss=0.1694, simple_loss=0.258, pruned_loss=0.04038, over 2379276.33 frames. ], batch size: 30, lr: 3.78e-03, grad_scale: 16.0 2023-05-16 20:19:13,631 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.7710, 2.8902, 3.8043, 4.5608, 4.0892, 4.6618, 3.9210, 3.3137], device='cuda:0'), covar=tensor([0.0032, 0.0372, 0.0134, 0.0051, 0.0100, 0.0066, 0.0134, 0.0341], device='cuda:0'), in_proj_covar=tensor([0.0088, 0.0124, 0.0106, 0.0080, 0.0103, 0.0117, 0.0098, 0.0139], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 20:19:14,321 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=242459.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 20:19:24,104 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.645e+02 2.795e+02 3.203e+02 3.816e+02 9.207e+02, threshold=6.406e+02, percent-clipped=4.0 2023-05-16 20:19:25,612 INFO [finetune.py:992] (0/2) Epoch 12, batch 7900, loss[loss=0.1796, simple_loss=0.2747, pruned_loss=0.04221, over 11418.00 frames. ], tot_loss[loss=0.1702, simple_loss=0.2588, pruned_loss=0.04077, over 2374840.70 frames. ], batch size: 55, lr: 3.78e-03, grad_scale: 16.0 2023-05-16 20:19:28,437 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.9337, 3.7672, 3.9319, 3.6044, 3.8013, 3.6234, 3.9485, 3.5587], device='cuda:0'), covar=tensor([0.0376, 0.0399, 0.0327, 0.0272, 0.0392, 0.0356, 0.0314, 0.1299], device='cuda:0'), in_proj_covar=tensor([0.0261, 0.0264, 0.0289, 0.0264, 0.0263, 0.0264, 0.0239, 0.0217], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 20:19:32,076 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=242483.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 20:19:34,956 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=242487.0, num_to_drop=1, layers_to_drop={0} 2023-05-16 20:19:39,795 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.9595, 4.5632, 4.5309, 4.8485, 4.6365, 4.8876, 4.6952, 2.0272], device='cuda:0'), covar=tensor([0.0124, 0.0094, 0.0153, 0.0083, 0.0065, 0.0119, 0.0099, 0.1138], device='cuda:0'), in_proj_covar=tensor([0.0069, 0.0079, 0.0081, 0.0073, 0.0060, 0.0092, 0.0081, 0.0097], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-16 20:19:48,941 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=242507.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 20:19:50,397 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=242509.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 20:20:01,097 INFO [finetune.py:992] (0/2) Epoch 12, batch 7950, loss[loss=0.1657, simple_loss=0.2572, pruned_loss=0.03709, over 12106.00 frames. ], tot_loss[loss=0.1697, simple_loss=0.2584, pruned_loss=0.04046, over 2381212.24 frames. ], batch size: 33, lr: 3.78e-03, grad_scale: 16.0 2023-05-16 20:20:02,711 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.2666, 5.2487, 5.0589, 4.6392, 4.6799, 5.2040, 4.8376, 4.6879], device='cuda:0'), covar=tensor([0.0792, 0.0901, 0.0688, 0.1560, 0.1221, 0.0738, 0.1529, 0.1046], device='cuda:0'), in_proj_covar=tensor([0.0619, 0.0558, 0.0515, 0.0633, 0.0412, 0.0710, 0.0777, 0.0570], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-05-16 20:20:08,970 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=242535.0, num_to_drop=1, layers_to_drop={0} 2023-05-16 20:20:16,153 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=242544.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 20:20:23,205 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.3548, 4.8770, 5.3283, 4.6327, 4.9661, 4.7324, 5.3433, 5.0875], device='cuda:0'), covar=tensor([0.0261, 0.0434, 0.0280, 0.0277, 0.0419, 0.0349, 0.0233, 0.0267], device='cuda:0'), in_proj_covar=tensor([0.0260, 0.0264, 0.0289, 0.0264, 0.0263, 0.0263, 0.0239, 0.0217], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 20:20:23,954 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=242555.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 20:20:35,875 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.882e+02 2.774e+02 3.170e+02 3.692e+02 7.403e+02, threshold=6.340e+02, percent-clipped=1.0 2023-05-16 20:20:37,245 INFO [finetune.py:992] (0/2) Epoch 12, batch 8000, loss[loss=0.1498, simple_loss=0.2437, pruned_loss=0.02792, over 12187.00 frames. ], tot_loss[loss=0.1691, simple_loss=0.258, pruned_loss=0.04012, over 2384605.35 frames. ], batch size: 31, lr: 3.78e-03, grad_scale: 16.0 2023-05-16 20:20:41,833 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=2.00 vs. limit=2.0 2023-05-16 20:20:42,875 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=242582.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 20:20:57,519 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=242602.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 20:20:59,592 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.7529, 3.0265, 3.5144, 4.6916, 2.8295, 4.5916, 4.6980, 4.8421], device='cuda:0'), covar=tensor([0.0113, 0.1053, 0.0422, 0.0129, 0.1068, 0.0227, 0.0113, 0.0083], device='cuda:0'), in_proj_covar=tensor([0.0124, 0.0201, 0.0183, 0.0119, 0.0188, 0.0180, 0.0177, 0.0119], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 20:21:08,070 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=242616.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 20:21:13,418 INFO [finetune.py:992] (0/2) Epoch 12, batch 8050, loss[loss=0.1724, simple_loss=0.2617, pruned_loss=0.04151, over 12249.00 frames. ], tot_loss[loss=0.168, simple_loss=0.2567, pruned_loss=0.0397, over 2391217.85 frames. ], batch size: 32, lr: 3.78e-03, grad_scale: 16.0 2023-05-16 20:21:19,421 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.5266, 2.4789, 3.7058, 4.3888, 3.9155, 4.4265, 3.7721, 2.9773], device='cuda:0'), covar=tensor([0.0041, 0.0413, 0.0140, 0.0053, 0.0117, 0.0080, 0.0139, 0.0397], device='cuda:0'), in_proj_covar=tensor([0.0088, 0.0124, 0.0105, 0.0079, 0.0103, 0.0116, 0.0097, 0.0138], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 20:21:25,606 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.2662, 6.1514, 5.6313, 5.6550, 6.2082, 5.5421, 5.7400, 5.6196], device='cuda:0'), covar=tensor([0.1388, 0.0971, 0.1106, 0.2221, 0.0867, 0.1967, 0.1722, 0.1098], device='cuda:0'), in_proj_covar=tensor([0.0358, 0.0496, 0.0396, 0.0442, 0.0469, 0.0436, 0.0395, 0.0386], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 20:21:39,163 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.5597, 3.5589, 3.3824, 3.2299, 2.9065, 2.7848, 3.6203, 2.2354], device='cuda:0'), covar=tensor([0.0376, 0.0164, 0.0156, 0.0174, 0.0391, 0.0345, 0.0117, 0.0499], device='cuda:0'), in_proj_covar=tensor([0.0193, 0.0161, 0.0160, 0.0189, 0.0202, 0.0199, 0.0173, 0.0203], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 20:21:41,368 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=242663.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 20:21:42,705 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=242665.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 20:21:47,628 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.120e+02 2.732e+02 3.297e+02 3.901e+02 7.745e+02, threshold=6.593e+02, percent-clipped=1.0 2023-05-16 20:21:48,985 INFO [finetune.py:992] (0/2) Epoch 12, batch 8100, loss[loss=0.1531, simple_loss=0.2461, pruned_loss=0.03008, over 12363.00 frames. ], tot_loss[loss=0.1684, simple_loss=0.2572, pruned_loss=0.03977, over 2385634.01 frames. ], batch size: 35, lr: 3.78e-03, grad_scale: 16.0 2023-05-16 20:21:54,799 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.0445, 4.5120, 4.0975, 4.8774, 4.5109, 2.8029, 4.1674, 3.0964], device='cuda:0'), covar=tensor([0.1016, 0.0816, 0.1287, 0.0534, 0.1132, 0.1651, 0.1019, 0.3011], device='cuda:0'), in_proj_covar=tensor([0.0302, 0.0377, 0.0352, 0.0311, 0.0365, 0.0268, 0.0341, 0.0365], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 20:22:02,057 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.93 vs. limit=2.0 2023-05-16 20:22:15,295 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.5643, 5.3640, 5.5369, 5.5157, 5.1392, 5.1724, 4.9312, 5.4366], device='cuda:0'), covar=tensor([0.0660, 0.0586, 0.0743, 0.0585, 0.1820, 0.1366, 0.0523, 0.1116], device='cuda:0'), in_proj_covar=tensor([0.0541, 0.0713, 0.0617, 0.0633, 0.0853, 0.0750, 0.0552, 0.0485], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0005, 0.0004, 0.0004, 0.0005, 0.0005, 0.0003, 0.0003], device='cuda:0') 2023-05-16 20:22:24,901 INFO [finetune.py:992] (0/2) Epoch 12, batch 8150, loss[loss=0.2603, simple_loss=0.3199, pruned_loss=0.1003, over 7697.00 frames. ], tot_loss[loss=0.1703, simple_loss=0.2589, pruned_loss=0.04085, over 2374182.98 frames. ], batch size: 98, lr: 3.78e-03, grad_scale: 16.0 2023-05-16 20:22:29,694 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.58 vs. limit=2.0 2023-05-16 20:22:59,371 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.881e+02 2.796e+02 3.192e+02 4.090e+02 9.009e+02, threshold=6.384e+02, percent-clipped=4.0 2023-05-16 20:22:59,545 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=242772.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 20:23:00,819 INFO [finetune.py:992] (0/2) Epoch 12, batch 8200, loss[loss=0.1626, simple_loss=0.2561, pruned_loss=0.03459, over 12189.00 frames. ], tot_loss[loss=0.1705, simple_loss=0.2589, pruned_loss=0.04102, over 2368256.40 frames. ], batch size: 31, lr: 3.78e-03, grad_scale: 16.0 2023-05-16 20:23:26,356 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=242809.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 20:23:37,163 INFO [finetune.py:992] (0/2) Epoch 12, batch 8250, loss[loss=0.2725, simple_loss=0.3286, pruned_loss=0.1082, over 7995.00 frames. ], tot_loss[loss=0.1707, simple_loss=0.2591, pruned_loss=0.04116, over 2361852.75 frames. ], batch size: 98, lr: 3.78e-03, grad_scale: 16.0 2023-05-16 20:23:44,349 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=242833.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 20:23:48,376 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=242839.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 20:23:49,944 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.3885, 3.7398, 3.4077, 3.3080, 2.9688, 2.8819, 3.7180, 2.1805], device='cuda:0'), covar=tensor([0.0470, 0.0132, 0.0196, 0.0211, 0.0358, 0.0336, 0.0127, 0.0538], device='cuda:0'), in_proj_covar=tensor([0.0194, 0.0161, 0.0161, 0.0190, 0.0203, 0.0200, 0.0173, 0.0204], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 20:24:01,108 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=242857.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 20:24:11,672 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.883e+02 2.670e+02 3.128e+02 3.945e+02 6.328e+02, threshold=6.256e+02, percent-clipped=0.0 2023-05-16 20:24:13,089 INFO [finetune.py:992] (0/2) Epoch 12, batch 8300, loss[loss=0.1433, simple_loss=0.2235, pruned_loss=0.03156, over 12369.00 frames. ], tot_loss[loss=0.1704, simple_loss=0.2587, pruned_loss=0.04106, over 2361624.97 frames. ], batch size: 30, lr: 3.77e-03, grad_scale: 16.0 2023-05-16 20:24:19,039 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=242882.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 20:24:23,610 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.58 vs. limit=2.0 2023-05-16 20:24:25,303 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.9800, 4.8530, 4.7469, 4.8095, 4.4854, 4.9991, 4.9208, 5.1747], device='cuda:0'), covar=tensor([0.0253, 0.0163, 0.0201, 0.0358, 0.0738, 0.0236, 0.0162, 0.0163], device='cuda:0'), in_proj_covar=tensor([0.0196, 0.0200, 0.0193, 0.0253, 0.0246, 0.0221, 0.0180, 0.0231], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0005, 0.0004, 0.0005, 0.0004, 0.0005], device='cuda:0') 2023-05-16 20:24:39,953 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=242911.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 20:24:40,067 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.2117, 2.2168, 2.6153, 3.2054, 2.1492, 3.2370, 3.1543, 3.3776], device='cuda:0'), covar=tensor([0.0171, 0.1032, 0.0541, 0.0179, 0.1069, 0.0384, 0.0323, 0.0120], device='cuda:0'), in_proj_covar=tensor([0.0126, 0.0203, 0.0185, 0.0119, 0.0190, 0.0182, 0.0179, 0.0120], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 20:24:47,270 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=242921.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 20:24:49,296 INFO [finetune.py:992] (0/2) Epoch 12, batch 8350, loss[loss=0.1427, simple_loss=0.2263, pruned_loss=0.0296, over 12112.00 frames. ], tot_loss[loss=0.169, simple_loss=0.2575, pruned_loss=0.04025, over 2365779.11 frames. ], batch size: 30, lr: 3.77e-03, grad_scale: 16.0 2023-05-16 20:24:53,478 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=242930.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 20:25:05,695 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.8495, 3.4970, 5.2580, 2.6769, 2.7497, 3.9689, 3.2623, 3.9996], device='cuda:0'), covar=tensor([0.0467, 0.1163, 0.0268, 0.1176, 0.2143, 0.1453, 0.1363, 0.1239], device='cuda:0'), in_proj_covar=tensor([0.0236, 0.0234, 0.0249, 0.0182, 0.0237, 0.0294, 0.0221, 0.0262], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-16 20:25:13,306 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=242958.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 20:25:18,237 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=242965.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 20:25:23,170 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.692e+02 2.758e+02 3.242e+02 3.889e+02 6.335e+02, threshold=6.485e+02, percent-clipped=1.0 2023-05-16 20:25:24,685 INFO [finetune.py:992] (0/2) Epoch 12, batch 8400, loss[loss=0.186, simple_loss=0.2857, pruned_loss=0.0432, over 11718.00 frames. ], tot_loss[loss=0.1696, simple_loss=0.2581, pruned_loss=0.04052, over 2362380.76 frames. ], batch size: 48, lr: 3.77e-03, grad_scale: 16.0 2023-05-16 20:25:31,255 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=242982.0, num_to_drop=1, layers_to_drop={3} 2023-05-16 20:25:53,262 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=243013.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 20:26:00,963 INFO [finetune.py:992] (0/2) Epoch 12, batch 8450, loss[loss=0.1777, simple_loss=0.2685, pruned_loss=0.04348, over 12389.00 frames. ], tot_loss[loss=0.1691, simple_loss=0.2577, pruned_loss=0.04027, over 2367758.36 frames. ], batch size: 38, lr: 3.77e-03, grad_scale: 16.0 2023-05-16 20:26:35,247 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.946e+02 2.724e+02 3.204e+02 3.577e+02 5.925e+02, threshold=6.408e+02, percent-clipped=0.0 2023-05-16 20:26:36,751 INFO [finetune.py:992] (0/2) Epoch 12, batch 8500, loss[loss=0.1557, simple_loss=0.2526, pruned_loss=0.02936, over 12368.00 frames. ], tot_loss[loss=0.1685, simple_loss=0.257, pruned_loss=0.04002, over 2375668.69 frames. ], batch size: 36, lr: 3.77e-03, grad_scale: 16.0 2023-05-16 20:26:40,632 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.60 vs. limit=2.0 2023-05-16 20:26:57,509 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.0440, 2.2253, 3.5750, 2.9342, 3.4274, 3.0640, 2.4016, 3.4420], device='cuda:0'), covar=tensor([0.0153, 0.0485, 0.0187, 0.0297, 0.0181, 0.0221, 0.0418, 0.0150], device='cuda:0'), in_proj_covar=tensor([0.0180, 0.0207, 0.0190, 0.0190, 0.0220, 0.0166, 0.0198, 0.0191], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 20:27:10,040 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.3855, 5.1909, 5.3126, 5.3447, 4.9480, 5.0303, 4.7690, 5.3017], device='cuda:0'), covar=tensor([0.0658, 0.0578, 0.0782, 0.0550, 0.1851, 0.1218, 0.0523, 0.0991], device='cuda:0'), in_proj_covar=tensor([0.0543, 0.0709, 0.0620, 0.0633, 0.0857, 0.0752, 0.0555, 0.0487], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0005, 0.0004, 0.0004, 0.0005, 0.0005, 0.0003, 0.0003], device='cuda:0') 2023-05-16 20:27:12,123 INFO [finetune.py:992] (0/2) Epoch 12, batch 8550, loss[loss=0.148, simple_loss=0.2346, pruned_loss=0.03073, over 12182.00 frames. ], tot_loss[loss=0.169, simple_loss=0.2573, pruned_loss=0.04032, over 2368894.78 frames. ], batch size: 29, lr: 3.77e-03, grad_scale: 16.0 2023-05-16 20:27:15,700 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=243128.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 20:27:19,498 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.45 vs. limit=2.0 2023-05-16 20:27:23,420 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=243139.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 20:27:30,642 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.3043, 5.0787, 5.2039, 5.2645, 4.8645, 4.9287, 4.6655, 5.2004], device='cuda:0'), covar=tensor([0.0678, 0.0635, 0.0897, 0.0584, 0.1852, 0.1309, 0.0562, 0.1079], device='cuda:0'), in_proj_covar=tensor([0.0545, 0.0712, 0.0623, 0.0634, 0.0859, 0.0756, 0.0557, 0.0489], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0005, 0.0004, 0.0004, 0.0005, 0.0005, 0.0003, 0.0003], device='cuda:0') 2023-05-16 20:27:46,823 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.919e+02 2.711e+02 3.242e+02 3.742e+02 1.088e+03, threshold=6.484e+02, percent-clipped=3.0 2023-05-16 20:27:48,278 INFO [finetune.py:992] (0/2) Epoch 12, batch 8600, loss[loss=0.1656, simple_loss=0.2608, pruned_loss=0.03523, over 12052.00 frames. ], tot_loss[loss=0.1692, simple_loss=0.2574, pruned_loss=0.04048, over 2363561.02 frames. ], batch size: 37, lr: 3.77e-03, grad_scale: 16.0 2023-05-16 20:27:49,897 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.1618, 5.9760, 5.5337, 5.5961, 6.0967, 5.4192, 5.6448, 5.5749], device='cuda:0'), covar=tensor([0.1214, 0.0871, 0.0917, 0.1601, 0.0813, 0.1915, 0.1509, 0.0986], device='cuda:0'), in_proj_covar=tensor([0.0352, 0.0489, 0.0387, 0.0436, 0.0461, 0.0425, 0.0388, 0.0381], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 20:27:57,588 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=243187.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 20:28:01,465 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 2023-05-16 20:28:15,640 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=243211.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 20:28:24,844 INFO [finetune.py:992] (0/2) Epoch 12, batch 8650, loss[loss=0.1794, simple_loss=0.272, pruned_loss=0.04345, over 10687.00 frames. ], tot_loss[loss=0.1695, simple_loss=0.2578, pruned_loss=0.04058, over 2356769.44 frames. ], batch size: 69, lr: 3.77e-03, grad_scale: 16.0 2023-05-16 20:28:27,933 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.5282, 3.5869, 3.2953, 3.2679, 2.9173, 2.7788, 3.6733, 2.3056], device='cuda:0'), covar=tensor([0.0430, 0.0160, 0.0196, 0.0205, 0.0373, 0.0402, 0.0133, 0.0498], device='cuda:0'), in_proj_covar=tensor([0.0195, 0.0162, 0.0164, 0.0191, 0.0204, 0.0202, 0.0174, 0.0206], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 20:28:48,828 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=243258.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 20:28:49,522 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=243259.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 20:28:59,526 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.984e+02 2.814e+02 3.184e+02 3.863e+02 6.884e+02, threshold=6.368e+02, percent-clipped=2.0 2023-05-16 20:29:00,496 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.6844, 3.3394, 5.1697, 2.5576, 2.8367, 3.7699, 3.1879, 3.7749], device='cuda:0'), covar=tensor([0.0460, 0.1128, 0.0259, 0.1155, 0.1835, 0.1443, 0.1343, 0.1200], device='cuda:0'), in_proj_covar=tensor([0.0237, 0.0235, 0.0251, 0.0182, 0.0238, 0.0295, 0.0221, 0.0262], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-16 20:29:00,951 INFO [finetune.py:992] (0/2) Epoch 12, batch 8700, loss[loss=0.1777, simple_loss=0.2668, pruned_loss=0.04434, over 12343.00 frames. ], tot_loss[loss=0.1689, simple_loss=0.2573, pruned_loss=0.04028, over 2366966.96 frames. ], batch size: 36, lr: 3.77e-03, grad_scale: 32.0 2023-05-16 20:29:03,196 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=243277.0, num_to_drop=1, layers_to_drop={2} 2023-05-16 20:29:14,451 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=243293.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 20:29:23,857 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=243306.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 20:29:24,337 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.36 vs. limit=2.0 2023-05-16 20:29:36,493 INFO [finetune.py:992] (0/2) Epoch 12, batch 8750, loss[loss=0.1753, simple_loss=0.2596, pruned_loss=0.04555, over 12088.00 frames. ], tot_loss[loss=0.1701, simple_loss=0.2579, pruned_loss=0.04113, over 2358123.13 frames. ], batch size: 32, lr: 3.77e-03, grad_scale: 32.0 2023-05-16 20:29:58,639 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=243354.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 20:30:00,847 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.3621, 3.3994, 3.1664, 3.1054, 2.7072, 2.5904, 3.4298, 2.2141], device='cuda:0'), covar=tensor([0.0426, 0.0153, 0.0200, 0.0195, 0.0404, 0.0362, 0.0133, 0.0516], device='cuda:0'), in_proj_covar=tensor([0.0196, 0.0162, 0.0165, 0.0192, 0.0204, 0.0203, 0.0174, 0.0207], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 20:30:10,894 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.791e+02 2.851e+02 3.290e+02 4.022e+02 1.120e+03, threshold=6.581e+02, percent-clipped=4.0 2023-05-16 20:30:12,369 INFO [finetune.py:992] (0/2) Epoch 12, batch 8800, loss[loss=0.1581, simple_loss=0.2439, pruned_loss=0.03612, over 12341.00 frames. ], tot_loss[loss=0.1707, simple_loss=0.2588, pruned_loss=0.04129, over 2360138.81 frames. ], batch size: 30, lr: 3.77e-03, grad_scale: 32.0 2023-05-16 20:30:30,448 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.1134, 4.5568, 3.9879, 4.8529, 4.4629, 2.7820, 4.1624, 3.0670], device='cuda:0'), covar=tensor([0.0963, 0.0830, 0.1539, 0.0591, 0.1115, 0.1759, 0.1119, 0.3319], device='cuda:0'), in_proj_covar=tensor([0.0308, 0.0383, 0.0359, 0.0317, 0.0372, 0.0273, 0.0347, 0.0372], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 20:30:49,066 INFO [finetune.py:992] (0/2) Epoch 12, batch 8850, loss[loss=0.1726, simple_loss=0.2634, pruned_loss=0.04093, over 12099.00 frames. ], tot_loss[loss=0.1714, simple_loss=0.2595, pruned_loss=0.04166, over 2350443.85 frames. ], batch size: 33, lr: 3.77e-03, grad_scale: 32.0 2023-05-16 20:30:51,872 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=243428.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 20:31:11,349 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.3620, 5.1737, 5.2829, 5.3184, 4.9503, 4.9860, 4.8071, 5.2325], device='cuda:0'), covar=tensor([0.0666, 0.0637, 0.0779, 0.0639, 0.2061, 0.1298, 0.0508, 0.1159], device='cuda:0'), in_proj_covar=tensor([0.0542, 0.0706, 0.0616, 0.0630, 0.0853, 0.0747, 0.0553, 0.0484], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0005, 0.0004, 0.0004, 0.0005, 0.0005, 0.0003, 0.0003], device='cuda:0') 2023-05-16 20:31:21,961 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.2588, 5.1946, 5.0173, 4.6354, 4.7351, 5.1724, 4.8173, 4.6420], device='cuda:0'), covar=tensor([0.0674, 0.0923, 0.0720, 0.1473, 0.1016, 0.0743, 0.1525, 0.0918], device='cuda:0'), in_proj_covar=tensor([0.0619, 0.0561, 0.0519, 0.0634, 0.0415, 0.0713, 0.0781, 0.0570], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-05-16 20:31:23,248 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.857e+02 2.679e+02 3.186e+02 3.892e+02 5.819e+02, threshold=6.373e+02, percent-clipped=0.0 2023-05-16 20:31:24,690 INFO [finetune.py:992] (0/2) Epoch 12, batch 8900, loss[loss=0.1534, simple_loss=0.2449, pruned_loss=0.03091, over 12352.00 frames. ], tot_loss[loss=0.1706, simple_loss=0.2592, pruned_loss=0.04102, over 2355559.23 frames. ], batch size: 31, lr: 3.77e-03, grad_scale: 16.0 2023-05-16 20:31:26,077 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=243476.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 20:31:26,294 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.1007, 4.4083, 3.8817, 4.7407, 4.3334, 2.7797, 4.0781, 3.0621], device='cuda:0'), covar=tensor([0.0840, 0.0835, 0.1549, 0.0508, 0.1043, 0.1651, 0.1007, 0.3173], device='cuda:0'), in_proj_covar=tensor([0.0306, 0.0381, 0.0357, 0.0314, 0.0370, 0.0272, 0.0344, 0.0370], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 20:32:00,865 INFO [finetune.py:992] (0/2) Epoch 12, batch 8950, loss[loss=0.1523, simple_loss=0.2389, pruned_loss=0.0329, over 12115.00 frames. ], tot_loss[loss=0.1695, simple_loss=0.2582, pruned_loss=0.04039, over 2363791.83 frames. ], batch size: 30, lr: 3.77e-03, grad_scale: 16.0 2023-05-16 20:32:36,325 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.876e+02 2.807e+02 3.224e+02 3.679e+02 7.560e+02, threshold=6.449e+02, percent-clipped=1.0 2023-05-16 20:32:36,857 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.39 vs. limit=2.0 2023-05-16 20:32:37,008 INFO [finetune.py:992] (0/2) Epoch 12, batch 9000, loss[loss=0.1693, simple_loss=0.2591, pruned_loss=0.03975, over 12312.00 frames. ], tot_loss[loss=0.1713, simple_loss=0.2598, pruned_loss=0.04139, over 2346619.13 frames. ], batch size: 34, lr: 3.77e-03, grad_scale: 16.0 2023-05-16 20:32:37,009 INFO [finetune.py:1017] (0/2) Computing validation loss 2023-05-16 20:32:46,897 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.8443, 4.0684, 3.8290, 4.5061, 4.1556, 2.5292, 3.8778, 2.9142], device='cuda:0'), covar=tensor([0.0981, 0.0995, 0.1368, 0.0554, 0.1049, 0.2028, 0.1249, 0.3883], device='cuda:0'), in_proj_covar=tensor([0.0306, 0.0380, 0.0356, 0.0313, 0.0369, 0.0271, 0.0344, 0.0369], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 20:32:55,019 INFO [finetune.py:1026] (0/2) Epoch 12, validation: loss=0.3264, simple_loss=0.3973, pruned_loss=0.1277, over 1020973.00 frames. 2023-05-16 20:32:55,020 INFO [finetune.py:1027] (0/2) Maximum memory allocated so far is 12827MB 2023-05-16 20:32:57,329 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=243577.0, num_to_drop=1, layers_to_drop={1} 2023-05-16 20:32:57,434 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.0477, 2.4934, 3.6149, 2.9617, 3.5259, 3.0977, 2.3729, 3.6046], device='cuda:0'), covar=tensor([0.0159, 0.0360, 0.0176, 0.0276, 0.0154, 0.0217, 0.0408, 0.0122], device='cuda:0'), in_proj_covar=tensor([0.0179, 0.0206, 0.0190, 0.0189, 0.0218, 0.0166, 0.0198, 0.0190], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 20:33:27,672 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.6151, 4.4972, 4.4064, 4.4531, 4.1511, 4.5793, 4.5515, 4.7520], device='cuda:0'), covar=tensor([0.0280, 0.0215, 0.0260, 0.0406, 0.0822, 0.0524, 0.0213, 0.0241], device='cuda:0'), in_proj_covar=tensor([0.0197, 0.0201, 0.0194, 0.0253, 0.0247, 0.0223, 0.0179, 0.0232], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0005, 0.0004, 0.0005, 0.0004, 0.0005], device='cuda:0') 2023-05-16 20:33:31,804 INFO [finetune.py:992] (0/2) Epoch 12, batch 9050, loss[loss=0.145, simple_loss=0.2244, pruned_loss=0.03277, over 12003.00 frames. ], tot_loss[loss=0.1703, simple_loss=0.2593, pruned_loss=0.04066, over 2355551.73 frames. ], batch size: 28, lr: 3.77e-03, grad_scale: 16.0 2023-05-16 20:33:32,604 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=243625.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 20:33:38,499 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.5988, 3.7608, 3.4896, 3.5472, 3.3310, 3.1612, 3.9002, 2.3778], device='cuda:0'), covar=tensor([0.0494, 0.0259, 0.0187, 0.0176, 0.0285, 0.0285, 0.0145, 0.0537], device='cuda:0'), in_proj_covar=tensor([0.0195, 0.0163, 0.0164, 0.0192, 0.0205, 0.0203, 0.0175, 0.0207], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 20:33:50,375 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=243649.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 20:34:04,638 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.2343, 4.7853, 5.0361, 5.0937, 4.9456, 5.0935, 5.0836, 2.5987], device='cuda:0'), covar=tensor([0.0088, 0.0069, 0.0063, 0.0052, 0.0040, 0.0086, 0.0090, 0.0787], device='cuda:0'), in_proj_covar=tensor([0.0069, 0.0079, 0.0081, 0.0074, 0.0060, 0.0093, 0.0082, 0.0097], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-16 20:34:06,805 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.3397, 4.6181, 2.8278, 2.7797, 3.9913, 2.5342, 3.9336, 3.2067], device='cuda:0'), covar=tensor([0.0686, 0.0474, 0.1170, 0.1366, 0.0255, 0.1369, 0.0458, 0.0802], device='cuda:0'), in_proj_covar=tensor([0.0192, 0.0259, 0.0177, 0.0200, 0.0143, 0.0183, 0.0201, 0.0178], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 20:34:07,276 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.960e+02 2.482e+02 3.089e+02 3.852e+02 1.074e+03, threshold=6.178e+02, percent-clipped=3.0 2023-05-16 20:34:07,976 INFO [finetune.py:992] (0/2) Epoch 12, batch 9100, loss[loss=0.2132, simple_loss=0.2874, pruned_loss=0.06951, over 8017.00 frames. ], tot_loss[loss=0.1698, simple_loss=0.2587, pruned_loss=0.04048, over 2352746.03 frames. ], batch size: 98, lr: 3.77e-03, grad_scale: 16.0 2023-05-16 20:34:23,865 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.62 vs. limit=2.0 2023-05-16 20:34:43,341 INFO [finetune.py:992] (0/2) Epoch 12, batch 9150, loss[loss=0.1853, simple_loss=0.2761, pruned_loss=0.04722, over 11760.00 frames. ], tot_loss[loss=0.169, simple_loss=0.258, pruned_loss=0.03999, over 2357422.45 frames. ], batch size: 44, lr: 3.77e-03, grad_scale: 16.0 2023-05-16 20:35:18,964 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.782e+02 2.646e+02 3.221e+02 3.838e+02 1.025e+03, threshold=6.442e+02, percent-clipped=3.0 2023-05-16 20:35:19,700 INFO [finetune.py:992] (0/2) Epoch 12, batch 9200, loss[loss=0.1709, simple_loss=0.2596, pruned_loss=0.04104, over 12158.00 frames. ], tot_loss[loss=0.1681, simple_loss=0.2574, pruned_loss=0.03944, over 2364457.66 frames. ], batch size: 36, lr: 3.77e-03, grad_scale: 16.0 2023-05-16 20:35:51,077 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.3370, 3.1973, 4.7879, 2.5726, 2.6401, 3.4464, 2.9911, 3.5719], device='cuda:0'), covar=tensor([0.0589, 0.1345, 0.0427, 0.1277, 0.2237, 0.1773, 0.1567, 0.1368], device='cuda:0'), in_proj_covar=tensor([0.0238, 0.0237, 0.0252, 0.0183, 0.0240, 0.0297, 0.0224, 0.0265], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-16 20:35:56,673 INFO [finetune.py:992] (0/2) Epoch 12, batch 9250, loss[loss=0.169, simple_loss=0.256, pruned_loss=0.04102, over 12301.00 frames. ], tot_loss[loss=0.1685, simple_loss=0.2578, pruned_loss=0.03961, over 2354294.60 frames. ], batch size: 34, lr: 3.77e-03, grad_scale: 16.0 2023-05-16 20:35:58,269 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.9779, 5.9551, 5.7340, 5.2573, 5.1438, 5.8493, 5.5055, 5.2909], device='cuda:0'), covar=tensor([0.0805, 0.1007, 0.0727, 0.1719, 0.0730, 0.0841, 0.1812, 0.1046], device='cuda:0'), in_proj_covar=tensor([0.0622, 0.0565, 0.0519, 0.0637, 0.0415, 0.0717, 0.0785, 0.0570], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0004, 0.0004, 0.0003], device='cuda:0') 2023-05-16 20:36:31,437 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.938e+02 2.684e+02 3.034e+02 3.618e+02 6.950e+02, threshold=6.067e+02, percent-clipped=3.0 2023-05-16 20:36:32,181 INFO [finetune.py:992] (0/2) Epoch 12, batch 9300, loss[loss=0.153, simple_loss=0.2381, pruned_loss=0.03393, over 12093.00 frames. ], tot_loss[loss=0.1687, simple_loss=0.2582, pruned_loss=0.03963, over 2352471.55 frames. ], batch size: 32, lr: 3.77e-03, grad_scale: 16.0 2023-05-16 20:37:09,173 INFO [finetune.py:992] (0/2) Epoch 12, batch 9350, loss[loss=0.1568, simple_loss=0.2502, pruned_loss=0.03173, over 12355.00 frames. ], tot_loss[loss=0.1679, simple_loss=0.2569, pruned_loss=0.03941, over 2358672.47 frames. ], batch size: 35, lr: 3.77e-03, grad_scale: 16.0 2023-05-16 20:37:26,879 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=243949.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 20:37:43,791 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.868e+02 2.953e+02 3.600e+02 4.345e+02 1.678e+03, threshold=7.200e+02, percent-clipped=9.0 2023-05-16 20:37:44,505 INFO [finetune.py:992] (0/2) Epoch 12, batch 9400, loss[loss=0.1717, simple_loss=0.2546, pruned_loss=0.04438, over 12342.00 frames. ], tot_loss[loss=0.1679, simple_loss=0.2566, pruned_loss=0.03956, over 2353789.85 frames. ], batch size: 31, lr: 3.77e-03, grad_scale: 16.0 2023-05-16 20:37:51,795 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.5711, 3.2251, 4.9632, 2.5091, 2.6692, 3.6857, 2.9999, 3.8465], device='cuda:0'), covar=tensor([0.0504, 0.1279, 0.0431, 0.1301, 0.2266, 0.1704, 0.1593, 0.1117], device='cuda:0'), in_proj_covar=tensor([0.0238, 0.0237, 0.0253, 0.0183, 0.0240, 0.0297, 0.0224, 0.0266], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-16 20:37:54,768 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.2998, 4.6565, 4.0778, 4.9895, 4.5337, 2.8629, 4.3001, 3.1417], device='cuda:0'), covar=tensor([0.0742, 0.0777, 0.1390, 0.0424, 0.1044, 0.1680, 0.0940, 0.3027], device='cuda:0'), in_proj_covar=tensor([0.0305, 0.0380, 0.0357, 0.0313, 0.0368, 0.0270, 0.0344, 0.0368], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 20:38:00,841 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=243997.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 20:38:03,159 INFO [checkpoint.py:75] (0/2) Saving checkpoint to pruned_transducer_stateless7/exp_giga_finetune/checkpoint-144000.pt 2023-05-16 20:38:24,496 INFO [finetune.py:992] (0/2) Epoch 12, batch 9450, loss[loss=0.1517, simple_loss=0.2347, pruned_loss=0.03438, over 12277.00 frames. ], tot_loss[loss=0.1681, simple_loss=0.2567, pruned_loss=0.03973, over 2359349.14 frames. ], batch size: 28, lr: 3.76e-03, grad_scale: 16.0 2023-05-16 20:38:59,561 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=4.64 vs. limit=5.0 2023-05-16 20:39:00,429 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.877e+02 2.575e+02 3.107e+02 4.027e+02 6.079e+02, threshold=6.214e+02, percent-clipped=0.0 2023-05-16 20:39:01,150 INFO [finetune.py:992] (0/2) Epoch 12, batch 9500, loss[loss=0.1844, simple_loss=0.2749, pruned_loss=0.04692, over 12021.00 frames. ], tot_loss[loss=0.1677, simple_loss=0.2568, pruned_loss=0.03932, over 2362407.78 frames. ], batch size: 40, lr: 3.76e-03, grad_scale: 16.0 2023-05-16 20:39:37,075 INFO [finetune.py:992] (0/2) Epoch 12, batch 9550, loss[loss=0.1483, simple_loss=0.2328, pruned_loss=0.03196, over 12356.00 frames. ], tot_loss[loss=0.167, simple_loss=0.2563, pruned_loss=0.03885, over 2371217.92 frames. ], batch size: 31, lr: 3.76e-03, grad_scale: 16.0 2023-05-16 20:40:12,709 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.763e+02 2.660e+02 3.055e+02 3.615e+02 7.485e+02, threshold=6.109e+02, percent-clipped=2.0 2023-05-16 20:40:13,438 INFO [finetune.py:992] (0/2) Epoch 12, batch 9600, loss[loss=0.1628, simple_loss=0.2513, pruned_loss=0.0371, over 12090.00 frames. ], tot_loss[loss=0.1674, simple_loss=0.2566, pruned_loss=0.03913, over 2373743.91 frames. ], batch size: 32, lr: 3.76e-03, grad_scale: 16.0 2023-05-16 20:40:50,242 INFO [finetune.py:992] (0/2) Epoch 12, batch 9650, loss[loss=0.1711, simple_loss=0.261, pruned_loss=0.04065, over 12364.00 frames. ], tot_loss[loss=0.1674, simple_loss=0.2567, pruned_loss=0.03906, over 2374351.72 frames. ], batch size: 38, lr: 3.76e-03, grad_scale: 16.0 2023-05-16 20:41:22,543 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.5711, 2.5070, 3.8286, 4.3966, 3.9529, 4.3358, 3.8778, 3.2028], device='cuda:0'), covar=tensor([0.0030, 0.0400, 0.0127, 0.0044, 0.0108, 0.0096, 0.0119, 0.0335], device='cuda:0'), in_proj_covar=tensor([0.0089, 0.0125, 0.0106, 0.0079, 0.0105, 0.0117, 0.0097, 0.0140], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 20:41:22,790 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2023-05-16 20:41:25,341 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.862e+02 2.732e+02 3.120e+02 3.808e+02 5.883e+02, threshold=6.241e+02, percent-clipped=0.0 2023-05-16 20:41:26,078 INFO [finetune.py:992] (0/2) Epoch 12, batch 9700, loss[loss=0.1545, simple_loss=0.2392, pruned_loss=0.03486, over 12132.00 frames. ], tot_loss[loss=0.1679, simple_loss=0.2571, pruned_loss=0.03937, over 2375704.87 frames. ], batch size: 30, lr: 3.76e-03, grad_scale: 16.0 2023-05-16 20:41:41,121 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.8268, 4.6452, 4.7847, 4.8017, 4.3776, 4.3577, 4.3079, 4.6876], device='cuda:0'), covar=tensor([0.0917, 0.0925, 0.1085, 0.0860, 0.2422, 0.1858, 0.0706, 0.1246], device='cuda:0'), in_proj_covar=tensor([0.0544, 0.0714, 0.0618, 0.0631, 0.0855, 0.0750, 0.0559, 0.0485], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0005, 0.0004, 0.0004, 0.0005, 0.0005, 0.0003, 0.0003], device='cuda:0') 2023-05-16 20:42:02,519 INFO [finetune.py:992] (0/2) Epoch 12, batch 9750, loss[loss=0.1842, simple_loss=0.2772, pruned_loss=0.04555, over 12296.00 frames. ], tot_loss[loss=0.1679, simple_loss=0.2569, pruned_loss=0.03944, over 2371655.12 frames. ], batch size: 34, lr: 3.76e-03, grad_scale: 16.0 2023-05-16 20:42:19,906 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.49 vs. limit=2.0 2023-05-16 20:42:23,956 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=244353.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 20:42:38,002 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.021e+02 2.796e+02 3.294e+02 3.819e+02 9.587e+02, threshold=6.587e+02, percent-clipped=2.0 2023-05-16 20:42:38,712 INFO [finetune.py:992] (0/2) Epoch 12, batch 9800, loss[loss=0.1745, simple_loss=0.2653, pruned_loss=0.04186, over 11553.00 frames. ], tot_loss[loss=0.168, simple_loss=0.2571, pruned_loss=0.03946, over 2373826.47 frames. ], batch size: 48, lr: 3.76e-03, grad_scale: 16.0 2023-05-16 20:42:40,487 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.66 vs. limit=2.0 2023-05-16 20:42:51,718 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.3326, 4.6625, 2.9007, 2.6586, 3.9637, 2.2504, 3.9663, 2.9602], device='cuda:0'), covar=tensor([0.0660, 0.0506, 0.1086, 0.1426, 0.0368, 0.1577, 0.0531, 0.0966], device='cuda:0'), in_proj_covar=tensor([0.0191, 0.0258, 0.0176, 0.0200, 0.0143, 0.0181, 0.0200, 0.0177], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 20:42:56,153 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.3582, 3.1193, 3.1812, 3.5355, 2.5573, 3.1907, 2.5209, 3.0186], device='cuda:0'), covar=tensor([0.1322, 0.0737, 0.0751, 0.0553, 0.0914, 0.0663, 0.1547, 0.0764], device='cuda:0'), in_proj_covar=tensor([0.0227, 0.0263, 0.0296, 0.0357, 0.0241, 0.0241, 0.0261, 0.0376], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 20:43:07,862 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=244414.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 20:43:14,897 INFO [finetune.py:992] (0/2) Epoch 12, batch 9850, loss[loss=0.1476, simple_loss=0.2408, pruned_loss=0.0272, over 12074.00 frames. ], tot_loss[loss=0.1687, simple_loss=0.2579, pruned_loss=0.03978, over 2371624.95 frames. ], batch size: 42, lr: 3.76e-03, grad_scale: 16.0 2023-05-16 20:43:26,860 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.90 vs. limit=2.0 2023-05-16 20:43:40,134 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.6373, 4.3000, 4.2472, 4.6930, 3.3746, 4.1768, 2.9165, 4.3479], device='cuda:0'), covar=tensor([0.1347, 0.0582, 0.0864, 0.0573, 0.1023, 0.0511, 0.1554, 0.1360], device='cuda:0'), in_proj_covar=tensor([0.0228, 0.0265, 0.0298, 0.0360, 0.0242, 0.0242, 0.0262, 0.0379], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 20:43:47,281 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.1412, 5.0369, 4.8915, 5.0308, 4.6502, 5.0416, 5.1064, 5.3543], device='cuda:0'), covar=tensor([0.0286, 0.0162, 0.0203, 0.0277, 0.0788, 0.0344, 0.0201, 0.0176], device='cuda:0'), in_proj_covar=tensor([0.0197, 0.0200, 0.0193, 0.0251, 0.0247, 0.0223, 0.0180, 0.0230], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0005, 0.0004, 0.0005, 0.0004, 0.0005], device='cuda:0') 2023-05-16 20:43:50,739 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.885e+02 2.647e+02 3.033e+02 3.603e+02 6.141e+02, threshold=6.066e+02, percent-clipped=0.0 2023-05-16 20:43:51,448 INFO [finetune.py:992] (0/2) Epoch 12, batch 9900, loss[loss=0.1788, simple_loss=0.2667, pruned_loss=0.04544, over 12152.00 frames. ], tot_loss[loss=0.1675, simple_loss=0.2567, pruned_loss=0.03914, over 2377457.49 frames. ], batch size: 34, lr: 3.76e-03, grad_scale: 16.0 2023-05-16 20:43:51,634 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.8416, 3.5165, 3.6217, 3.7679, 3.7544, 3.7962, 3.6463, 2.4285], device='cuda:0'), covar=tensor([0.0097, 0.0129, 0.0130, 0.0083, 0.0060, 0.0113, 0.0098, 0.0734], device='cuda:0'), in_proj_covar=tensor([0.0070, 0.0081, 0.0083, 0.0076, 0.0062, 0.0095, 0.0084, 0.0099], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 20:43:54,510 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.1429, 2.3036, 3.0903, 4.0657, 2.2374, 4.1508, 4.1646, 4.2772], device='cuda:0'), covar=tensor([0.0182, 0.1466, 0.0560, 0.0163, 0.1482, 0.0262, 0.0193, 0.0111], device='cuda:0'), in_proj_covar=tensor([0.0127, 0.0205, 0.0186, 0.0122, 0.0193, 0.0184, 0.0180, 0.0122], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 20:44:28,027 INFO [finetune.py:992] (0/2) Epoch 12, batch 9950, loss[loss=0.1517, simple_loss=0.2433, pruned_loss=0.03002, over 12302.00 frames. ], tot_loss[loss=0.1677, simple_loss=0.2571, pruned_loss=0.03911, over 2376995.58 frames. ], batch size: 34, lr: 3.76e-03, grad_scale: 16.0 2023-05-16 20:44:32,274 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.9650, 5.9940, 5.7077, 5.3136, 5.1984, 5.8538, 5.4877, 5.3011], device='cuda:0'), covar=tensor([0.0821, 0.0817, 0.0712, 0.1495, 0.0675, 0.0692, 0.1522, 0.1133], device='cuda:0'), in_proj_covar=tensor([0.0620, 0.0559, 0.0520, 0.0634, 0.0411, 0.0714, 0.0784, 0.0571], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-05-16 20:45:02,736 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.659e+02 2.695e+02 3.297e+02 3.922e+02 9.943e+02, threshold=6.593e+02, percent-clipped=6.0 2023-05-16 20:45:03,465 INFO [finetune.py:992] (0/2) Epoch 12, batch 10000, loss[loss=0.1598, simple_loss=0.2412, pruned_loss=0.03922, over 12408.00 frames. ], tot_loss[loss=0.1678, simple_loss=0.2573, pruned_loss=0.03915, over 2382854.98 frames. ], batch size: 32, lr: 3.76e-03, grad_scale: 16.0 2023-05-16 20:45:19,985 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=4.31 vs. limit=5.0 2023-05-16 20:45:21,002 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.1332, 4.7392, 4.9616, 4.9636, 4.8767, 5.0109, 4.9566, 2.5856], device='cuda:0'), covar=tensor([0.0096, 0.0067, 0.0082, 0.0067, 0.0047, 0.0093, 0.0069, 0.0822], device='cuda:0'), in_proj_covar=tensor([0.0071, 0.0081, 0.0084, 0.0076, 0.0062, 0.0095, 0.0084, 0.0099], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 20:45:39,699 INFO [finetune.py:992] (0/2) Epoch 12, batch 10050, loss[loss=0.1641, simple_loss=0.2526, pruned_loss=0.03779, over 12423.00 frames. ], tot_loss[loss=0.1683, simple_loss=0.2574, pruned_loss=0.03961, over 2379011.80 frames. ], batch size: 32, lr: 3.76e-03, grad_scale: 16.0 2023-05-16 20:46:15,313 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.753e+02 2.632e+02 3.142e+02 3.893e+02 5.782e+02, threshold=6.284e+02, percent-clipped=0.0 2023-05-16 20:46:16,041 INFO [finetune.py:992] (0/2) Epoch 12, batch 10100, loss[loss=0.1661, simple_loss=0.2638, pruned_loss=0.03422, over 12355.00 frames. ], tot_loss[loss=0.1681, simple_loss=0.2573, pruned_loss=0.03948, over 2380590.11 frames. ], batch size: 35, lr: 3.76e-03, grad_scale: 16.0 2023-05-16 20:46:33,325 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.4214, 4.9155, 4.1794, 5.0418, 4.5855, 2.5946, 4.1754, 3.0021], device='cuda:0'), covar=tensor([0.0728, 0.0641, 0.1382, 0.0573, 0.1094, 0.1840, 0.1118, 0.3306], device='cuda:0'), in_proj_covar=tensor([0.0303, 0.0378, 0.0355, 0.0312, 0.0366, 0.0269, 0.0344, 0.0366], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 20:46:40,971 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=244709.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 20:46:46,671 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=244717.0, num_to_drop=1, layers_to_drop={1} 2023-05-16 20:46:51,513 INFO [finetune.py:992] (0/2) Epoch 12, batch 10150, loss[loss=0.1483, simple_loss=0.2292, pruned_loss=0.03371, over 11776.00 frames. ], tot_loss[loss=0.1687, simple_loss=0.2577, pruned_loss=0.03978, over 2382893.26 frames. ], batch size: 26, lr: 3.76e-03, grad_scale: 16.0 2023-05-16 20:47:12,777 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.81 vs. limit=2.0 2023-05-16 20:47:27,291 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.751e+02 2.627e+02 3.161e+02 3.759e+02 7.063e+02, threshold=6.321e+02, percent-clipped=1.0 2023-05-16 20:47:27,964 INFO [finetune.py:992] (0/2) Epoch 12, batch 10200, loss[loss=0.1801, simple_loss=0.2743, pruned_loss=0.0429, over 12189.00 frames. ], tot_loss[loss=0.1691, simple_loss=0.2581, pruned_loss=0.0401, over 2370113.65 frames. ], batch size: 35, lr: 3.76e-03, grad_scale: 16.0 2023-05-16 20:47:31,216 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=244778.0, num_to_drop=1, layers_to_drop={3} 2023-05-16 20:47:32,683 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.2742, 2.7272, 3.7871, 3.1769, 3.7096, 3.3643, 2.7043, 3.7306], device='cuda:0'), covar=tensor([0.0140, 0.0337, 0.0129, 0.0232, 0.0128, 0.0165, 0.0350, 0.0113], device='cuda:0'), in_proj_covar=tensor([0.0179, 0.0206, 0.0190, 0.0187, 0.0218, 0.0166, 0.0197, 0.0190], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 20:47:52,796 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.4735, 4.7102, 2.9756, 2.3393, 4.0582, 2.2985, 4.0916, 3.1391], device='cuda:0'), covar=tensor([0.0544, 0.0588, 0.1021, 0.1807, 0.0398, 0.1619, 0.0446, 0.0915], device='cuda:0'), in_proj_covar=tensor([0.0192, 0.0260, 0.0178, 0.0201, 0.0144, 0.0181, 0.0201, 0.0177], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 20:48:04,516 INFO [finetune.py:992] (0/2) Epoch 12, batch 10250, loss[loss=0.1937, simple_loss=0.2832, pruned_loss=0.05211, over 12129.00 frames. ], tot_loss[loss=0.1695, simple_loss=0.2585, pruned_loss=0.04025, over 2360873.74 frames. ], batch size: 38, lr: 3.76e-03, grad_scale: 16.0 2023-05-16 20:48:39,155 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.907e+02 2.708e+02 3.036e+02 3.649e+02 6.724e+02, threshold=6.073e+02, percent-clipped=1.0 2023-05-16 20:48:39,901 INFO [finetune.py:992] (0/2) Epoch 12, batch 10300, loss[loss=0.2198, simple_loss=0.2982, pruned_loss=0.07069, over 8071.00 frames. ], tot_loss[loss=0.1693, simple_loss=0.2585, pruned_loss=0.04008, over 2357014.35 frames. ], batch size: 98, lr: 3.76e-03, grad_scale: 16.0 2023-05-16 20:48:49,222 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=244887.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 20:49:16,452 INFO [finetune.py:992] (0/2) Epoch 12, batch 10350, loss[loss=0.1597, simple_loss=0.2461, pruned_loss=0.03664, over 12037.00 frames. ], tot_loss[loss=0.1691, simple_loss=0.2585, pruned_loss=0.03986, over 2363880.67 frames. ], batch size: 31, lr: 3.76e-03, grad_scale: 16.0 2023-05-16 20:49:21,290 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=4.93 vs. limit=5.0 2023-05-16 20:49:33,680 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=244948.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 20:49:37,914 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=244954.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 20:49:39,437 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.1163, 4.1010, 4.2069, 4.5254, 3.0186, 4.0689, 2.7423, 4.1238], device='cuda:0'), covar=tensor([0.1589, 0.0630, 0.0815, 0.0587, 0.1092, 0.0560, 0.1699, 0.1120], device='cuda:0'), in_proj_covar=tensor([0.0228, 0.0267, 0.0300, 0.0362, 0.0243, 0.0243, 0.0264, 0.0381], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 20:49:51,291 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.813e+02 2.759e+02 3.200e+02 3.746e+02 7.063e+02, threshold=6.400e+02, percent-clipped=2.0 2023-05-16 20:49:52,045 INFO [finetune.py:992] (0/2) Epoch 12, batch 10400, loss[loss=0.1444, simple_loss=0.2277, pruned_loss=0.0306, over 12186.00 frames. ], tot_loss[loss=0.1683, simple_loss=0.2576, pruned_loss=0.0395, over 2370320.86 frames. ], batch size: 29, lr: 3.76e-03, grad_scale: 16.0 2023-05-16 20:50:06,567 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=2.85 vs. limit=5.0 2023-05-16 20:50:12,333 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.5605, 2.9144, 4.5040, 4.6993, 2.8314, 2.4989, 2.9837, 2.2295], device='cuda:0'), covar=tensor([0.1588, 0.2810, 0.0473, 0.0390, 0.1327, 0.2496, 0.2679, 0.3906], device='cuda:0'), in_proj_covar=tensor([0.0295, 0.0384, 0.0272, 0.0301, 0.0272, 0.0305, 0.0381, 0.0371], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-16 20:50:17,212 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=245009.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 20:50:21,451 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=245015.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 20:50:27,509 INFO [finetune.py:992] (0/2) Epoch 12, batch 10450, loss[loss=0.1646, simple_loss=0.2461, pruned_loss=0.04155, over 12004.00 frames. ], tot_loss[loss=0.1684, simple_loss=0.2575, pruned_loss=0.03961, over 2372650.45 frames. ], batch size: 28, lr: 3.76e-03, grad_scale: 16.0 2023-05-16 20:50:48,956 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.4836, 5.0231, 5.4022, 4.7600, 5.0859, 4.8651, 5.4796, 5.0536], device='cuda:0'), covar=tensor([0.0253, 0.0410, 0.0308, 0.0250, 0.0380, 0.0369, 0.0228, 0.0352], device='cuda:0'), in_proj_covar=tensor([0.0263, 0.0268, 0.0293, 0.0264, 0.0265, 0.0267, 0.0240, 0.0219], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 20:50:51,728 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=245057.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 20:51:03,629 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.731e+02 2.694e+02 3.120e+02 3.983e+02 1.025e+03, threshold=6.241e+02, percent-clipped=4.0 2023-05-16 20:51:03,729 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=245073.0, num_to_drop=1, layers_to_drop={3} 2023-05-16 20:51:04,384 INFO [finetune.py:992] (0/2) Epoch 12, batch 10500, loss[loss=0.1476, simple_loss=0.2254, pruned_loss=0.03491, over 12163.00 frames. ], tot_loss[loss=0.1689, simple_loss=0.258, pruned_loss=0.03989, over 2365397.57 frames. ], batch size: 29, lr: 3.76e-03, grad_scale: 16.0 2023-05-16 20:51:13,901 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.51 vs. limit=2.0 2023-05-16 20:51:14,737 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.87 vs. limit=2.0 2023-05-16 20:51:40,158 INFO [finetune.py:992] (0/2) Epoch 12, batch 10550, loss[loss=0.1834, simple_loss=0.268, pruned_loss=0.04945, over 10510.00 frames. ], tot_loss[loss=0.1681, simple_loss=0.257, pruned_loss=0.03958, over 2369974.34 frames. ], batch size: 68, lr: 3.76e-03, grad_scale: 16.0 2023-05-16 20:52:14,923 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.581e+02 2.721e+02 3.204e+02 3.875e+02 6.361e+02, threshold=6.407e+02, percent-clipped=1.0 2023-05-16 20:52:15,669 INFO [finetune.py:992] (0/2) Epoch 12, batch 10600, loss[loss=0.1421, simple_loss=0.2348, pruned_loss=0.02472, over 12037.00 frames. ], tot_loss[loss=0.1671, simple_loss=0.2562, pruned_loss=0.03905, over 2373935.20 frames. ], batch size: 31, lr: 3.75e-03, grad_scale: 16.0 2023-05-16 20:52:33,382 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.35 vs. limit=2.0 2023-05-16 20:52:42,790 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.8664, 3.3152, 2.4305, 2.1533, 2.9440, 2.2610, 3.1583, 2.6439], device='cuda:0'), covar=tensor([0.0598, 0.0650, 0.0961, 0.1392, 0.0296, 0.1155, 0.0507, 0.0770], device='cuda:0'), in_proj_covar=tensor([0.0194, 0.0262, 0.0179, 0.0203, 0.0145, 0.0183, 0.0202, 0.0179], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 20:52:53,135 INFO [finetune.py:992] (0/2) Epoch 12, batch 10650, loss[loss=0.1638, simple_loss=0.2473, pruned_loss=0.04016, over 12245.00 frames. ], tot_loss[loss=0.1675, simple_loss=0.2564, pruned_loss=0.03924, over 2372508.61 frames. ], batch size: 32, lr: 3.75e-03, grad_scale: 16.0 2023-05-16 20:53:06,613 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=245243.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 20:53:17,514 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.9425, 4.5675, 4.1198, 4.2384, 4.6674, 3.9957, 4.2042, 4.0669], device='cuda:0'), covar=tensor([0.1664, 0.1304, 0.1689, 0.2091, 0.1220, 0.2569, 0.1917, 0.1555], device='cuda:0'), in_proj_covar=tensor([0.0368, 0.0507, 0.0400, 0.0454, 0.0479, 0.0444, 0.0399, 0.0398], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 20:53:25,652 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.7463, 2.8784, 4.4927, 4.6764, 2.7504, 2.5842, 2.9335, 2.1095], device='cuda:0'), covar=tensor([0.1497, 0.2854, 0.0456, 0.0412, 0.1360, 0.2374, 0.2606, 0.4072], device='cuda:0'), in_proj_covar=tensor([0.0297, 0.0388, 0.0275, 0.0303, 0.0274, 0.0308, 0.0385, 0.0374], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-16 20:53:28,253 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.009e+02 2.654e+02 3.047e+02 3.752e+02 8.474e+02, threshold=6.094e+02, percent-clipped=2.0 2023-05-16 20:53:28,990 INFO [finetune.py:992] (0/2) Epoch 12, batch 10700, loss[loss=0.1683, simple_loss=0.2576, pruned_loss=0.03957, over 12360.00 frames. ], tot_loss[loss=0.1684, simple_loss=0.2573, pruned_loss=0.03979, over 2365217.74 frames. ], batch size: 36, lr: 3.75e-03, grad_scale: 16.0 2023-05-16 20:53:53,838 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.5567, 2.7586, 3.7366, 4.4981, 3.9110, 4.4939, 3.8762, 3.1220], device='cuda:0'), covar=tensor([0.0039, 0.0369, 0.0131, 0.0048, 0.0107, 0.0070, 0.0130, 0.0368], device='cuda:0'), in_proj_covar=tensor([0.0090, 0.0127, 0.0108, 0.0080, 0.0106, 0.0120, 0.0099, 0.0142], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-16 20:53:54,458 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=245310.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 20:54:04,502 INFO [finetune.py:992] (0/2) Epoch 12, batch 10750, loss[loss=0.1442, simple_loss=0.2248, pruned_loss=0.03179, over 12276.00 frames. ], tot_loss[loss=0.1682, simple_loss=0.257, pruned_loss=0.03973, over 2360567.32 frames. ], batch size: 28, lr: 3.75e-03, grad_scale: 16.0 2023-05-16 20:54:14,637 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.1355, 3.7076, 3.8095, 4.3083, 2.9262, 3.7373, 2.5805, 3.8694], device='cuda:0'), covar=tensor([0.1671, 0.0824, 0.0930, 0.0677, 0.1097, 0.0632, 0.1810, 0.1059], device='cuda:0'), in_proj_covar=tensor([0.0231, 0.0267, 0.0300, 0.0362, 0.0243, 0.0243, 0.0265, 0.0382], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 20:54:16,779 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.1171, 2.2130, 2.6217, 3.0651, 2.2584, 3.1888, 3.1122, 3.2949], device='cuda:0'), covar=tensor([0.0195, 0.1060, 0.0527, 0.0245, 0.1019, 0.0315, 0.0339, 0.0136], device='cuda:0'), in_proj_covar=tensor([0.0124, 0.0203, 0.0184, 0.0121, 0.0189, 0.0182, 0.0177, 0.0121], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 20:54:33,148 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.4933, 3.5865, 3.3019, 3.2112, 3.0060, 2.6707, 3.7123, 2.5180], device='cuda:0'), covar=tensor([0.0407, 0.0143, 0.0188, 0.0224, 0.0358, 0.0362, 0.0115, 0.0439], device='cuda:0'), in_proj_covar=tensor([0.0194, 0.0163, 0.0166, 0.0191, 0.0204, 0.0201, 0.0174, 0.0205], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 20:54:40,586 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.894e+02 2.769e+02 3.205e+02 3.871e+02 6.976e+02, threshold=6.409e+02, percent-clipped=1.0 2023-05-16 20:54:40,732 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=245373.0, num_to_drop=1, layers_to_drop={0} 2023-05-16 20:54:41,350 INFO [finetune.py:992] (0/2) Epoch 12, batch 10800, loss[loss=0.1506, simple_loss=0.2329, pruned_loss=0.03415, over 12359.00 frames. ], tot_loss[loss=0.1678, simple_loss=0.2568, pruned_loss=0.03943, over 2366322.38 frames. ], batch size: 30, lr: 3.75e-03, grad_scale: 16.0 2023-05-16 20:54:48,605 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.4840, 4.7090, 4.3063, 5.1375, 4.6683, 3.0731, 4.3845, 3.1579], device='cuda:0'), covar=tensor([0.0761, 0.0931, 0.1395, 0.0484, 0.1156, 0.1568, 0.1008, 0.3361], device='cuda:0'), in_proj_covar=tensor([0.0303, 0.0378, 0.0356, 0.0311, 0.0366, 0.0268, 0.0344, 0.0366], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 20:54:54,154 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.0883, 3.8926, 4.0306, 4.4671, 3.0975, 3.8323, 2.6532, 4.0865], device='cuda:0'), covar=tensor([0.1622, 0.0753, 0.0833, 0.0556, 0.1058, 0.0625, 0.1701, 0.1028], device='cuda:0'), in_proj_covar=tensor([0.0231, 0.0267, 0.0300, 0.0362, 0.0243, 0.0243, 0.0264, 0.0382], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 20:55:14,316 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.7186, 5.3737, 4.8535, 4.9766, 5.5295, 4.7784, 5.0247, 4.8523], device='cuda:0'), covar=tensor([0.1538, 0.1152, 0.1392, 0.2136, 0.1145, 0.2544, 0.1838, 0.1315], device='cuda:0'), in_proj_covar=tensor([0.0366, 0.0503, 0.0399, 0.0452, 0.0476, 0.0443, 0.0398, 0.0398], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 20:55:15,012 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=245421.0, num_to_drop=1, layers_to_drop={1} 2023-05-16 20:55:17,000 INFO [finetune.py:992] (0/2) Epoch 12, batch 10850, loss[loss=0.1683, simple_loss=0.2616, pruned_loss=0.03752, over 12103.00 frames. ], tot_loss[loss=0.1671, simple_loss=0.2562, pruned_loss=0.03906, over 2372072.03 frames. ], batch size: 33, lr: 3.75e-03, grad_scale: 16.0 2023-05-16 20:55:22,204 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.2705, 2.3978, 3.0951, 4.0542, 2.2389, 4.2267, 4.2176, 4.3568], device='cuda:0'), covar=tensor([0.0114, 0.1191, 0.0511, 0.0174, 0.1316, 0.0215, 0.0186, 0.0100], device='cuda:0'), in_proj_covar=tensor([0.0124, 0.0203, 0.0184, 0.0122, 0.0190, 0.0182, 0.0178, 0.0121], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 20:55:43,544 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=245460.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 20:55:53,052 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.872e+02 2.731e+02 3.179e+02 3.624e+02 5.889e+02, threshold=6.357e+02, percent-clipped=0.0 2023-05-16 20:55:53,700 INFO [finetune.py:992] (0/2) Epoch 12, batch 10900, loss[loss=0.173, simple_loss=0.264, pruned_loss=0.04097, over 12251.00 frames. ], tot_loss[loss=0.1672, simple_loss=0.2565, pruned_loss=0.03895, over 2374564.75 frames. ], batch size: 32, lr: 3.75e-03, grad_scale: 32.0 2023-05-16 20:56:28,734 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=245521.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 20:56:30,641 INFO [finetune.py:992] (0/2) Epoch 12, batch 10950, loss[loss=0.1584, simple_loss=0.2521, pruned_loss=0.03236, over 12352.00 frames. ], tot_loss[loss=0.1674, simple_loss=0.2565, pruned_loss=0.03911, over 2373363.13 frames. ], batch size: 36, lr: 3.75e-03, grad_scale: 32.0 2023-05-16 20:56:44,455 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=245543.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 20:56:46,427 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=245546.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 20:56:58,567 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.3837, 3.3632, 3.1775, 3.1019, 2.8827, 2.5600, 3.4987, 2.3772], device='cuda:0'), covar=tensor([0.0402, 0.0167, 0.0194, 0.0197, 0.0360, 0.0350, 0.0137, 0.0464], device='cuda:0'), in_proj_covar=tensor([0.0194, 0.0163, 0.0165, 0.0191, 0.0204, 0.0201, 0.0174, 0.0206], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 20:57:05,482 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.160e+02 2.725e+02 3.222e+02 4.178e+02 8.072e+02, threshold=6.445e+02, percent-clipped=5.0 2023-05-16 20:57:06,142 INFO [finetune.py:992] (0/2) Epoch 12, batch 11000, loss[loss=0.2166, simple_loss=0.3131, pruned_loss=0.06002, over 11285.00 frames. ], tot_loss[loss=0.1705, simple_loss=0.2593, pruned_loss=0.04086, over 2347266.56 frames. ], batch size: 55, lr: 3.75e-03, grad_scale: 32.0 2023-05-16 20:57:18,326 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=245591.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 20:57:29,968 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=245607.0, num_to_drop=1, layers_to_drop={1} 2023-05-16 20:57:32,614 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=245610.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 20:57:42,942 INFO [finetune.py:992] (0/2) Epoch 12, batch 11050, loss[loss=0.1556, simple_loss=0.2445, pruned_loss=0.03333, over 12360.00 frames. ], tot_loss[loss=0.1742, simple_loss=0.2626, pruned_loss=0.04286, over 2309139.01 frames. ], batch size: 30, lr: 3.75e-03, grad_scale: 32.0 2023-05-16 20:58:02,149 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.5435, 2.5537, 3.2911, 4.4424, 2.3446, 4.4475, 4.5087, 4.7082], device='cuda:0'), covar=tensor([0.0174, 0.1300, 0.0503, 0.0172, 0.1451, 0.0286, 0.0190, 0.0100], device='cuda:0'), in_proj_covar=tensor([0.0123, 0.0201, 0.0183, 0.0121, 0.0189, 0.0181, 0.0177, 0.0121], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 20:58:05,345 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.69 vs. limit=2.0 2023-05-16 20:58:06,921 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=245658.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 20:58:17,295 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.303e+02 3.015e+02 3.784e+02 4.411e+02 6.473e+02, threshold=7.569e+02, percent-clipped=1.0 2023-05-16 20:58:17,989 INFO [finetune.py:992] (0/2) Epoch 12, batch 11100, loss[loss=0.1585, simple_loss=0.2517, pruned_loss=0.03267, over 12112.00 frames. ], tot_loss[loss=0.1784, simple_loss=0.2666, pruned_loss=0.04505, over 2267278.71 frames. ], batch size: 33, lr: 3.75e-03, grad_scale: 32.0 2023-05-16 20:58:26,545 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=245686.0, num_to_drop=1, layers_to_drop={0} 2023-05-16 20:58:54,225 INFO [finetune.py:992] (0/2) Epoch 12, batch 11150, loss[loss=0.1933, simple_loss=0.2849, pruned_loss=0.05088, over 11575.00 frames. ], tot_loss[loss=0.186, simple_loss=0.2737, pruned_loss=0.04919, over 2204946.32 frames. ], batch size: 48, lr: 3.75e-03, grad_scale: 32.0 2023-05-16 20:58:59,276 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.8334, 2.3267, 2.8836, 2.5715, 2.9753, 2.8973, 2.8905, 2.3719], device='cuda:0'), covar=tensor([0.0080, 0.0315, 0.0147, 0.0131, 0.0106, 0.0114, 0.0138, 0.0345], device='cuda:0'), in_proj_covar=tensor([0.0088, 0.0124, 0.0105, 0.0079, 0.0104, 0.0117, 0.0097, 0.0139], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 20:59:10,536 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=245747.0, num_to_drop=1, layers_to_drop={0} 2023-05-16 20:59:29,233 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.300e+02 3.344e+02 3.973e+02 4.906e+02 8.289e+02, threshold=7.946e+02, percent-clipped=2.0 2023-05-16 20:59:29,262 INFO [finetune.py:992] (0/2) Epoch 12, batch 11200, loss[loss=0.1675, simple_loss=0.2576, pruned_loss=0.0387, over 12187.00 frames. ], tot_loss[loss=0.1935, simple_loss=0.2802, pruned_loss=0.05342, over 2143508.74 frames. ], batch size: 31, lr: 3.75e-03, grad_scale: 16.0 2023-05-16 20:59:43,286 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=245792.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 20:59:56,100 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([1.9535, 2.1583, 2.2661, 2.1675, 2.1326, 1.8667, 2.0915, 1.6520], device='cuda:0'), covar=tensor([0.0285, 0.0168, 0.0140, 0.0174, 0.0253, 0.0218, 0.0177, 0.0376], device='cuda:0'), in_proj_covar=tensor([0.0191, 0.0161, 0.0162, 0.0188, 0.0202, 0.0198, 0.0172, 0.0204], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 21:00:00,147 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=245816.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 21:00:05,671 INFO [finetune.py:992] (0/2) Epoch 12, batch 11250, loss[loss=0.2648, simple_loss=0.3596, pruned_loss=0.08497, over 10313.00 frames. ], tot_loss[loss=0.2017, simple_loss=0.2874, pruned_loss=0.05805, over 2081099.05 frames. ], batch size: 68, lr: 3.75e-03, grad_scale: 16.0 2023-05-16 21:00:09,705 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2023-05-16 21:00:26,508 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=245853.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 21:00:41,265 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.328e+02 3.283e+02 4.161e+02 5.005e+02 1.339e+03, threshold=8.323e+02, percent-clipped=2.0 2023-05-16 21:00:41,283 INFO [finetune.py:992] (0/2) Epoch 12, batch 11300, loss[loss=0.1908, simple_loss=0.2924, pruned_loss=0.04457, over 12211.00 frames. ], tot_loss[loss=0.2074, simple_loss=0.2923, pruned_loss=0.06122, over 2030575.01 frames. ], batch size: 35, lr: 3.75e-03, grad_scale: 16.0 2023-05-16 21:00:52,983 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=245891.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 21:01:00,482 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=245902.0, num_to_drop=1, layers_to_drop={0} 2023-05-16 21:01:07,992 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.6775, 3.8947, 3.5636, 3.5269, 3.3005, 3.0933, 3.9119, 2.6340], device='cuda:0'), covar=tensor([0.0394, 0.0133, 0.0157, 0.0188, 0.0299, 0.0290, 0.0149, 0.0462], device='cuda:0'), in_proj_covar=tensor([0.0192, 0.0161, 0.0162, 0.0189, 0.0203, 0.0198, 0.0173, 0.0204], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 21:01:15,890 INFO [finetune.py:992] (0/2) Epoch 12, batch 11350, loss[loss=0.1963, simple_loss=0.2825, pruned_loss=0.05501, over 12000.00 frames. ], tot_loss[loss=0.2122, simple_loss=0.2963, pruned_loss=0.06406, over 1990247.39 frames. ], batch size: 40, lr: 3.75e-03, grad_scale: 16.0 2023-05-16 21:01:22,881 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.1518, 4.9483, 5.0832, 5.1366, 4.8077, 4.8276, 4.6810, 5.0154], device='cuda:0'), covar=tensor([0.0807, 0.0645, 0.0772, 0.0567, 0.1649, 0.1309, 0.0527, 0.1126], device='cuda:0'), in_proj_covar=tensor([0.0532, 0.0694, 0.0603, 0.0620, 0.0822, 0.0730, 0.0546, 0.0475], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0005, 0.0004, 0.0003, 0.0003], device='cuda:0') 2023-05-16 21:01:35,695 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=245952.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 21:01:50,523 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.992e+02 3.504e+02 4.241e+02 5.083e+02 1.276e+03, threshold=8.481e+02, percent-clipped=3.0 2023-05-16 21:01:50,542 INFO [finetune.py:992] (0/2) Epoch 12, batch 11400, loss[loss=0.2483, simple_loss=0.3214, pruned_loss=0.08757, over 6881.00 frames. ], tot_loss[loss=0.2164, simple_loss=0.2995, pruned_loss=0.06664, over 1943527.64 frames. ], batch size: 98, lr: 3.75e-03, grad_scale: 16.0 2023-05-16 21:02:10,103 INFO [checkpoint.py:75] (0/2) Saving checkpoint to pruned_transducer_stateless7/exp_giga_finetune/checkpoint-146000.pt 2023-05-16 21:02:19,018 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.8504, 2.3378, 3.4297, 3.5233, 3.0098, 2.6880, 2.6061, 2.4528], device='cuda:0'), covar=tensor([0.1268, 0.3027, 0.0663, 0.0450, 0.0819, 0.2189, 0.2725, 0.3844], device='cuda:0'), in_proj_covar=tensor([0.0291, 0.0378, 0.0268, 0.0295, 0.0267, 0.0301, 0.0378, 0.0368], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-16 21:02:29,775 INFO [finetune.py:992] (0/2) Epoch 12, batch 11450, loss[loss=0.2785, simple_loss=0.3459, pruned_loss=0.1056, over 7262.00 frames. ], tot_loss[loss=0.2211, simple_loss=0.3031, pruned_loss=0.06959, over 1890423.12 frames. ], batch size: 99, lr: 3.75e-03, grad_scale: 16.0 2023-05-16 21:02:34,799 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 2023-05-16 21:02:41,845 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=246042.0, num_to_drop=1, layers_to_drop={1} 2023-05-16 21:03:04,523 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.706e+02 3.486e+02 3.950e+02 4.755e+02 6.665e+02, threshold=7.899e+02, percent-clipped=0.0 2023-05-16 21:03:04,542 INFO [finetune.py:992] (0/2) Epoch 12, batch 11500, loss[loss=0.2589, simple_loss=0.3301, pruned_loss=0.09383, over 7215.00 frames. ], tot_loss[loss=0.225, simple_loss=0.306, pruned_loss=0.07198, over 1848019.16 frames. ], batch size: 98, lr: 3.75e-03, grad_scale: 16.0 2023-05-16 21:03:25,770 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.1985, 4.2552, 2.7001, 2.1992, 3.7901, 2.0039, 3.8280, 2.6313], device='cuda:0'), covar=tensor([0.0723, 0.0443, 0.1157, 0.1783, 0.0240, 0.1918, 0.0389, 0.1093], device='cuda:0'), in_proj_covar=tensor([0.0188, 0.0252, 0.0175, 0.0198, 0.0141, 0.0180, 0.0195, 0.0174], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 21:03:27,160 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.5616, 2.6817, 3.7852, 4.3944, 4.0488, 4.3369, 4.0208, 2.9270], device='cuda:0'), covar=tensor([0.0034, 0.0392, 0.0118, 0.0049, 0.0109, 0.0090, 0.0098, 0.0443], device='cuda:0'), in_proj_covar=tensor([0.0086, 0.0122, 0.0102, 0.0077, 0.0101, 0.0114, 0.0095, 0.0136], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 21:03:29,907 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=246111.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 21:03:31,390 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.8618, 2.8305, 4.5058, 2.4472, 2.4151, 3.7663, 2.9325, 3.7427], device='cuda:0'), covar=tensor([0.0679, 0.1481, 0.0184, 0.1356, 0.2204, 0.1045, 0.1632, 0.0828], device='cuda:0'), in_proj_covar=tensor([0.0226, 0.0226, 0.0237, 0.0177, 0.0230, 0.0281, 0.0214, 0.0253], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:0') 2023-05-16 21:03:34,726 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=246116.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 21:03:40,071 INFO [finetune.py:992] (0/2) Epoch 12, batch 11550, loss[loss=0.2188, simple_loss=0.3016, pruned_loss=0.06796, over 10193.00 frames. ], tot_loss[loss=0.2268, simple_loss=0.3075, pruned_loss=0.07307, over 1843174.99 frames. ], batch size: 68, lr: 3.75e-03, grad_scale: 16.0 2023-05-16 21:03:55,902 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-05-16 21:03:56,235 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=246148.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 21:04:07,117 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=246164.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 21:04:12,598 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=246172.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 21:04:14,370 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.699e+02 3.686e+02 4.061e+02 4.741e+02 8.417e+02, threshold=8.121e+02, percent-clipped=2.0 2023-05-16 21:04:14,393 INFO [finetune.py:992] (0/2) Epoch 12, batch 11600, loss[loss=0.2335, simple_loss=0.309, pruned_loss=0.07897, over 7563.00 frames. ], tot_loss[loss=0.2299, simple_loss=0.3095, pruned_loss=0.0751, over 1798211.48 frames. ], batch size: 100, lr: 3.75e-03, grad_scale: 16.0 2023-05-16 21:04:30,403 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=246196.0, num_to_drop=1, layers_to_drop={1} 2023-05-16 21:04:35,057 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=246202.0, num_to_drop=1, layers_to_drop={2} 2023-05-16 21:04:48,982 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=246221.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 21:04:51,085 INFO [finetune.py:992] (0/2) Epoch 12, batch 11650, loss[loss=0.2035, simple_loss=0.2894, pruned_loss=0.05881, over 11040.00 frames. ], tot_loss[loss=0.2289, simple_loss=0.3084, pruned_loss=0.07473, over 1801307.09 frames. ], batch size: 55, lr: 3.75e-03, grad_scale: 16.0 2023-05-16 21:04:55,132 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.5491, 4.1740, 4.2489, 4.5423, 4.4276, 4.5066, 4.5206, 2.3297], device='cuda:0'), covar=tensor([0.0104, 0.0120, 0.0175, 0.0075, 0.0066, 0.0139, 0.0095, 0.1261], device='cuda:0'), in_proj_covar=tensor([0.0067, 0.0077, 0.0079, 0.0072, 0.0059, 0.0090, 0.0080, 0.0095], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-16 21:05:08,452 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=246247.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 21:05:10,587 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=246250.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 21:05:15,225 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=246257.0, num_to_drop=1, layers_to_drop={0} 2023-05-16 21:05:26,395 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.526e+02 3.461e+02 3.908e+02 4.540e+02 6.500e+02, threshold=7.816e+02, percent-clipped=0.0 2023-05-16 21:05:26,414 INFO [finetune.py:992] (0/2) Epoch 12, batch 11700, loss[loss=0.2722, simple_loss=0.3277, pruned_loss=0.1083, over 7016.00 frames. ], tot_loss[loss=0.2313, simple_loss=0.3094, pruned_loss=0.07657, over 1759049.86 frames. ], batch size: 100, lr: 3.74e-03, grad_scale: 16.0 2023-05-16 21:05:32,046 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=246282.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 21:06:01,227 INFO [finetune.py:992] (0/2) Epoch 12, batch 11750, loss[loss=0.2809, simple_loss=0.3295, pruned_loss=0.1161, over 6652.00 frames. ], tot_loss[loss=0.2318, simple_loss=0.3097, pruned_loss=0.07694, over 1740536.35 frames. ], batch size: 98, lr: 3.74e-03, grad_scale: 16.0 2023-05-16 21:06:13,562 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=246342.0, num_to_drop=1, layers_to_drop={2} 2023-05-16 21:06:36,452 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.683e+02 3.527e+02 4.253e+02 4.983e+02 1.025e+03, threshold=8.505e+02, percent-clipped=3.0 2023-05-16 21:06:36,471 INFO [finetune.py:992] (0/2) Epoch 12, batch 11800, loss[loss=0.2747, simple_loss=0.337, pruned_loss=0.1062, over 6253.00 frames. ], tot_loss[loss=0.2342, simple_loss=0.3118, pruned_loss=0.07827, over 1731695.75 frames. ], batch size: 98, lr: 3.74e-03, grad_scale: 16.0 2023-05-16 21:06:43,435 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.1484, 2.0018, 2.1783, 1.9783, 2.1806, 2.2784, 1.8114, 2.2106], device='cuda:0'), covar=tensor([0.0119, 0.0283, 0.0127, 0.0196, 0.0152, 0.0138, 0.0258, 0.0126], device='cuda:0'), in_proj_covar=tensor([0.0170, 0.0198, 0.0179, 0.0180, 0.0207, 0.0157, 0.0190, 0.0181], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 21:06:46,788 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.7298, 3.7336, 3.6930, 3.8147, 3.6459, 3.6740, 3.5711, 3.7444], device='cuda:0'), covar=tensor([0.1282, 0.0687, 0.1356, 0.0663, 0.1467, 0.1137, 0.0532, 0.0907], device='cuda:0'), in_proj_covar=tensor([0.0506, 0.0659, 0.0573, 0.0587, 0.0776, 0.0689, 0.0518, 0.0454], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0004, 0.0004, 0.0005, 0.0004, 0.0003, 0.0003], device='cuda:0') 2023-05-16 21:06:47,411 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=246390.0, num_to_drop=1, layers_to_drop={1} 2023-05-16 21:07:11,791 INFO [finetune.py:992] (0/2) Epoch 12, batch 11850, loss[loss=0.271, simple_loss=0.3336, pruned_loss=0.1042, over 6968.00 frames. ], tot_loss[loss=0.2368, simple_loss=0.3138, pruned_loss=0.07987, over 1701573.43 frames. ], batch size: 98, lr: 3.74e-03, grad_scale: 16.0 2023-05-16 21:07:18,067 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.6513, 4.3674, 4.5991, 4.1395, 4.3923, 4.1984, 4.6086, 4.2863], device='cuda:0'), covar=tensor([0.0273, 0.0349, 0.0289, 0.0262, 0.0360, 0.0319, 0.0237, 0.0608], device='cuda:0'), in_proj_covar=tensor([0.0239, 0.0242, 0.0265, 0.0240, 0.0240, 0.0242, 0.0216, 0.0196], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 21:07:28,096 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=246448.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 21:07:40,994 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=246467.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 21:07:42,363 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.8772, 4.5234, 4.1673, 4.2224, 4.5891, 3.9950, 4.2203, 4.0449], device='cuda:0'), covar=tensor([0.1562, 0.1004, 0.1273, 0.1858, 0.0993, 0.2152, 0.1610, 0.1358], device='cuda:0'), in_proj_covar=tensor([0.0346, 0.0470, 0.0380, 0.0427, 0.0447, 0.0416, 0.0371, 0.0375], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 21:07:45,471 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.582e+02 3.580e+02 4.280e+02 5.047e+02 8.856e+02, threshold=8.560e+02, percent-clipped=2.0 2023-05-16 21:07:45,490 INFO [finetune.py:992] (0/2) Epoch 12, batch 11900, loss[loss=0.2541, simple_loss=0.3205, pruned_loss=0.09378, over 6989.00 frames. ], tot_loss[loss=0.2354, simple_loss=0.3132, pruned_loss=0.07878, over 1700788.65 frames. ], batch size: 100, lr: 3.74e-03, grad_scale: 16.0 2023-05-16 21:07:52,561 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.2770, 3.8996, 3.8670, 4.4423, 2.9801, 3.7314, 2.3322, 3.8811], device='cuda:0'), covar=tensor([0.1986, 0.0868, 0.1104, 0.0658, 0.1380, 0.0879, 0.2530, 0.1192], device='cuda:0'), in_proj_covar=tensor([0.0227, 0.0263, 0.0290, 0.0348, 0.0237, 0.0240, 0.0260, 0.0371], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 21:08:01,867 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=246496.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 21:08:20,795 INFO [finetune.py:992] (0/2) Epoch 12, batch 11950, loss[loss=0.1795, simple_loss=0.2748, pruned_loss=0.04205, over 11472.00 frames. ], tot_loss[loss=0.2306, simple_loss=0.3099, pruned_loss=0.07567, over 1698012.47 frames. ], batch size: 48, lr: 3.74e-03, grad_scale: 16.0 2023-05-16 21:08:37,887 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=246547.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 21:08:41,224 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=246552.0, num_to_drop=1, layers_to_drop={2} 2023-05-16 21:08:46,192 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2023-05-16 21:08:55,986 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.884e+02 2.981e+02 3.497e+02 4.415e+02 8.803e+02, threshold=6.994e+02, percent-clipped=1.0 2023-05-16 21:08:56,004 INFO [finetune.py:992] (0/2) Epoch 12, batch 12000, loss[loss=0.1958, simple_loss=0.2935, pruned_loss=0.04902, over 10242.00 frames. ], tot_loss[loss=0.2239, simple_loss=0.3049, pruned_loss=0.07147, over 1701948.88 frames. ], batch size: 68, lr: 3.74e-03, grad_scale: 16.0 2023-05-16 21:08:56,005 INFO [finetune.py:1017] (0/2) Computing validation loss 2023-05-16 21:09:12,811 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.5598, 4.0739, 3.6542, 4.3635, 3.8264, 2.4327, 3.8110, 2.6556], device='cuda:0'), covar=tensor([0.1203, 0.1002, 0.1597, 0.0515, 0.1750, 0.2349, 0.1318, 0.4381], device='cuda:0'), in_proj_covar=tensor([0.0287, 0.0356, 0.0333, 0.0287, 0.0345, 0.0257, 0.0322, 0.0346], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-16 21:09:14,139 INFO [finetune.py:1026] (0/2) Epoch 12, validation: loss=0.2869, simple_loss=0.3623, pruned_loss=0.1058, over 1020973.00 frames. 2023-05-16 21:09:14,140 INFO [finetune.py:1027] (0/2) Maximum memory allocated so far is 12827MB 2023-05-16 21:09:16,293 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=246577.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 21:09:29,252 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=246595.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 21:09:44,703 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2023-05-16 21:09:49,053 INFO [finetune.py:992] (0/2) Epoch 12, batch 12050, loss[loss=0.1852, simple_loss=0.2856, pruned_loss=0.04246, over 10323.00 frames. ], tot_loss[loss=0.2185, simple_loss=0.3007, pruned_loss=0.06816, over 1710036.44 frames. ], batch size: 68, lr: 3.74e-03, grad_scale: 16.0 2023-05-16 21:10:22,291 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.133e+02 2.926e+02 3.521e+02 4.054e+02 6.335e+02, threshold=7.041e+02, percent-clipped=0.0 2023-05-16 21:10:22,310 INFO [finetune.py:992] (0/2) Epoch 12, batch 12100, loss[loss=0.2418, simple_loss=0.3094, pruned_loss=0.08714, over 7017.00 frames. ], tot_loss[loss=0.2167, simple_loss=0.2993, pruned_loss=0.06705, over 1707024.32 frames. ], batch size: 98, lr: 3.74e-03, grad_scale: 16.0 2023-05-16 21:10:24,335 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=246677.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 21:10:54,392 INFO [finetune.py:992] (0/2) Epoch 12, batch 12150, loss[loss=0.2412, simple_loss=0.3179, pruned_loss=0.08224, over 6986.00 frames. ], tot_loss[loss=0.2175, simple_loss=0.2998, pruned_loss=0.06763, over 1704272.94 frames. ], batch size: 98, lr: 3.74e-03, grad_scale: 16.0 2023-05-16 21:10:54,568 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.8615, 2.1003, 2.6406, 2.9356, 2.1822, 3.0139, 2.9196, 3.0453], device='cuda:0'), covar=tensor([0.0171, 0.1189, 0.0505, 0.0229, 0.1147, 0.0298, 0.0371, 0.0172], device='cuda:0'), in_proj_covar=tensor([0.0116, 0.0195, 0.0175, 0.0114, 0.0183, 0.0171, 0.0166, 0.0116], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 21:10:59,648 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.8844, 2.2105, 2.8598, 2.8336, 2.9371, 2.9884, 2.8480, 2.4392], device='cuda:0'), covar=tensor([0.0078, 0.0377, 0.0165, 0.0085, 0.0138, 0.0103, 0.0153, 0.0327], device='cuda:0'), in_proj_covar=tensor([0.0084, 0.0121, 0.0099, 0.0075, 0.0100, 0.0111, 0.0093, 0.0135], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 21:11:03,491 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=246738.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 21:11:09,888 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.1557, 2.3627, 3.6343, 4.0174, 3.8399, 4.0686, 3.7391, 3.0388], device='cuda:0'), covar=tensor([0.0052, 0.0455, 0.0126, 0.0061, 0.0110, 0.0086, 0.0110, 0.0390], device='cuda:0'), in_proj_covar=tensor([0.0084, 0.0121, 0.0099, 0.0075, 0.0100, 0.0111, 0.0093, 0.0135], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 21:11:21,770 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=246767.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 21:11:25,893 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.965e+02 3.313e+02 3.843e+02 4.649e+02 1.350e+03, threshold=7.686e+02, percent-clipped=5.0 2023-05-16 21:11:25,912 INFO [finetune.py:992] (0/2) Epoch 12, batch 12200, loss[loss=0.2122, simple_loss=0.2898, pruned_loss=0.06727, over 6512.00 frames. ], tot_loss[loss=0.2202, simple_loss=0.3015, pruned_loss=0.06944, over 1671160.26 frames. ], batch size: 98, lr: 3.74e-03, grad_scale: 16.0 2023-05-16 21:11:33,913 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.5752, 4.5089, 4.4406, 4.1011, 4.1439, 4.5383, 4.3115, 4.1549], device='cuda:0'), covar=tensor([0.0783, 0.0953, 0.0645, 0.1199, 0.1693, 0.0833, 0.1267, 0.0976], device='cuda:0'), in_proj_covar=tensor([0.0578, 0.0517, 0.0479, 0.0579, 0.0386, 0.0657, 0.0702, 0.0526], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002], device='cuda:0') 2023-05-16 21:11:48,102 INFO [checkpoint.py:75] (0/2) Saving checkpoint to pruned_transducer_stateless7/exp_giga_finetune/epoch-12.pt 2023-05-16 21:12:10,844 INFO [finetune.py:992] (0/2) Epoch 13, batch 0, loss[loss=0.1926, simple_loss=0.2862, pruned_loss=0.04947, over 12165.00 frames. ], tot_loss[loss=0.1926, simple_loss=0.2862, pruned_loss=0.04947, over 12165.00 frames. ], batch size: 34, lr: 3.74e-03, grad_scale: 8.0 2023-05-16 21:12:10,845 INFO [finetune.py:1017] (0/2) Computing validation loss 2023-05-16 21:12:28,456 INFO [finetune.py:1026] (0/2) Epoch 13, validation: loss=0.2846, simple_loss=0.3601, pruned_loss=0.1046, over 1020973.00 frames. 2023-05-16 21:12:28,457 INFO [finetune.py:1027] (0/2) Maximum memory allocated so far is 12827MB 2023-05-16 21:12:33,555 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=246815.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 21:12:45,791 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.7001, 2.7767, 4.6743, 4.7759, 2.8870, 2.6169, 2.9376, 2.0031], device='cuda:0'), covar=tensor([0.1805, 0.3554, 0.0463, 0.0409, 0.1368, 0.2776, 0.3131, 0.4971], device='cuda:0'), in_proj_covar=tensor([0.0283, 0.0369, 0.0260, 0.0286, 0.0259, 0.0295, 0.0370, 0.0363], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-16 21:12:51,739 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 2023-05-16 21:12:59,892 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=246852.0, num_to_drop=1, layers_to_drop={1} 2023-05-16 21:13:01,309 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.3599, 6.0599, 5.7557, 5.6351, 6.1557, 5.4288, 5.6476, 5.6781], device='cuda:0'), covar=tensor([0.1526, 0.0926, 0.0954, 0.2197, 0.0827, 0.2292, 0.1787, 0.1165], device='cuda:0'), in_proj_covar=tensor([0.0343, 0.0465, 0.0375, 0.0423, 0.0441, 0.0411, 0.0369, 0.0369], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002], device='cuda:0') 2023-05-16 21:13:02,820 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.7053, 4.6477, 4.5611, 4.5713, 4.3128, 4.6200, 4.6825, 4.8416], device='cuda:0'), covar=tensor([0.0246, 0.0163, 0.0206, 0.0376, 0.0747, 0.0343, 0.0183, 0.0235], device='cuda:0'), in_proj_covar=tensor([0.0171, 0.0173, 0.0168, 0.0217, 0.0213, 0.0192, 0.0156, 0.0202], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:0') 2023-05-16 21:13:04,069 INFO [finetune.py:992] (0/2) Epoch 13, batch 50, loss[loss=0.2144, simple_loss=0.3113, pruned_loss=0.05878, over 12126.00 frames. ], tot_loss[loss=0.1775, simple_loss=0.2676, pruned_loss=0.04367, over 531731.00 frames. ], batch size: 39, lr: 3.74e-03, grad_scale: 8.0 2023-05-16 21:13:16,859 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.322e+02 2.937e+02 3.620e+02 4.068e+02 8.964e+02, threshold=7.240e+02, percent-clipped=1.0 2023-05-16 21:13:18,358 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=246877.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 21:13:32,029 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.0309, 2.2627, 2.2610, 2.3238, 2.1325, 1.9914, 2.2812, 1.8070], device='cuda:0'), covar=tensor([0.0334, 0.0201, 0.0198, 0.0222, 0.0382, 0.0269, 0.0191, 0.0420], device='cuda:0'), in_proj_covar=tensor([0.0184, 0.0151, 0.0153, 0.0180, 0.0192, 0.0189, 0.0163, 0.0194], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 21:13:35,413 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=246900.0, num_to_drop=1, layers_to_drop={0} 2023-05-16 21:13:41,139 INFO [finetune.py:992] (0/2) Epoch 13, batch 100, loss[loss=0.1795, simple_loss=0.2762, pruned_loss=0.04135, over 12150.00 frames. ], tot_loss[loss=0.1781, simple_loss=0.2679, pruned_loss=0.04416, over 937546.88 frames. ], batch size: 34, lr: 3.74e-03, grad_scale: 8.0 2023-05-16 21:13:53,424 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=246925.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 21:13:59,974 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=246934.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 21:14:17,261 INFO [finetune.py:992] (0/2) Epoch 13, batch 150, loss[loss=0.1736, simple_loss=0.2793, pruned_loss=0.03388, over 12159.00 frames. ], tot_loss[loss=0.1745, simple_loss=0.2647, pruned_loss=0.04219, over 1260258.11 frames. ], batch size: 34, lr: 3.74e-03, grad_scale: 8.0 2023-05-16 21:14:29,051 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.053e+02 2.751e+02 3.175e+02 3.588e+02 8.004e+02, threshold=6.350e+02, percent-clipped=1.0 2023-05-16 21:14:43,250 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=246995.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 21:14:48,545 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.0216, 4.7430, 4.8998, 4.9407, 4.7684, 5.0639, 4.8506, 2.8835], device='cuda:0'), covar=tensor([0.0080, 0.0083, 0.0088, 0.0073, 0.0054, 0.0094, 0.0082, 0.0738], device='cuda:0'), in_proj_covar=tensor([0.0066, 0.0077, 0.0079, 0.0071, 0.0058, 0.0090, 0.0079, 0.0095], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-16 21:14:52,554 INFO [finetune.py:992] (0/2) Epoch 13, batch 200, loss[loss=0.1585, simple_loss=0.2437, pruned_loss=0.03665, over 12013.00 frames. ], tot_loss[loss=0.1726, simple_loss=0.2627, pruned_loss=0.04124, over 1513960.37 frames. ], batch size: 28, lr: 3.74e-03, grad_scale: 8.0 2023-05-16 21:15:09,985 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=2.24 vs. limit=5.0 2023-05-16 21:15:11,151 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=247033.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 21:15:29,836 INFO [finetune.py:992] (0/2) Epoch 13, batch 250, loss[loss=0.1785, simple_loss=0.2623, pruned_loss=0.04729, over 12017.00 frames. ], tot_loss[loss=0.1706, simple_loss=0.2605, pruned_loss=0.04039, over 1708700.03 frames. ], batch size: 40, lr: 3.74e-03, grad_scale: 8.0 2023-05-16 21:15:41,772 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.900e+02 2.777e+02 3.206e+02 3.750e+02 2.533e+03, threshold=6.411e+02, percent-clipped=1.0 2023-05-16 21:15:51,815 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.2605, 3.9851, 4.0185, 4.4619, 2.9322, 3.9423, 2.5930, 4.1687], device='cuda:0'), covar=tensor([0.1793, 0.0766, 0.1114, 0.0699, 0.1291, 0.0670, 0.2002, 0.1187], device='cuda:0'), in_proj_covar=tensor([0.0229, 0.0265, 0.0291, 0.0351, 0.0238, 0.0242, 0.0261, 0.0371], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 21:15:54,604 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.3115, 3.2252, 3.0631, 3.0449, 2.7320, 2.6432, 3.2908, 2.2071], device='cuda:0'), covar=tensor([0.0420, 0.0167, 0.0205, 0.0224, 0.0406, 0.0358, 0.0148, 0.0540], device='cuda:0'), in_proj_covar=tensor([0.0184, 0.0152, 0.0154, 0.0181, 0.0194, 0.0190, 0.0164, 0.0196], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 21:16:05,094 INFO [finetune.py:992] (0/2) Epoch 13, batch 300, loss[loss=0.1657, simple_loss=0.2677, pruned_loss=0.03184, over 11098.00 frames. ], tot_loss[loss=0.1697, simple_loss=0.2588, pruned_loss=0.0403, over 1852315.58 frames. ], batch size: 55, lr: 3.74e-03, grad_scale: 8.0 2023-05-16 21:16:40,535 INFO [finetune.py:992] (0/2) Epoch 13, batch 350, loss[loss=0.1717, simple_loss=0.2564, pruned_loss=0.04352, over 12348.00 frames. ], tot_loss[loss=0.171, simple_loss=0.2603, pruned_loss=0.04085, over 1978750.70 frames. ], batch size: 31, lr: 3.74e-03, grad_scale: 8.0 2023-05-16 21:16:52,656 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.785e+02 2.818e+02 3.215e+02 3.898e+02 9.522e+02, threshold=6.431e+02, percent-clipped=2.0 2023-05-16 21:16:56,287 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.7186, 3.2668, 5.1559, 2.6129, 2.7524, 3.6819, 3.1198, 3.6689], device='cuda:0'), covar=tensor([0.0490, 0.1254, 0.0251, 0.1271, 0.2164, 0.1798, 0.1576, 0.1419], device='cuda:0'), in_proj_covar=tensor([0.0232, 0.0233, 0.0242, 0.0182, 0.0235, 0.0287, 0.0221, 0.0259], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-16 21:17:05,822 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.41 vs. limit=2.0 2023-05-16 21:17:17,897 INFO [finetune.py:992] (0/2) Epoch 13, batch 400, loss[loss=0.1855, simple_loss=0.2739, pruned_loss=0.04851, over 11687.00 frames. ], tot_loss[loss=0.1716, simple_loss=0.2612, pruned_loss=0.041, over 2070949.31 frames. ], batch size: 48, lr: 3.73e-03, grad_scale: 8.0 2023-05-16 21:17:23,171 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.14 vs. limit=2.0 2023-05-16 21:17:28,609 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.7383, 4.5695, 4.4967, 4.5949, 4.2645, 4.6623, 4.6731, 4.9120], device='cuda:0'), covar=tensor([0.0289, 0.0187, 0.0246, 0.0378, 0.0830, 0.0350, 0.0208, 0.0216], device='cuda:0'), in_proj_covar=tensor([0.0181, 0.0185, 0.0179, 0.0232, 0.0228, 0.0205, 0.0167, 0.0214], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:0') 2023-05-16 21:17:39,326 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.8688, 3.3238, 2.3969, 2.0944, 2.9251, 2.2798, 3.1472, 2.6221], device='cuda:0'), covar=tensor([0.0688, 0.0718, 0.1069, 0.1515, 0.0334, 0.1227, 0.0514, 0.0864], device='cuda:0'), in_proj_covar=tensor([0.0188, 0.0251, 0.0177, 0.0199, 0.0140, 0.0181, 0.0194, 0.0176], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 21:17:53,511 INFO [finetune.py:992] (0/2) Epoch 13, batch 450, loss[loss=0.1725, simple_loss=0.2682, pruned_loss=0.03843, over 10509.00 frames. ], tot_loss[loss=0.1711, simple_loss=0.2607, pruned_loss=0.04072, over 2137456.44 frames. ], batch size: 68, lr: 3.73e-03, grad_scale: 8.0 2023-05-16 21:18:05,454 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.820e+02 2.692e+02 3.169e+02 3.840e+02 7.314e+02, threshold=6.338e+02, percent-clipped=2.0 2023-05-16 21:18:16,351 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=247290.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 21:18:28,764 INFO [finetune.py:992] (0/2) Epoch 13, batch 500, loss[loss=0.1812, simple_loss=0.2647, pruned_loss=0.04883, over 12263.00 frames. ], tot_loss[loss=0.1713, simple_loss=0.2607, pruned_loss=0.04094, over 2185786.00 frames. ], batch size: 28, lr: 3.73e-03, grad_scale: 8.0 2023-05-16 21:18:32,247 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.1609, 4.7515, 5.1162, 4.4069, 4.8350, 4.5672, 5.1583, 4.8323], device='cuda:0'), covar=tensor([0.0290, 0.0428, 0.0329, 0.0336, 0.0377, 0.0377, 0.0236, 0.0366], device='cuda:0'), in_proj_covar=tensor([0.0249, 0.0250, 0.0274, 0.0249, 0.0250, 0.0252, 0.0225, 0.0204], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 21:18:47,175 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=247333.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 21:19:05,201 INFO [finetune.py:992] (0/2) Epoch 13, batch 550, loss[loss=0.1704, simple_loss=0.2664, pruned_loss=0.03719, over 12302.00 frames. ], tot_loss[loss=0.1697, simple_loss=0.2592, pruned_loss=0.04013, over 2237133.67 frames. ], batch size: 34, lr: 3.73e-03, grad_scale: 8.0 2023-05-16 21:19:16,748 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.011e+02 2.711e+02 3.240e+02 3.910e+02 6.136e+02, threshold=6.479e+02, percent-clipped=0.0 2023-05-16 21:19:19,815 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=247379.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 21:19:21,118 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=247381.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 21:19:40,650 INFO [finetune.py:992] (0/2) Epoch 13, batch 600, loss[loss=0.1607, simple_loss=0.2482, pruned_loss=0.0366, over 12340.00 frames. ], tot_loss[loss=0.1696, simple_loss=0.2586, pruned_loss=0.0403, over 2263802.58 frames. ], batch size: 30, lr: 3.73e-03, grad_scale: 8.0 2023-05-16 21:19:53,568 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=247426.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 21:19:54,513 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 2023-05-16 21:20:03,346 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=247440.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 21:20:15,701 INFO [finetune.py:992] (0/2) Epoch 13, batch 650, loss[loss=0.1455, simple_loss=0.2295, pruned_loss=0.03075, over 12176.00 frames. ], tot_loss[loss=0.169, simple_loss=0.2583, pruned_loss=0.03985, over 2293303.99 frames. ], batch size: 31, lr: 3.73e-03, grad_scale: 8.0 2023-05-16 21:20:21,766 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.4723, 2.6677, 3.2014, 4.3689, 2.0870, 4.3132, 4.5324, 4.5496], device='cuda:0'), covar=tensor([0.0141, 0.1334, 0.0545, 0.0191, 0.1691, 0.0310, 0.0128, 0.0142], device='cuda:0'), in_proj_covar=tensor([0.0120, 0.0200, 0.0180, 0.0116, 0.0187, 0.0176, 0.0171, 0.0118], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 21:20:27,742 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.951e+02 2.699e+02 3.159e+02 3.694e+02 7.659e+02, threshold=6.318e+02, percent-clipped=2.0 2023-05-16 21:20:37,257 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=247487.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 21:20:39,403 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.7907, 2.2981, 3.2937, 3.7320, 3.5286, 3.6748, 3.2903, 2.8191], device='cuda:0'), covar=tensor([0.0052, 0.0393, 0.0132, 0.0047, 0.0103, 0.0096, 0.0123, 0.0325], device='cuda:0'), in_proj_covar=tensor([0.0087, 0.0123, 0.0102, 0.0076, 0.0101, 0.0114, 0.0095, 0.0137], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 21:20:47,248 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=247500.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 21:20:52,801 INFO [finetune.py:992] (0/2) Epoch 13, batch 700, loss[loss=0.1823, simple_loss=0.276, pruned_loss=0.04425, over 12296.00 frames. ], tot_loss[loss=0.1681, simple_loss=0.2578, pruned_loss=0.03924, over 2310084.41 frames. ], batch size: 34, lr: 3.73e-03, grad_scale: 8.0 2023-05-16 21:21:05,286 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=247525.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 21:21:25,697 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=247554.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 21:21:28,431 INFO [finetune.py:992] (0/2) Epoch 13, batch 750, loss[loss=0.18, simple_loss=0.2695, pruned_loss=0.04528, over 12355.00 frames. ], tot_loss[loss=0.1675, simple_loss=0.2572, pruned_loss=0.03892, over 2323660.59 frames. ], batch size: 36, lr: 3.73e-03, grad_scale: 8.0 2023-05-16 21:21:30,747 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=247561.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 21:21:40,618 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.002e+02 2.608e+02 3.056e+02 3.668e+02 6.856e+02, threshold=6.112e+02, percent-clipped=0.0 2023-05-16 21:21:48,699 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=247586.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 21:21:51,548 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=247590.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 21:21:54,619 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.7398, 3.4878, 5.1405, 2.8116, 2.9143, 3.8392, 3.2595, 3.8248], device='cuda:0'), covar=tensor([0.0401, 0.1131, 0.0334, 0.1226, 0.2040, 0.1492, 0.1383, 0.1172], device='cuda:0'), in_proj_covar=tensor([0.0237, 0.0237, 0.0247, 0.0185, 0.0239, 0.0293, 0.0224, 0.0264], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-16 21:22:04,528 INFO [finetune.py:992] (0/2) Epoch 13, batch 800, loss[loss=0.1549, simple_loss=0.2448, pruned_loss=0.03253, over 12087.00 frames. ], tot_loss[loss=0.1674, simple_loss=0.2572, pruned_loss=0.03883, over 2340753.77 frames. ], batch size: 32, lr: 3.73e-03, grad_scale: 8.0 2023-05-16 21:22:07,511 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.7782, 3.3639, 5.2173, 2.7668, 2.8278, 3.7785, 3.1701, 3.8032], device='cuda:0'), covar=tensor([0.0455, 0.1253, 0.0259, 0.1247, 0.2075, 0.1696, 0.1566, 0.1295], device='cuda:0'), in_proj_covar=tensor([0.0237, 0.0237, 0.0247, 0.0185, 0.0239, 0.0293, 0.0224, 0.0264], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-16 21:22:09,625 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=247615.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 21:22:26,573 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=247638.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 21:22:31,707 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.8469, 4.7361, 4.6405, 4.6654, 4.3999, 4.8761, 4.8281, 5.0880], device='cuda:0'), covar=tensor([0.0241, 0.0184, 0.0209, 0.0332, 0.0831, 0.0290, 0.0179, 0.0177], device='cuda:0'), in_proj_covar=tensor([0.0188, 0.0193, 0.0186, 0.0242, 0.0236, 0.0212, 0.0174, 0.0220], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0003, 0.0005, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:0') 2023-05-16 21:22:41,666 INFO [finetune.py:992] (0/2) Epoch 13, batch 850, loss[loss=0.1685, simple_loss=0.2527, pruned_loss=0.04221, over 12002.00 frames. ], tot_loss[loss=0.1668, simple_loss=0.2563, pruned_loss=0.03866, over 2357609.79 frames. ], batch size: 40, lr: 3.73e-03, grad_scale: 8.0 2023-05-16 21:22:53,629 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.987e+02 2.684e+02 3.226e+02 3.703e+02 6.597e+02, threshold=6.452e+02, percent-clipped=2.0 2023-05-16 21:23:16,902 INFO [finetune.py:992] (0/2) Epoch 13, batch 900, loss[loss=0.1642, simple_loss=0.2742, pruned_loss=0.02715, over 12170.00 frames. ], tot_loss[loss=0.1665, simple_loss=0.2556, pruned_loss=0.03865, over 2362756.65 frames. ], batch size: 35, lr: 3.73e-03, grad_scale: 8.0 2023-05-16 21:23:36,050 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=247735.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 21:23:52,232 INFO [finetune.py:992] (0/2) Epoch 13, batch 950, loss[loss=0.1873, simple_loss=0.2862, pruned_loss=0.04418, over 12350.00 frames. ], tot_loss[loss=0.1664, simple_loss=0.2552, pruned_loss=0.03879, over 2372990.67 frames. ], batch size: 35, lr: 3.73e-03, grad_scale: 8.0 2023-05-16 21:24:05,177 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.622e+02 2.716e+02 3.279e+02 3.877e+02 9.094e+02, threshold=6.558e+02, percent-clipped=4.0 2023-05-16 21:24:10,100 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=247782.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 21:24:11,540 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=247784.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 21:24:13,546 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.6770, 2.8265, 3.8581, 4.5381, 4.1142, 4.5705, 3.8121, 3.3878], device='cuda:0'), covar=tensor([0.0038, 0.0361, 0.0126, 0.0042, 0.0108, 0.0067, 0.0157, 0.0323], device='cuda:0'), in_proj_covar=tensor([0.0087, 0.0123, 0.0101, 0.0077, 0.0102, 0.0114, 0.0095, 0.0137], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 21:24:25,573 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.5417, 5.1450, 5.4751, 4.7385, 5.0809, 4.9304, 5.4872, 5.1933], device='cuda:0'), covar=tensor([0.0313, 0.0384, 0.0356, 0.0292, 0.0412, 0.0335, 0.0307, 0.0231], device='cuda:0'), in_proj_covar=tensor([0.0255, 0.0256, 0.0281, 0.0254, 0.0256, 0.0258, 0.0232, 0.0209], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 21:24:26,406 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.9501, 3.0721, 4.8099, 4.9862, 3.2274, 2.7011, 3.1080, 2.2630], device='cuda:0'), covar=tensor([0.1492, 0.2888, 0.0449, 0.0394, 0.1188, 0.2443, 0.2649, 0.4063], device='cuda:0'), in_proj_covar=tensor([0.0291, 0.0381, 0.0269, 0.0294, 0.0267, 0.0303, 0.0379, 0.0373], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-16 21:24:29,069 INFO [finetune.py:992] (0/2) Epoch 13, batch 1000, loss[loss=0.1533, simple_loss=0.2467, pruned_loss=0.02996, over 12193.00 frames. ], tot_loss[loss=0.1669, simple_loss=0.2557, pruned_loss=0.03906, over 2372507.68 frames. ], batch size: 35, lr: 3.73e-03, grad_scale: 8.0 2023-05-16 21:24:43,634 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.5351, 5.3282, 5.4517, 5.4897, 5.1271, 5.1532, 4.8868, 5.4410], device='cuda:0'), covar=tensor([0.0659, 0.0609, 0.0779, 0.0505, 0.1869, 0.1303, 0.0531, 0.0978], device='cuda:0'), in_proj_covar=tensor([0.0542, 0.0695, 0.0609, 0.0623, 0.0832, 0.0736, 0.0547, 0.0480], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0005, 0.0005, 0.0003, 0.0003], device='cuda:0') 2023-05-16 21:24:55,885 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=247845.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 21:25:00,447 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=4.58 vs. limit=5.0 2023-05-16 21:25:03,484 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=247856.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 21:25:04,892 INFO [finetune.py:992] (0/2) Epoch 13, batch 1050, loss[loss=0.1712, simple_loss=0.262, pruned_loss=0.04014, over 11604.00 frames. ], tot_loss[loss=0.1673, simple_loss=0.2562, pruned_loss=0.03918, over 2368714.77 frames. ], batch size: 48, lr: 3.73e-03, grad_scale: 8.0 2023-05-16 21:25:16,975 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.962e+02 2.693e+02 3.221e+02 3.799e+02 6.659e+02, threshold=6.442e+02, percent-clipped=1.0 2023-05-16 21:25:21,262 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=247881.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 21:25:40,642 INFO [finetune.py:992] (0/2) Epoch 13, batch 1100, loss[loss=0.1525, simple_loss=0.2333, pruned_loss=0.03589, over 11837.00 frames. ], tot_loss[loss=0.1682, simple_loss=0.2575, pruned_loss=0.03945, over 2374181.69 frames. ], batch size: 26, lr: 3.73e-03, grad_scale: 8.0 2023-05-16 21:25:42,207 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=247910.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 21:26:17,640 INFO [finetune.py:992] (0/2) Epoch 13, batch 1150, loss[loss=0.1417, simple_loss=0.2326, pruned_loss=0.02537, over 12119.00 frames. ], tot_loss[loss=0.1677, simple_loss=0.2571, pruned_loss=0.03914, over 2382105.18 frames. ], batch size: 30, lr: 3.73e-03, grad_scale: 8.0 2023-05-16 21:26:19,127 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.1704, 4.9698, 5.1034, 5.1461, 4.7713, 4.8310, 4.5286, 5.0868], device='cuda:0'), covar=tensor([0.0776, 0.0671, 0.0875, 0.0536, 0.2022, 0.1360, 0.0612, 0.1034], device='cuda:0'), in_proj_covar=tensor([0.0544, 0.0699, 0.0611, 0.0627, 0.0838, 0.0736, 0.0549, 0.0483], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0005, 0.0005, 0.0003, 0.0003], device='cuda:0') 2023-05-16 21:26:29,596 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.864e+02 2.775e+02 3.254e+02 3.807e+02 5.397e+02, threshold=6.508e+02, percent-clipped=0.0 2023-05-16 21:26:36,372 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=247984.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 21:26:45,059 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=247996.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 21:26:48,065 INFO [checkpoint.py:75] (0/2) Saving checkpoint to pruned_transducer_stateless7/exp_giga_finetune/checkpoint-148000.pt 2023-05-16 21:26:56,675 INFO [finetune.py:992] (0/2) Epoch 13, batch 1200, loss[loss=0.1668, simple_loss=0.2589, pruned_loss=0.03736, over 11815.00 frames. ], tot_loss[loss=0.1674, simple_loss=0.2565, pruned_loss=0.03911, over 2374347.59 frames. ], batch size: 44, lr: 3.73e-03, grad_scale: 8.0 2023-05-16 21:27:16,161 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=248035.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 21:27:22,692 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.4035, 4.7824, 3.9519, 5.0283, 4.5849, 2.6741, 4.1478, 2.9410], device='cuda:0'), covar=tensor([0.0695, 0.0695, 0.1472, 0.0493, 0.1092, 0.1776, 0.1131, 0.3379], device='cuda:0'), in_proj_covar=tensor([0.0302, 0.0377, 0.0355, 0.0307, 0.0365, 0.0271, 0.0342, 0.0366], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 21:27:23,360 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=248045.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 21:27:31,142 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.7366, 2.5211, 3.2258, 4.5521, 2.4311, 4.4307, 4.6459, 4.7388], device='cuda:0'), covar=tensor([0.0089, 0.1395, 0.0505, 0.0138, 0.1377, 0.0330, 0.0151, 0.0102], device='cuda:0'), in_proj_covar=tensor([0.0120, 0.0201, 0.0180, 0.0116, 0.0187, 0.0176, 0.0172, 0.0120], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 21:27:31,906 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=248057.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 21:27:32,429 INFO [finetune.py:992] (0/2) Epoch 13, batch 1250, loss[loss=0.1488, simple_loss=0.2347, pruned_loss=0.03151, over 12303.00 frames. ], tot_loss[loss=0.1673, simple_loss=0.2565, pruned_loss=0.03902, over 2378306.56 frames. ], batch size: 28, lr: 3.73e-03, grad_scale: 8.0 2023-05-16 21:27:43,095 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 2023-05-16 21:27:45,373 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.871e+02 2.704e+02 3.068e+02 3.784e+02 5.230e+02, threshold=6.136e+02, percent-clipped=0.0 2023-05-16 21:27:50,556 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=248082.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 21:27:50,628 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=248082.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 21:27:51,245 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=248083.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 21:28:09,360 INFO [finetune.py:992] (0/2) Epoch 13, batch 1300, loss[loss=0.1763, simple_loss=0.264, pruned_loss=0.04431, over 12159.00 frames. ], tot_loss[loss=0.1674, simple_loss=0.257, pruned_loss=0.03892, over 2381692.94 frames. ], batch size: 34, lr: 3.73e-03, grad_scale: 8.0 2023-05-16 21:28:12,922 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.83 vs. limit=2.0 2023-05-16 21:28:25,283 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=248130.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 21:28:31,096 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.9235, 3.5163, 5.2083, 2.7555, 2.8529, 3.8700, 3.2823, 3.8293], device='cuda:0'), covar=tensor([0.0391, 0.1088, 0.0294, 0.1299, 0.2049, 0.1635, 0.1466, 0.1266], device='cuda:0'), in_proj_covar=tensor([0.0239, 0.0237, 0.0250, 0.0186, 0.0240, 0.0297, 0.0226, 0.0265], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-16 21:28:32,434 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=248140.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 21:28:34,680 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=248143.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 21:28:43,929 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=248156.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 21:28:45,182 INFO [finetune.py:992] (0/2) Epoch 13, batch 1350, loss[loss=0.1651, simple_loss=0.2636, pruned_loss=0.03327, over 12009.00 frames. ], tot_loss[loss=0.1668, simple_loss=0.2563, pruned_loss=0.03868, over 2385170.22 frames. ], batch size: 40, lr: 3.73e-03, grad_scale: 8.0 2023-05-16 21:28:52,642 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.5626, 4.9008, 3.1505, 2.8687, 4.1008, 2.6121, 4.1585, 3.4282], device='cuda:0'), covar=tensor([0.0615, 0.0516, 0.1093, 0.1457, 0.0303, 0.1415, 0.0422, 0.0778], device='cuda:0'), in_proj_covar=tensor([0.0187, 0.0253, 0.0176, 0.0199, 0.0141, 0.0182, 0.0196, 0.0175], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 21:28:57,392 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.630e+02 2.440e+02 2.860e+02 3.425e+02 6.326e+02, threshold=5.721e+02, percent-clipped=1.0 2023-05-16 21:29:01,728 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=248181.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 21:29:18,489 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=248204.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 21:29:21,282 INFO [finetune.py:992] (0/2) Epoch 13, batch 1400, loss[loss=0.181, simple_loss=0.2682, pruned_loss=0.04684, over 11602.00 frames. ], tot_loss[loss=0.1662, simple_loss=0.2559, pruned_loss=0.03831, over 2389478.68 frames. ], batch size: 48, lr: 3.73e-03, grad_scale: 8.0 2023-05-16 21:29:22,786 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=248210.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 21:29:36,814 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=248229.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 21:29:58,109 INFO [finetune.py:992] (0/2) Epoch 13, batch 1450, loss[loss=0.138, simple_loss=0.2248, pruned_loss=0.02558, over 11385.00 frames. ], tot_loss[loss=0.1675, simple_loss=0.2573, pruned_loss=0.03892, over 2392136.77 frames. ], batch size: 25, lr: 3.73e-03, grad_scale: 8.0 2023-05-16 21:29:58,172 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=248258.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 21:30:09,063 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.0086, 4.6560, 4.7853, 4.8993, 4.6964, 4.9396, 4.7675, 2.5745], device='cuda:0'), covar=tensor([0.0095, 0.0081, 0.0098, 0.0060, 0.0062, 0.0087, 0.0131, 0.0813], device='cuda:0'), in_proj_covar=tensor([0.0068, 0.0079, 0.0082, 0.0073, 0.0060, 0.0092, 0.0082, 0.0097], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-16 21:30:10,280 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.666e+02 2.647e+02 2.996e+02 3.630e+02 5.731e+02, threshold=5.992e+02, percent-clipped=1.0 2023-05-16 21:30:25,612 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.2909, 4.4621, 2.8002, 2.5169, 3.8521, 2.4869, 3.8470, 3.0255], device='cuda:0'), covar=tensor([0.0759, 0.0576, 0.1187, 0.1631, 0.0293, 0.1495, 0.0520, 0.0933], device='cuda:0'), in_proj_covar=tensor([0.0188, 0.0254, 0.0176, 0.0199, 0.0140, 0.0181, 0.0195, 0.0175], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 21:30:34,007 INFO [finetune.py:992] (0/2) Epoch 13, batch 1500, loss[loss=0.1395, simple_loss=0.2254, pruned_loss=0.02677, over 12134.00 frames. ], tot_loss[loss=0.1673, simple_loss=0.2568, pruned_loss=0.03892, over 2391518.08 frames. ], batch size: 30, lr: 3.73e-03, grad_scale: 8.0 2023-05-16 21:30:51,311 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=248332.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 21:30:57,006 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=248340.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 21:31:05,606 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=248352.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 21:31:09,756 INFO [finetune.py:992] (0/2) Epoch 13, batch 1550, loss[loss=0.1708, simple_loss=0.2639, pruned_loss=0.03885, over 11260.00 frames. ], tot_loss[loss=0.1676, simple_loss=0.257, pruned_loss=0.03906, over 2376592.87 frames. ], batch size: 55, lr: 3.72e-03, grad_scale: 8.0 2023-05-16 21:31:22,534 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.091e+02 2.825e+02 3.179e+02 3.763e+02 8.904e+02, threshold=6.357e+02, percent-clipped=2.0 2023-05-16 21:31:28,514 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=248383.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 21:31:36,342 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=248393.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 21:31:43,413 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.3279, 5.0877, 5.2471, 5.3049, 4.8786, 4.9326, 4.6697, 5.1908], device='cuda:0'), covar=tensor([0.0712, 0.0724, 0.0917, 0.0607, 0.2280, 0.1404, 0.0611, 0.1153], device='cuda:0'), in_proj_covar=tensor([0.0545, 0.0695, 0.0611, 0.0626, 0.0840, 0.0738, 0.0548, 0.0482], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0005, 0.0005, 0.0003, 0.0003], device='cuda:0') 2023-05-16 21:31:46,851 INFO [finetune.py:992] (0/2) Epoch 13, batch 1600, loss[loss=0.1494, simple_loss=0.2385, pruned_loss=0.03016, over 12105.00 frames. ], tot_loss[loss=0.1678, simple_loss=0.2573, pruned_loss=0.03913, over 2376266.06 frames. ], batch size: 30, lr: 3.72e-03, grad_scale: 8.0 2023-05-16 21:32:08,651 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=248438.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 21:32:08,974 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.40 vs. limit=2.0 2023-05-16 21:32:10,091 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=248440.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 21:32:13,088 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=248444.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 21:32:22,779 INFO [finetune.py:992] (0/2) Epoch 13, batch 1650, loss[loss=0.1616, simple_loss=0.2552, pruned_loss=0.03401, over 12157.00 frames. ], tot_loss[loss=0.1684, simple_loss=0.2581, pruned_loss=0.03934, over 2369612.01 frames. ], batch size: 34, lr: 3.72e-03, grad_scale: 8.0 2023-05-16 21:32:35,086 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.963e+02 2.657e+02 3.152e+02 3.883e+02 1.000e+03, threshold=6.303e+02, percent-clipped=1.0 2023-05-16 21:32:44,440 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=248488.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 21:32:58,425 INFO [finetune.py:992] (0/2) Epoch 13, batch 1700, loss[loss=0.1885, simple_loss=0.2771, pruned_loss=0.04993, over 12355.00 frames. ], tot_loss[loss=0.1681, simple_loss=0.2578, pruned_loss=0.03922, over 2380126.51 frames. ], batch size: 38, lr: 3.72e-03, grad_scale: 8.0 2023-05-16 21:33:02,444 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.36 vs. limit=2.0 2023-05-16 21:33:09,870 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.3916, 3.4464, 3.2929, 3.1474, 2.8609, 2.6578, 3.5514, 2.3578], device='cuda:0'), covar=tensor([0.0450, 0.0191, 0.0216, 0.0213, 0.0415, 0.0399, 0.0150, 0.0517], device='cuda:0'), in_proj_covar=tensor([0.0191, 0.0159, 0.0161, 0.0188, 0.0202, 0.0199, 0.0170, 0.0203], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 21:33:14,041 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.0455, 2.1700, 2.6237, 3.1084, 2.1678, 3.0959, 3.0643, 3.2314], device='cuda:0'), covar=tensor([0.0180, 0.1074, 0.0518, 0.0194, 0.1140, 0.0322, 0.0335, 0.0146], device='cuda:0'), in_proj_covar=tensor([0.0121, 0.0201, 0.0181, 0.0116, 0.0187, 0.0176, 0.0172, 0.0119], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 21:33:28,318 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.6182, 2.6610, 3.2296, 4.4831, 2.4794, 4.4593, 4.5925, 4.6662], device='cuda:0'), covar=tensor([0.0117, 0.1228, 0.0472, 0.0109, 0.1244, 0.0228, 0.0134, 0.0094], device='cuda:0'), in_proj_covar=tensor([0.0120, 0.0200, 0.0181, 0.0116, 0.0187, 0.0176, 0.0172, 0.0119], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 21:33:35,239 INFO [finetune.py:992] (0/2) Epoch 13, batch 1750, loss[loss=0.1838, simple_loss=0.2743, pruned_loss=0.04664, over 11747.00 frames. ], tot_loss[loss=0.1677, simple_loss=0.2573, pruned_loss=0.03901, over 2378333.29 frames. ], batch size: 44, lr: 3.72e-03, grad_scale: 8.0 2023-05-16 21:33:47,383 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.720e+02 2.588e+02 3.157e+02 3.850e+02 6.228e+02, threshold=6.313e+02, percent-clipped=0.0 2023-05-16 21:34:10,938 INFO [finetune.py:992] (0/2) Epoch 13, batch 1800, loss[loss=0.1729, simple_loss=0.2667, pruned_loss=0.0396, over 12290.00 frames. ], tot_loss[loss=0.1676, simple_loss=0.2573, pruned_loss=0.03901, over 2372357.18 frames. ], batch size: 33, lr: 3.72e-03, grad_scale: 8.0 2023-05-16 21:34:33,754 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=248640.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 21:34:42,275 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=248652.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 21:34:46,422 INFO [finetune.py:992] (0/2) Epoch 13, batch 1850, loss[loss=0.1415, simple_loss=0.2239, pruned_loss=0.02961, over 11866.00 frames. ], tot_loss[loss=0.1665, simple_loss=0.2558, pruned_loss=0.03855, over 2381052.81 frames. ], batch size: 26, lr: 3.72e-03, grad_scale: 8.0 2023-05-16 21:34:59,278 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.539e+02 2.662e+02 3.210e+02 3.660e+02 5.824e+02, threshold=6.421e+02, percent-clipped=0.0 2023-05-16 21:35:08,719 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=248688.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 21:35:08,734 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=248688.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 21:35:10,970 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=248691.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 21:35:17,802 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=248700.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 21:35:19,392 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.3398, 2.8945, 3.8402, 3.2271, 3.5162, 3.3560, 2.7289, 3.6667], device='cuda:0'), covar=tensor([0.0133, 0.0302, 0.0133, 0.0225, 0.0183, 0.0185, 0.0338, 0.0135], device='cuda:0'), in_proj_covar=tensor([0.0175, 0.0202, 0.0184, 0.0184, 0.0214, 0.0162, 0.0197, 0.0185], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 21:35:23,540 INFO [finetune.py:992] (0/2) Epoch 13, batch 1900, loss[loss=0.1504, simple_loss=0.2352, pruned_loss=0.03278, over 12339.00 frames. ], tot_loss[loss=0.1664, simple_loss=0.2558, pruned_loss=0.03846, over 2379849.14 frames. ], batch size: 30, lr: 3.72e-03, grad_scale: 8.0 2023-05-16 21:35:28,710 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=248715.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 21:35:34,487 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.0697, 3.5882, 5.3488, 3.0075, 2.9632, 3.9840, 3.3312, 4.0264], device='cuda:0'), covar=tensor([0.0407, 0.1088, 0.0354, 0.1130, 0.2004, 0.1587, 0.1413, 0.1124], device='cuda:0'), in_proj_covar=tensor([0.0242, 0.0240, 0.0253, 0.0187, 0.0243, 0.0299, 0.0228, 0.0267], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-16 21:35:44,907 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=248738.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 21:35:45,590 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=248739.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 21:35:54,892 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=248752.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 21:35:58,879 INFO [finetune.py:992] (0/2) Epoch 13, batch 1950, loss[loss=0.1697, simple_loss=0.2561, pruned_loss=0.04162, over 11868.00 frames. ], tot_loss[loss=0.1665, simple_loss=0.2561, pruned_loss=0.03846, over 2381790.19 frames. ], batch size: 44, lr: 3.72e-03, grad_scale: 8.0 2023-05-16 21:36:08,344 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.5668, 4.2817, 4.3427, 4.4874, 4.3562, 4.5116, 4.3810, 2.6220], device='cuda:0'), covar=tensor([0.0111, 0.0081, 0.0105, 0.0076, 0.0057, 0.0105, 0.0093, 0.0769], device='cuda:0'), in_proj_covar=tensor([0.0068, 0.0078, 0.0081, 0.0073, 0.0059, 0.0092, 0.0082, 0.0096], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-16 21:36:11,034 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.694e+02 2.728e+02 3.157e+02 3.801e+02 8.869e+02, threshold=6.314e+02, percent-clipped=3.0 2023-05-16 21:36:11,979 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=248776.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 21:36:18,858 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=248786.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 21:36:30,725 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.5942, 2.5703, 3.5502, 4.4722, 4.0014, 4.4682, 3.6800, 3.3428], device='cuda:0'), covar=tensor([0.0039, 0.0426, 0.0158, 0.0055, 0.0131, 0.0077, 0.0154, 0.0328], device='cuda:0'), in_proj_covar=tensor([0.0087, 0.0123, 0.0102, 0.0077, 0.0102, 0.0114, 0.0095, 0.0137], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 21:36:31,552 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.7099, 2.9543, 4.6605, 4.6599, 2.8953, 2.5679, 3.0616, 2.1889], device='cuda:0'), covar=tensor([0.1627, 0.3100, 0.0416, 0.0454, 0.1343, 0.2483, 0.2618, 0.3910], device='cuda:0'), in_proj_covar=tensor([0.0297, 0.0389, 0.0274, 0.0300, 0.0271, 0.0308, 0.0385, 0.0378], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-16 21:36:34,924 INFO [finetune.py:992] (0/2) Epoch 13, batch 2000, loss[loss=0.231, simple_loss=0.3052, pruned_loss=0.07838, over 8238.00 frames. ], tot_loss[loss=0.1662, simple_loss=0.2557, pruned_loss=0.03836, over 2380486.77 frames. ], batch size: 98, lr: 3.72e-03, grad_scale: 16.0 2023-05-16 21:37:11,891 INFO [finetune.py:992] (0/2) Epoch 13, batch 2050, loss[loss=0.1607, simple_loss=0.2511, pruned_loss=0.03519, over 12271.00 frames. ], tot_loss[loss=0.167, simple_loss=0.2561, pruned_loss=0.03901, over 2377580.00 frames. ], batch size: 37, lr: 3.72e-03, grad_scale: 16.0 2023-05-16 21:37:24,171 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.873e+02 2.825e+02 3.221e+02 3.921e+02 8.647e+02, threshold=6.442e+02, percent-clipped=3.0 2023-05-16 21:37:47,437 INFO [finetune.py:992] (0/2) Epoch 13, batch 2100, loss[loss=0.1512, simple_loss=0.2477, pruned_loss=0.02731, over 12341.00 frames. ], tot_loss[loss=0.1674, simple_loss=0.2566, pruned_loss=0.03909, over 2372383.14 frames. ], batch size: 36, lr: 3.72e-03, grad_scale: 16.0 2023-05-16 21:37:55,106 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.39 vs. limit=2.0 2023-05-16 21:38:23,495 INFO [finetune.py:992] (0/2) Epoch 13, batch 2150, loss[loss=0.1498, simple_loss=0.2385, pruned_loss=0.03048, over 12246.00 frames. ], tot_loss[loss=0.1678, simple_loss=0.2569, pruned_loss=0.03936, over 2373522.58 frames. ], batch size: 32, lr: 3.72e-03, grad_scale: 16.0 2023-05-16 21:38:27,566 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=4.93 vs. limit=5.0 2023-05-16 21:38:35,674 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.922e+02 2.706e+02 3.066e+02 3.846e+02 6.335e+02, threshold=6.132e+02, percent-clipped=0.0 2023-05-16 21:38:35,950 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.1617, 3.9598, 3.9962, 4.4615, 2.7656, 3.9252, 2.6024, 4.0560], device='cuda:0'), covar=tensor([0.1701, 0.0777, 0.0941, 0.0635, 0.1289, 0.0623, 0.1895, 0.1225], device='cuda:0'), in_proj_covar=tensor([0.0232, 0.0264, 0.0293, 0.0360, 0.0241, 0.0244, 0.0262, 0.0380], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 21:38:45,890 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=248988.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 21:38:57,750 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=3.97 vs. limit=5.0 2023-05-16 21:38:58,923 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.9759, 5.9617, 5.7131, 5.2630, 5.0476, 5.7841, 5.4116, 5.2194], device='cuda:0'), covar=tensor([0.0695, 0.0854, 0.0723, 0.1483, 0.0782, 0.0776, 0.1635, 0.1203], device='cuda:0'), in_proj_covar=tensor([0.0625, 0.0563, 0.0523, 0.0638, 0.0417, 0.0723, 0.0783, 0.0577], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0004, 0.0004, 0.0003], device='cuda:0') 2023-05-16 21:39:00,271 INFO [finetune.py:992] (0/2) Epoch 13, batch 2200, loss[loss=0.1538, simple_loss=0.2348, pruned_loss=0.03644, over 12163.00 frames. ], tot_loss[loss=0.1676, simple_loss=0.2566, pruned_loss=0.03934, over 2367163.04 frames. ], batch size: 29, lr: 3.72e-03, grad_scale: 16.0 2023-05-16 21:39:08,436 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.4117, 2.2196, 3.1636, 4.1882, 2.2388, 4.3221, 4.4830, 4.4516], device='cuda:0'), covar=tensor([0.0158, 0.1399, 0.0528, 0.0192, 0.1373, 0.0254, 0.0145, 0.0129], device='cuda:0'), in_proj_covar=tensor([0.0121, 0.0202, 0.0183, 0.0118, 0.0188, 0.0178, 0.0173, 0.0120], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 21:39:11,393 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.6416, 2.8005, 4.6225, 4.6568, 2.7309, 2.6142, 2.8321, 2.1360], device='cuda:0'), covar=tensor([0.1709, 0.3257, 0.0463, 0.0516, 0.1513, 0.2545, 0.3039, 0.4345], device='cuda:0'), in_proj_covar=tensor([0.0298, 0.0391, 0.0276, 0.0301, 0.0272, 0.0309, 0.0387, 0.0379], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-16 21:39:20,424 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=249036.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 21:39:22,616 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=249039.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 21:39:28,323 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=249047.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 21:39:32,618 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.0820, 4.8989, 4.8355, 4.8952, 4.3134, 5.1038, 5.0424, 5.2221], device='cuda:0'), covar=tensor([0.0189, 0.0176, 0.0194, 0.0348, 0.1030, 0.0279, 0.0172, 0.0199], device='cuda:0'), in_proj_covar=tensor([0.0193, 0.0197, 0.0190, 0.0247, 0.0241, 0.0219, 0.0178, 0.0225], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0005, 0.0004, 0.0004, 0.0003, 0.0005], device='cuda:0') 2023-05-16 21:39:35,911 INFO [finetune.py:992] (0/2) Epoch 13, batch 2250, loss[loss=0.1779, simple_loss=0.2679, pruned_loss=0.04395, over 12151.00 frames. ], tot_loss[loss=0.1675, simple_loss=0.2564, pruned_loss=0.0393, over 2370037.54 frames. ], batch size: 39, lr: 3.72e-03, grad_scale: 16.0 2023-05-16 21:39:45,072 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=249071.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 21:39:47,832 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.903e+02 2.541e+02 3.006e+02 3.631e+02 8.139e+02, threshold=6.012e+02, percent-clipped=2.0 2023-05-16 21:39:56,337 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=249087.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 21:40:11,563 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.9648, 5.9832, 5.6875, 5.2174, 5.0566, 5.8195, 5.3826, 5.2184], device='cuda:0'), covar=tensor([0.0867, 0.0881, 0.0846, 0.1546, 0.0755, 0.0828, 0.2003, 0.1240], device='cuda:0'), in_proj_covar=tensor([0.0625, 0.0560, 0.0522, 0.0636, 0.0417, 0.0721, 0.0783, 0.0576], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0004, 0.0004, 0.0003], device='cuda:0') 2023-05-16 21:40:12,163 INFO [finetune.py:992] (0/2) Epoch 13, batch 2300, loss[loss=0.1702, simple_loss=0.2675, pruned_loss=0.03643, over 12150.00 frames. ], tot_loss[loss=0.1673, simple_loss=0.2562, pruned_loss=0.03919, over 2371536.93 frames. ], batch size: 36, lr: 3.72e-03, grad_scale: 16.0 2023-05-16 21:40:25,121 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=2.57 vs. limit=5.0 2023-05-16 21:40:48,114 INFO [finetune.py:992] (0/2) Epoch 13, batch 2350, loss[loss=0.1555, simple_loss=0.2407, pruned_loss=0.03517, over 12344.00 frames. ], tot_loss[loss=0.1673, simple_loss=0.2564, pruned_loss=0.03907, over 2382185.53 frames. ], batch size: 31, lr: 3.72e-03, grad_scale: 16.0 2023-05-16 21:41:00,163 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.825e+02 2.710e+02 3.210e+02 3.953e+02 6.628e+02, threshold=6.419e+02, percent-clipped=3.0 2023-05-16 21:41:24,141 INFO [finetune.py:992] (0/2) Epoch 13, batch 2400, loss[loss=0.1535, simple_loss=0.2529, pruned_loss=0.02708, over 12303.00 frames. ], tot_loss[loss=0.1669, simple_loss=0.2561, pruned_loss=0.03879, over 2381892.87 frames. ], batch size: 34, lr: 3.72e-03, grad_scale: 16.0 2023-05-16 21:41:25,383 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.85 vs. limit=2.0 2023-05-16 21:41:29,265 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=249215.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 21:42:00,318 INFO [finetune.py:992] (0/2) Epoch 13, batch 2450, loss[loss=0.1681, simple_loss=0.2567, pruned_loss=0.03969, over 11845.00 frames. ], tot_loss[loss=0.1678, simple_loss=0.2571, pruned_loss=0.03928, over 2379820.60 frames. ], batch size: 44, lr: 3.72e-03, grad_scale: 16.0 2023-05-16 21:42:13,064 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.973e+02 2.619e+02 3.067e+02 3.670e+02 7.694e+02, threshold=6.133e+02, percent-clipped=1.0 2023-05-16 21:42:14,055 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=249276.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 21:42:14,741 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.3907, 5.1626, 5.2996, 5.3734, 4.9989, 5.0123, 4.7819, 5.2981], device='cuda:0'), covar=tensor([0.0596, 0.0609, 0.0768, 0.0552, 0.1852, 0.1282, 0.0556, 0.0933], device='cuda:0'), in_proj_covar=tensor([0.0554, 0.0712, 0.0625, 0.0644, 0.0856, 0.0758, 0.0564, 0.0491], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0005, 0.0004, 0.0004, 0.0005, 0.0005, 0.0003, 0.0003], device='cuda:0') 2023-05-16 21:42:36,734 INFO [finetune.py:992] (0/2) Epoch 13, batch 2500, loss[loss=0.1559, simple_loss=0.2547, pruned_loss=0.02854, over 12277.00 frames. ], tot_loss[loss=0.1672, simple_loss=0.2563, pruned_loss=0.03902, over 2372603.26 frames. ], batch size: 33, lr: 3.72e-03, grad_scale: 16.0 2023-05-16 21:42:43,490 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.36 vs. limit=2.0 2023-05-16 21:42:59,864 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.3342, 4.6993, 2.9479, 2.7666, 4.0638, 2.5182, 3.9075, 3.2449], device='cuda:0'), covar=tensor([0.0716, 0.0505, 0.1142, 0.1421, 0.0327, 0.1407, 0.0522, 0.0792], device='cuda:0'), in_proj_covar=tensor([0.0190, 0.0257, 0.0177, 0.0202, 0.0143, 0.0182, 0.0198, 0.0176], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 21:43:04,664 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=249347.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 21:43:12,507 INFO [finetune.py:992] (0/2) Epoch 13, batch 2550, loss[loss=0.1852, simple_loss=0.2819, pruned_loss=0.04427, over 12057.00 frames. ], tot_loss[loss=0.1671, simple_loss=0.2566, pruned_loss=0.03882, over 2372094.29 frames. ], batch size: 40, lr: 3.72e-03, grad_scale: 16.0 2023-05-16 21:43:21,878 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=249371.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 21:43:24,533 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.730e+02 2.612e+02 3.049e+02 3.624e+02 1.436e+03, threshold=6.098e+02, percent-clipped=1.0 2023-05-16 21:43:37,018 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.1981, 2.0549, 2.4067, 2.1747, 2.3080, 2.4049, 1.9115, 2.4177], device='cuda:0'), covar=tensor([0.0117, 0.0305, 0.0168, 0.0189, 0.0157, 0.0162, 0.0280, 0.0152], device='cuda:0'), in_proj_covar=tensor([0.0178, 0.0204, 0.0188, 0.0186, 0.0216, 0.0164, 0.0198, 0.0188], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 21:43:39,013 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=249395.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 21:43:46,209 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.82 vs. limit=2.0 2023-05-16 21:43:49,947 INFO [finetune.py:992] (0/2) Epoch 13, batch 2600, loss[loss=0.229, simple_loss=0.2997, pruned_loss=0.07916, over 8215.00 frames. ], tot_loss[loss=0.1665, simple_loss=0.2554, pruned_loss=0.03876, over 2359273.50 frames. ], batch size: 98, lr: 3.72e-03, grad_scale: 16.0 2023-05-16 21:43:57,710 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=249419.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 21:44:25,246 INFO [finetune.py:992] (0/2) Epoch 13, batch 2650, loss[loss=0.1649, simple_loss=0.2674, pruned_loss=0.03118, over 12349.00 frames. ], tot_loss[loss=0.1661, simple_loss=0.2554, pruned_loss=0.03845, over 2366936.34 frames. ], batch size: 36, lr: 3.72e-03, grad_scale: 16.0 2023-05-16 21:44:37,131 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.956e+02 2.791e+02 3.288e+02 3.836e+02 9.634e+02, threshold=6.576e+02, percent-clipped=1.0 2023-05-16 21:45:00,943 INFO [finetune.py:992] (0/2) Epoch 13, batch 2700, loss[loss=0.1482, simple_loss=0.2258, pruned_loss=0.03523, over 12196.00 frames. ], tot_loss[loss=0.1656, simple_loss=0.2545, pruned_loss=0.03837, over 2377592.78 frames. ], batch size: 29, lr: 3.71e-03, grad_scale: 16.0 2023-05-16 21:45:37,646 INFO [finetune.py:992] (0/2) Epoch 13, batch 2750, loss[loss=0.1671, simple_loss=0.2502, pruned_loss=0.042, over 12293.00 frames. ], tot_loss[loss=0.1654, simple_loss=0.254, pruned_loss=0.03838, over 2370461.47 frames. ], batch size: 33, lr: 3.71e-03, grad_scale: 16.0 2023-05-16 21:45:46,837 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=249571.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 21:45:49,608 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.027e+02 2.689e+02 3.252e+02 4.031e+02 1.393e+03, threshold=6.504e+02, percent-clipped=2.0 2023-05-16 21:46:13,560 INFO [finetune.py:992] (0/2) Epoch 13, batch 2800, loss[loss=0.1446, simple_loss=0.232, pruned_loss=0.02862, over 12179.00 frames. ], tot_loss[loss=0.1658, simple_loss=0.2546, pruned_loss=0.03853, over 2364955.43 frames. ], batch size: 31, lr: 3.71e-03, grad_scale: 16.0 2023-05-16 21:46:24,444 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.2101, 4.6602, 2.9571, 2.8528, 3.8678, 2.5620, 3.9252, 3.2525], device='cuda:0'), covar=tensor([0.0732, 0.0551, 0.1123, 0.1389, 0.0457, 0.1372, 0.0494, 0.0774], device='cuda:0'), in_proj_covar=tensor([0.0190, 0.0256, 0.0177, 0.0201, 0.0142, 0.0181, 0.0197, 0.0176], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 21:46:25,778 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.4447, 5.0375, 5.4482, 4.7607, 5.0489, 4.9365, 5.4901, 5.1723], device='cuda:0'), covar=tensor([0.0259, 0.0376, 0.0277, 0.0247, 0.0404, 0.0297, 0.0178, 0.0237], device='cuda:0'), in_proj_covar=tensor([0.0265, 0.0266, 0.0291, 0.0263, 0.0264, 0.0265, 0.0240, 0.0217], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 21:46:49,000 INFO [finetune.py:992] (0/2) Epoch 13, batch 2850, loss[loss=0.1698, simple_loss=0.2678, pruned_loss=0.03594, over 12385.00 frames. ], tot_loss[loss=0.1663, simple_loss=0.2554, pruned_loss=0.03863, over 2368295.01 frames. ], batch size: 38, lr: 3.71e-03, grad_scale: 16.0 2023-05-16 21:47:00,882 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.106e+02 2.639e+02 3.024e+02 3.770e+02 5.817e+02, threshold=6.047e+02, percent-clipped=0.0 2023-05-16 21:47:16,924 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.0390, 4.8454, 4.9642, 5.0211, 4.6764, 4.7358, 4.4970, 4.9180], device='cuda:0'), covar=tensor([0.0782, 0.0680, 0.0866, 0.0701, 0.1966, 0.1305, 0.0622, 0.1071], device='cuda:0'), in_proj_covar=tensor([0.0550, 0.0712, 0.0625, 0.0642, 0.0856, 0.0756, 0.0561, 0.0488], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0005, 0.0004, 0.0004, 0.0005, 0.0005, 0.0003, 0.0003], device='cuda:0') 2023-05-16 21:47:26,086 INFO [finetune.py:992] (0/2) Epoch 13, batch 2900, loss[loss=0.1504, simple_loss=0.2277, pruned_loss=0.03652, over 12340.00 frames. ], tot_loss[loss=0.1663, simple_loss=0.2556, pruned_loss=0.03848, over 2369971.88 frames. ], batch size: 31, lr: 3.71e-03, grad_scale: 16.0 2023-05-16 21:48:01,925 INFO [finetune.py:992] (0/2) Epoch 13, batch 2950, loss[loss=0.1411, simple_loss=0.2258, pruned_loss=0.02823, over 12188.00 frames. ], tot_loss[loss=0.1669, simple_loss=0.2562, pruned_loss=0.0388, over 2367167.38 frames. ], batch size: 31, lr: 3.71e-03, grad_scale: 16.0 2023-05-16 21:48:10,726 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.33 vs. limit=2.0 2023-05-16 21:48:13,837 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.042e+02 2.615e+02 3.019e+02 3.612e+02 8.318e+02, threshold=6.038e+02, percent-clipped=3.0 2023-05-16 21:48:37,092 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.4269, 5.0357, 5.3247, 4.6990, 5.1233, 4.8078, 5.3653, 5.1790], device='cuda:0'), covar=tensor([0.0364, 0.0495, 0.0620, 0.0330, 0.0384, 0.0398, 0.0374, 0.0267], device='cuda:0'), in_proj_covar=tensor([0.0263, 0.0266, 0.0290, 0.0262, 0.0262, 0.0264, 0.0238, 0.0217], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 21:48:37,664 INFO [finetune.py:992] (0/2) Epoch 13, batch 3000, loss[loss=0.166, simple_loss=0.252, pruned_loss=0.04001, over 12353.00 frames. ], tot_loss[loss=0.1665, simple_loss=0.2558, pruned_loss=0.0386, over 2369419.71 frames. ], batch size: 30, lr: 3.71e-03, grad_scale: 16.0 2023-05-16 21:48:37,665 INFO [finetune.py:1017] (0/2) Computing validation loss 2023-05-16 21:48:49,466 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.3402, 2.4394, 3.1657, 4.2524, 2.4047, 4.3363, 4.3981, 4.4065], device='cuda:0'), covar=tensor([0.0144, 0.1296, 0.0509, 0.0134, 0.1305, 0.0211, 0.0144, 0.0101], device='cuda:0'), in_proj_covar=tensor([0.0120, 0.0200, 0.0180, 0.0118, 0.0186, 0.0176, 0.0171, 0.0119], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 21:48:54,539 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.7996, 5.6793, 5.5953, 5.0032, 5.0307, 5.6710, 5.1994, 5.2334], device='cuda:0'), covar=tensor([0.0576, 0.0838, 0.0568, 0.1660, 0.0488, 0.0651, 0.1556, 0.0967], device='cuda:0'), in_proj_covar=tensor([0.0617, 0.0557, 0.0516, 0.0630, 0.0414, 0.0716, 0.0774, 0.0570], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-05-16 21:48:56,558 INFO [finetune.py:1026] (0/2) Epoch 13, validation: loss=0.3128, simple_loss=0.3917, pruned_loss=0.117, over 1020973.00 frames. 2023-05-16 21:48:56,558 INFO [finetune.py:1027] (0/2) Maximum memory allocated so far is 12827MB 2023-05-16 21:49:01,033 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=249814.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 21:49:07,282 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.8255, 5.7987, 5.5198, 5.1511, 5.0222, 5.7291, 5.3416, 5.1372], device='cuda:0'), covar=tensor([0.0738, 0.0896, 0.0748, 0.1479, 0.0834, 0.0702, 0.1544, 0.1148], device='cuda:0'), in_proj_covar=tensor([0.0619, 0.0558, 0.0517, 0.0631, 0.0415, 0.0716, 0.0776, 0.0571], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-05-16 21:49:15,243 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.2407, 2.6430, 3.8478, 3.2085, 3.6039, 3.3450, 2.5816, 3.6811], device='cuda:0'), covar=tensor([0.0143, 0.0364, 0.0115, 0.0223, 0.0153, 0.0173, 0.0382, 0.0133], device='cuda:0'), in_proj_covar=tensor([0.0179, 0.0204, 0.0187, 0.0187, 0.0217, 0.0164, 0.0197, 0.0188], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 21:49:17,531 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.80 vs. limit=2.0 2023-05-16 21:49:31,972 INFO [finetune.py:992] (0/2) Epoch 13, batch 3050, loss[loss=0.165, simple_loss=0.2469, pruned_loss=0.04154, over 12192.00 frames. ], tot_loss[loss=0.1669, simple_loss=0.2561, pruned_loss=0.03885, over 2368525.07 frames. ], batch size: 31, lr: 3.71e-03, grad_scale: 16.0 2023-05-16 21:49:34,896 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.5253, 5.1775, 4.7141, 4.7063, 5.3036, 4.6390, 4.7045, 4.6360], device='cuda:0'), covar=tensor([0.1665, 0.1068, 0.1340, 0.2185, 0.1135, 0.2318, 0.2097, 0.1387], device='cuda:0'), in_proj_covar=tensor([0.0365, 0.0500, 0.0399, 0.0449, 0.0468, 0.0437, 0.0397, 0.0396], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 21:49:41,327 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=249871.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 21:49:44,045 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.775e+02 2.733e+02 3.086e+02 3.787e+02 9.762e+02, threshold=6.172e+02, percent-clipped=2.0 2023-05-16 21:49:44,266 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=249875.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 21:49:45,585 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.0552, 3.8866, 4.0434, 3.6731, 3.9001, 3.7692, 4.0538, 3.5577], device='cuda:0'), covar=tensor([0.0382, 0.0393, 0.0343, 0.0276, 0.0372, 0.0326, 0.0258, 0.1409], device='cuda:0'), in_proj_covar=tensor([0.0263, 0.0266, 0.0291, 0.0263, 0.0263, 0.0264, 0.0238, 0.0217], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 21:50:07,198 INFO [finetune.py:992] (0/2) Epoch 13, batch 3100, loss[loss=0.1816, simple_loss=0.2822, pruned_loss=0.04052, over 11738.00 frames. ], tot_loss[loss=0.1669, simple_loss=0.2564, pruned_loss=0.03872, over 2368578.81 frames. ], batch size: 44, lr: 3.71e-03, grad_scale: 16.0 2023-05-16 21:50:15,294 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=249919.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 21:50:20,257 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([1.9822, 2.2870, 2.3563, 2.3310, 2.1985, 1.9490, 2.2509, 1.6047], device='cuda:0'), covar=tensor([0.0376, 0.0213, 0.0199, 0.0212, 0.0317, 0.0222, 0.0163, 0.0471], device='cuda:0'), in_proj_covar=tensor([0.0192, 0.0159, 0.0162, 0.0187, 0.0202, 0.0198, 0.0170, 0.0202], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 21:50:41,554 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=249955.0, num_to_drop=1, layers_to_drop={0} 2023-05-16 21:50:44,104 INFO [finetune.py:992] (0/2) Epoch 13, batch 3150, loss[loss=0.1661, simple_loss=0.2635, pruned_loss=0.03435, over 11811.00 frames. ], tot_loss[loss=0.1678, simple_loss=0.2573, pruned_loss=0.03911, over 2359339.50 frames. ], batch size: 44, lr: 3.71e-03, grad_scale: 16.0 2023-05-16 21:50:56,310 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.617e+02 2.680e+02 3.172e+02 3.678e+02 6.003e+02, threshold=6.344e+02, percent-clipped=0.0 2023-05-16 21:50:57,260 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=249976.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 21:51:05,798 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.7672, 4.3729, 4.4686, 4.6754, 4.5260, 4.6022, 4.4400, 2.5100], device='cuda:0'), covar=tensor([0.0087, 0.0080, 0.0100, 0.0057, 0.0050, 0.0100, 0.0088, 0.0773], device='cuda:0'), in_proj_covar=tensor([0.0068, 0.0079, 0.0082, 0.0073, 0.0060, 0.0093, 0.0082, 0.0096], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-16 21:51:14,657 INFO [checkpoint.py:75] (0/2) Saving checkpoint to pruned_transducer_stateless7/exp_giga_finetune/checkpoint-150000.pt 2023-05-16 21:51:23,492 INFO [finetune.py:992] (0/2) Epoch 13, batch 3200, loss[loss=0.1583, simple_loss=0.2433, pruned_loss=0.03667, over 12192.00 frames. ], tot_loss[loss=0.1668, simple_loss=0.2562, pruned_loss=0.03868, over 2366919.92 frames. ], batch size: 29, lr: 3.71e-03, grad_scale: 16.0 2023-05-16 21:51:29,422 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=250016.0, num_to_drop=1, layers_to_drop={1} 2023-05-16 21:51:42,572 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.1104, 3.7105, 3.9559, 4.3780, 2.9522, 3.7529, 2.5333, 3.9253], device='cuda:0'), covar=tensor([0.1768, 0.0857, 0.0829, 0.0600, 0.1125, 0.0712, 0.1857, 0.1029], device='cuda:0'), in_proj_covar=tensor([0.0230, 0.0263, 0.0294, 0.0358, 0.0240, 0.0244, 0.0261, 0.0377], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 21:51:44,551 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=250037.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 21:51:59,358 INFO [finetune.py:992] (0/2) Epoch 13, batch 3250, loss[loss=0.1836, simple_loss=0.2824, pruned_loss=0.04234, over 11794.00 frames. ], tot_loss[loss=0.1671, simple_loss=0.2568, pruned_loss=0.03864, over 2370308.14 frames. ], batch size: 44, lr: 3.71e-03, grad_scale: 16.0 2023-05-16 21:52:11,453 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.040e+02 2.740e+02 3.037e+02 3.497e+02 7.472e+02, threshold=6.074e+02, percent-clipped=1.0 2023-05-16 21:52:35,810 INFO [finetune.py:992] (0/2) Epoch 13, batch 3300, loss[loss=0.1405, simple_loss=0.2241, pruned_loss=0.02842, over 12190.00 frames. ], tot_loss[loss=0.1671, simple_loss=0.2567, pruned_loss=0.03876, over 2361464.39 frames. ], batch size: 29, lr: 3.71e-03, grad_scale: 16.0 2023-05-16 21:52:38,037 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.3701, 5.1345, 5.3043, 5.3618, 4.9702, 5.0219, 4.7806, 5.2980], device='cuda:0'), covar=tensor([0.0722, 0.0705, 0.0772, 0.0589, 0.1915, 0.1355, 0.0563, 0.1045], device='cuda:0'), in_proj_covar=tensor([0.0551, 0.0715, 0.0626, 0.0643, 0.0862, 0.0761, 0.0566, 0.0489], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0005, 0.0004, 0.0004, 0.0005, 0.0005, 0.0004, 0.0003], device='cuda:0') 2023-05-16 21:53:04,282 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.9435, 5.9257, 5.6535, 5.1165, 5.0970, 5.7797, 5.4162, 5.1993], device='cuda:0'), covar=tensor([0.0782, 0.0976, 0.0761, 0.1811, 0.0689, 0.0856, 0.1785, 0.1154], device='cuda:0'), in_proj_covar=tensor([0.0624, 0.0566, 0.0524, 0.0638, 0.0418, 0.0725, 0.0786, 0.0574], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0004, 0.0004, 0.0003], device='cuda:0') 2023-05-16 21:53:11,467 INFO [finetune.py:992] (0/2) Epoch 13, batch 3350, loss[loss=0.1855, simple_loss=0.2869, pruned_loss=0.04207, over 12055.00 frames. ], tot_loss[loss=0.1673, simple_loss=0.2568, pruned_loss=0.03891, over 2364456.45 frames. ], batch size: 42, lr: 3.71e-03, grad_scale: 16.0 2023-05-16 21:53:19,977 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=250170.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 21:53:23,281 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.069e+02 2.870e+02 3.367e+02 3.931e+02 8.641e+02, threshold=6.735e+02, percent-clipped=4.0 2023-05-16 21:53:47,302 INFO [finetune.py:992] (0/2) Epoch 13, batch 3400, loss[loss=0.1535, simple_loss=0.2548, pruned_loss=0.02607, over 12192.00 frames. ], tot_loss[loss=0.1664, simple_loss=0.2557, pruned_loss=0.03855, over 2369785.74 frames. ], batch size: 35, lr: 3.71e-03, grad_scale: 16.0 2023-05-16 21:53:52,498 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.0625, 4.7315, 4.9213, 5.0212, 4.8184, 4.9723, 4.8670, 2.5042], device='cuda:0'), covar=tensor([0.0073, 0.0065, 0.0076, 0.0055, 0.0051, 0.0093, 0.0077, 0.0814], device='cuda:0'), in_proj_covar=tensor([0.0068, 0.0079, 0.0081, 0.0073, 0.0060, 0.0093, 0.0082, 0.0096], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-16 21:54:05,609 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2023-05-16 21:54:22,377 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.4959, 5.2929, 5.4255, 5.4752, 5.0937, 5.0907, 4.8727, 5.3374], device='cuda:0'), covar=tensor([0.0687, 0.0610, 0.0789, 0.0548, 0.1850, 0.1351, 0.0573, 0.1244], device='cuda:0'), in_proj_covar=tensor([0.0556, 0.0719, 0.0632, 0.0647, 0.0869, 0.0766, 0.0568, 0.0493], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0005, 0.0004, 0.0004, 0.0005, 0.0005, 0.0004, 0.0003], device='cuda:0') 2023-05-16 21:54:24,367 INFO [finetune.py:992] (0/2) Epoch 13, batch 3450, loss[loss=0.1587, simple_loss=0.2538, pruned_loss=0.03179, over 12037.00 frames. ], tot_loss[loss=0.1669, simple_loss=0.2564, pruned_loss=0.03873, over 2375598.08 frames. ], batch size: 40, lr: 3.71e-03, grad_scale: 16.0 2023-05-16 21:54:36,533 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.009e+02 2.724e+02 3.154e+02 3.692e+02 7.544e+02, threshold=6.308e+02, percent-clipped=1.0 2023-05-16 21:54:59,963 INFO [finetune.py:992] (0/2) Epoch 13, batch 3500, loss[loss=0.1956, simple_loss=0.2905, pruned_loss=0.05039, over 11334.00 frames. ], tot_loss[loss=0.166, simple_loss=0.2556, pruned_loss=0.0382, over 2383578.55 frames. ], batch size: 55, lr: 3.71e-03, grad_scale: 16.0 2023-05-16 21:55:00,632 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.94 vs. limit=2.0 2023-05-16 21:55:02,164 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=250311.0, num_to_drop=1, layers_to_drop={0} 2023-05-16 21:55:17,433 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=250332.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 21:55:35,829 INFO [finetune.py:992] (0/2) Epoch 13, batch 3550, loss[loss=0.1926, simple_loss=0.2795, pruned_loss=0.05286, over 11562.00 frames. ], tot_loss[loss=0.1652, simple_loss=0.2546, pruned_loss=0.03788, over 2376282.56 frames. ], batch size: 48, lr: 3.71e-03, grad_scale: 16.0 2023-05-16 21:55:43,184 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.1829, 2.4961, 3.6681, 3.1336, 3.5708, 3.2782, 2.6126, 3.5446], device='cuda:0'), covar=tensor([0.0135, 0.0374, 0.0148, 0.0250, 0.0151, 0.0168, 0.0330, 0.0140], device='cuda:0'), in_proj_covar=tensor([0.0179, 0.0206, 0.0189, 0.0189, 0.0218, 0.0165, 0.0198, 0.0189], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 21:55:47,983 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.993e+02 2.648e+02 3.157e+02 3.792e+02 7.894e+02, threshold=6.315e+02, percent-clipped=2.0 2023-05-16 21:55:51,990 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.75 vs. limit=2.0 2023-05-16 21:55:59,211 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=4.37 vs. limit=5.0 2023-05-16 21:56:13,476 INFO [finetune.py:992] (0/2) Epoch 13, batch 3600, loss[loss=0.1584, simple_loss=0.2493, pruned_loss=0.03374, over 12144.00 frames. ], tot_loss[loss=0.1656, simple_loss=0.2549, pruned_loss=0.03821, over 2372067.72 frames. ], batch size: 30, lr: 3.71e-03, grad_scale: 16.0 2023-05-16 21:56:20,846 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.2210, 2.5969, 3.7277, 3.2230, 3.4955, 3.3357, 2.7047, 3.5312], device='cuda:0'), covar=tensor([0.0139, 0.0385, 0.0170, 0.0248, 0.0205, 0.0211, 0.0356, 0.0182], device='cuda:0'), in_proj_covar=tensor([0.0179, 0.0205, 0.0189, 0.0188, 0.0218, 0.0164, 0.0198, 0.0189], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 21:56:49,022 INFO [finetune.py:992] (0/2) Epoch 13, batch 3650, loss[loss=0.1853, simple_loss=0.2737, pruned_loss=0.0485, over 11629.00 frames. ], tot_loss[loss=0.1659, simple_loss=0.2557, pruned_loss=0.03807, over 2377856.72 frames. ], batch size: 48, lr: 3.71e-03, grad_scale: 16.0 2023-05-16 21:56:57,713 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=250470.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 21:56:59,919 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.3726, 3.4713, 3.2469, 3.1947, 2.9491, 2.7092, 3.5555, 2.1897], device='cuda:0'), covar=tensor([0.0402, 0.0162, 0.0195, 0.0194, 0.0373, 0.0380, 0.0129, 0.0514], device='cuda:0'), in_proj_covar=tensor([0.0190, 0.0158, 0.0162, 0.0187, 0.0201, 0.0197, 0.0170, 0.0201], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 21:57:01,131 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.768e+02 2.465e+02 2.932e+02 3.493e+02 5.778e+02, threshold=5.863e+02, percent-clipped=0.0 2023-05-16 21:57:24,631 INFO [finetune.py:992] (0/2) Epoch 13, batch 3700, loss[loss=0.171, simple_loss=0.2581, pruned_loss=0.0419, over 12186.00 frames. ], tot_loss[loss=0.1659, simple_loss=0.2554, pruned_loss=0.03822, over 2379371.08 frames. ], batch size: 31, lr: 3.71e-03, grad_scale: 8.0 2023-05-16 21:57:31,728 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=250518.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 21:58:01,068 INFO [finetune.py:992] (0/2) Epoch 13, batch 3750, loss[loss=0.191, simple_loss=0.2717, pruned_loss=0.05521, over 10510.00 frames. ], tot_loss[loss=0.167, simple_loss=0.256, pruned_loss=0.03897, over 2371488.38 frames. ], batch size: 68, lr: 3.71e-03, grad_scale: 8.0 2023-05-16 21:58:06,250 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.2157, 4.5783, 2.8952, 2.5161, 3.9545, 2.4231, 3.9279, 3.1144], device='cuda:0'), covar=tensor([0.0759, 0.0488, 0.1062, 0.1659, 0.0307, 0.1461, 0.0459, 0.0872], device='cuda:0'), in_proj_covar=tensor([0.0190, 0.0255, 0.0177, 0.0200, 0.0141, 0.0181, 0.0197, 0.0175], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 21:58:13,885 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.851e+02 2.934e+02 3.432e+02 4.063e+02 1.664e+03, threshold=6.864e+02, percent-clipped=5.0 2023-05-16 21:58:37,149 INFO [finetune.py:992] (0/2) Epoch 13, batch 3800, loss[loss=0.1651, simple_loss=0.2614, pruned_loss=0.03443, over 12091.00 frames. ], tot_loss[loss=0.1667, simple_loss=0.256, pruned_loss=0.03873, over 2362314.61 frames. ], batch size: 32, lr: 3.71e-03, grad_scale: 8.0 2023-05-16 21:58:39,369 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=250611.0, num_to_drop=1, layers_to_drop={0} 2023-05-16 21:58:54,410 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=250632.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 21:59:04,752 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.38 vs. limit=2.0 2023-05-16 21:59:12,577 INFO [finetune.py:992] (0/2) Epoch 13, batch 3850, loss[loss=0.171, simple_loss=0.2645, pruned_loss=0.03882, over 11661.00 frames. ], tot_loss[loss=0.1671, simple_loss=0.2566, pruned_loss=0.03877, over 2364488.12 frames. ], batch size: 48, lr: 3.70e-03, grad_scale: 8.0 2023-05-16 21:59:13,333 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=250659.0, num_to_drop=1, layers_to_drop={0} 2023-05-16 21:59:15,507 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=250662.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 21:59:25,629 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.835e+02 2.662e+02 3.077e+02 3.752e+02 6.691e+02, threshold=6.154e+02, percent-clipped=0.0 2023-05-16 21:59:29,041 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=250680.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 21:59:42,461 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.8781, 4.6080, 4.6493, 4.7782, 4.7311, 4.8861, 4.6190, 2.4626], device='cuda:0'), covar=tensor([0.0122, 0.0079, 0.0121, 0.0080, 0.0057, 0.0109, 0.0120, 0.0935], device='cuda:0'), in_proj_covar=tensor([0.0068, 0.0078, 0.0082, 0.0073, 0.0060, 0.0093, 0.0082, 0.0096], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-16 21:59:49,229 INFO [finetune.py:992] (0/2) Epoch 13, batch 3900, loss[loss=0.1738, simple_loss=0.267, pruned_loss=0.04032, over 11809.00 frames. ], tot_loss[loss=0.1671, simple_loss=0.2566, pruned_loss=0.0388, over 2372554.78 frames. ], batch size: 44, lr: 3.70e-03, grad_scale: 8.0 2023-05-16 21:59:59,980 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=250723.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 22:00:24,739 INFO [finetune.py:992] (0/2) Epoch 13, batch 3950, loss[loss=0.1946, simple_loss=0.2863, pruned_loss=0.0515, over 12307.00 frames. ], tot_loss[loss=0.1671, simple_loss=0.2565, pruned_loss=0.0388, over 2374160.67 frames. ], batch size: 34, lr: 3.70e-03, grad_scale: 8.0 2023-05-16 22:00:37,726 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.771e+02 2.733e+02 3.127e+02 3.686e+02 8.040e+02, threshold=6.254e+02, percent-clipped=1.0 2023-05-16 22:00:46,624 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.5635, 4.0217, 4.0647, 4.5712, 3.1214, 3.9643, 2.7769, 4.2521], device='cuda:0'), covar=tensor([0.1471, 0.0854, 0.1126, 0.0711, 0.1199, 0.0702, 0.1786, 0.1079], device='cuda:0'), in_proj_covar=tensor([0.0229, 0.0264, 0.0294, 0.0358, 0.0240, 0.0243, 0.0259, 0.0375], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 22:01:01,039 INFO [finetune.py:992] (0/2) Epoch 13, batch 4000, loss[loss=0.1536, simple_loss=0.24, pruned_loss=0.03358, over 12085.00 frames. ], tot_loss[loss=0.1668, simple_loss=0.2564, pruned_loss=0.03863, over 2368281.69 frames. ], batch size: 32, lr: 3.70e-03, grad_scale: 8.0 2023-05-16 22:01:14,844 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.6235, 3.0560, 3.7302, 4.4476, 3.8258, 4.4636, 3.8483, 3.2610], device='cuda:0'), covar=tensor([0.0036, 0.0288, 0.0150, 0.0053, 0.0135, 0.0071, 0.0131, 0.0336], device='cuda:0'), in_proj_covar=tensor([0.0087, 0.0121, 0.0102, 0.0077, 0.0102, 0.0114, 0.0094, 0.0136], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 22:01:19,693 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=250833.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 22:01:28,853 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=250845.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 22:01:37,944 INFO [finetune.py:992] (0/2) Epoch 13, batch 4050, loss[loss=0.1943, simple_loss=0.2854, pruned_loss=0.05162, over 12011.00 frames. ], tot_loss[loss=0.167, simple_loss=0.2564, pruned_loss=0.0388, over 2364986.38 frames. ], batch size: 40, lr: 3.70e-03, grad_scale: 8.0 2023-05-16 22:01:50,782 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.725e+02 2.811e+02 3.319e+02 3.784e+02 9.660e+02, threshold=6.638e+02, percent-clipped=5.0 2023-05-16 22:02:03,607 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=250894.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 22:02:08,034 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.9288, 3.4878, 5.2622, 2.8507, 2.8653, 3.8210, 3.2679, 3.6418], device='cuda:0'), covar=tensor([0.0414, 0.1096, 0.0385, 0.1113, 0.2029, 0.1768, 0.1335, 0.1514], device='cuda:0'), in_proj_covar=tensor([0.0240, 0.0239, 0.0254, 0.0186, 0.0241, 0.0299, 0.0228, 0.0268], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-16 22:02:12,237 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=250906.0, num_to_drop=1, layers_to_drop={1} 2023-05-16 22:02:13,337 INFO [finetune.py:992] (0/2) Epoch 13, batch 4100, loss[loss=0.161, simple_loss=0.2442, pruned_loss=0.03887, over 12258.00 frames. ], tot_loss[loss=0.1671, simple_loss=0.2563, pruned_loss=0.039, over 2361832.41 frames. ], batch size: 32, lr: 3.70e-03, grad_scale: 8.0 2023-05-16 22:02:20,632 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.7100, 4.6303, 4.5147, 4.5512, 4.2910, 4.7160, 4.6590, 4.8993], device='cuda:0'), covar=tensor([0.0334, 0.0164, 0.0231, 0.0375, 0.0782, 0.0466, 0.0217, 0.0220], device='cuda:0'), in_proj_covar=tensor([0.0196, 0.0198, 0.0194, 0.0250, 0.0245, 0.0222, 0.0182, 0.0230], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0005, 0.0004, 0.0005, 0.0004, 0.0005], device='cuda:0') 2023-05-16 22:02:48,842 INFO [finetune.py:992] (0/2) Epoch 13, batch 4150, loss[loss=0.1431, simple_loss=0.2316, pruned_loss=0.02726, over 12331.00 frames. ], tot_loss[loss=0.167, simple_loss=0.2562, pruned_loss=0.0389, over 2364709.84 frames. ], batch size: 30, lr: 3.70e-03, grad_scale: 8.0 2023-05-16 22:03:02,402 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.041e+02 2.619e+02 3.090e+02 3.951e+02 3.498e+03, threshold=6.180e+02, percent-clipped=5.0 2023-05-16 22:03:07,429 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.2039, 4.9529, 5.2536, 5.1890, 4.4083, 4.5527, 4.6125, 4.9413], device='cuda:0'), covar=tensor([0.1088, 0.1148, 0.0874, 0.1169, 0.3681, 0.2224, 0.0783, 0.1884], device='cuda:0'), in_proj_covar=tensor([0.0553, 0.0710, 0.0632, 0.0644, 0.0862, 0.0751, 0.0568, 0.0491], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0005, 0.0004, 0.0004, 0.0005, 0.0005, 0.0004, 0.0003], device='cuda:0') 2023-05-16 22:03:15,153 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.5790, 3.5538, 3.2342, 3.1474, 2.9170, 2.7591, 3.6572, 2.3331], device='cuda:0'), covar=tensor([0.0367, 0.0136, 0.0209, 0.0234, 0.0372, 0.0331, 0.0123, 0.0501], device='cuda:0'), in_proj_covar=tensor([0.0192, 0.0160, 0.0164, 0.0188, 0.0203, 0.0198, 0.0170, 0.0202], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 22:03:19,515 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2023-05-16 22:03:26,016 INFO [finetune.py:992] (0/2) Epoch 13, batch 4200, loss[loss=0.1658, simple_loss=0.2529, pruned_loss=0.03935, over 12108.00 frames. ], tot_loss[loss=0.1679, simple_loss=0.257, pruned_loss=0.03945, over 2370524.31 frames. ], batch size: 42, lr: 3.70e-03, grad_scale: 8.0 2023-05-16 22:03:26,917 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.6071, 2.9023, 3.3279, 4.5317, 2.4908, 4.4117, 4.6473, 4.6683], device='cuda:0'), covar=tensor([0.0135, 0.1029, 0.0476, 0.0128, 0.1320, 0.0273, 0.0122, 0.0084], device='cuda:0'), in_proj_covar=tensor([0.0120, 0.0199, 0.0181, 0.0119, 0.0189, 0.0178, 0.0173, 0.0120], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 22:03:33,377 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=251018.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 22:04:01,505 INFO [finetune.py:992] (0/2) Epoch 13, batch 4250, loss[loss=0.1629, simple_loss=0.2558, pruned_loss=0.03498, over 12085.00 frames. ], tot_loss[loss=0.1684, simple_loss=0.2577, pruned_loss=0.03957, over 2370335.78 frames. ], batch size: 32, lr: 3.70e-03, grad_scale: 8.0 2023-05-16 22:04:14,456 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.860e+02 2.557e+02 2.991e+02 3.941e+02 6.668e+02, threshold=5.982e+02, percent-clipped=2.0 2023-05-16 22:04:37,363 INFO [finetune.py:992] (0/2) Epoch 13, batch 4300, loss[loss=0.1953, simple_loss=0.2873, pruned_loss=0.05169, over 11590.00 frames. ], tot_loss[loss=0.1684, simple_loss=0.2578, pruned_loss=0.03954, over 2363698.63 frames. ], batch size: 48, lr: 3.70e-03, grad_scale: 8.0 2023-05-16 22:05:14,514 INFO [finetune.py:992] (0/2) Epoch 13, batch 4350, loss[loss=0.1689, simple_loss=0.2652, pruned_loss=0.03632, over 12192.00 frames. ], tot_loss[loss=0.1683, simple_loss=0.2576, pruned_loss=0.0395, over 2366589.81 frames. ], batch size: 35, lr: 3.70e-03, grad_scale: 8.0 2023-05-16 22:05:27,216 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.037e+02 2.611e+02 3.160e+02 3.927e+02 7.515e+02, threshold=6.319e+02, percent-clipped=3.0 2023-05-16 22:05:36,399 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=251189.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 22:05:42,948 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=251198.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 22:05:45,314 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=251201.0, num_to_drop=1, layers_to_drop={2} 2023-05-16 22:05:50,023 INFO [finetune.py:992] (0/2) Epoch 13, batch 4400, loss[loss=0.1853, simple_loss=0.273, pruned_loss=0.04883, over 12085.00 frames. ], tot_loss[loss=0.1681, simple_loss=0.2572, pruned_loss=0.0395, over 2363810.91 frames. ], batch size: 32, lr: 3.70e-03, grad_scale: 8.0 2023-05-16 22:06:09,360 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 2023-05-16 22:06:25,247 INFO [finetune.py:992] (0/2) Epoch 13, batch 4450, loss[loss=0.1511, simple_loss=0.2315, pruned_loss=0.03538, over 12363.00 frames. ], tot_loss[loss=0.1681, simple_loss=0.2575, pruned_loss=0.03941, over 2366147.35 frames. ], batch size: 30, lr: 3.70e-03, grad_scale: 8.0 2023-05-16 22:06:26,186 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=251259.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 22:06:39,509 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.058e+02 2.761e+02 3.162e+02 4.023e+02 1.114e+03, threshold=6.324e+02, percent-clipped=6.0 2023-05-16 22:06:41,728 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.2554, 4.9120, 5.1482, 5.1882, 5.0527, 5.1809, 5.0381, 3.0849], device='cuda:0'), covar=tensor([0.0092, 0.0069, 0.0065, 0.0054, 0.0046, 0.0088, 0.0088, 0.0678], device='cuda:0'), in_proj_covar=tensor([0.0070, 0.0080, 0.0083, 0.0075, 0.0062, 0.0095, 0.0083, 0.0097], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 22:06:57,521 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=251301.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 22:07:02,421 INFO [finetune.py:992] (0/2) Epoch 13, batch 4500, loss[loss=0.1601, simple_loss=0.2542, pruned_loss=0.033, over 12115.00 frames. ], tot_loss[loss=0.1679, simple_loss=0.257, pruned_loss=0.03937, over 2368293.15 frames. ], batch size: 33, lr: 3.70e-03, grad_scale: 8.0 2023-05-16 22:07:09,822 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=251318.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 22:07:12,594 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.7390, 3.2737, 5.1533, 2.6675, 2.6606, 3.7632, 3.3075, 3.7811], device='cuda:0'), covar=tensor([0.0413, 0.1165, 0.0247, 0.1186, 0.1977, 0.1359, 0.1278, 0.1095], device='cuda:0'), in_proj_covar=tensor([0.0239, 0.0238, 0.0254, 0.0185, 0.0241, 0.0298, 0.0226, 0.0266], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-16 22:07:36,447 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.3026, 6.1527, 5.7638, 5.8000, 6.1945, 5.5710, 5.5793, 5.7526], device='cuda:0'), covar=tensor([0.1411, 0.0852, 0.1193, 0.1652, 0.0823, 0.1965, 0.2014, 0.1157], device='cuda:0'), in_proj_covar=tensor([0.0363, 0.0501, 0.0398, 0.0450, 0.0470, 0.0439, 0.0396, 0.0396], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 22:07:37,813 INFO [finetune.py:992] (0/2) Epoch 13, batch 4550, loss[loss=0.1722, simple_loss=0.267, pruned_loss=0.03872, over 11329.00 frames. ], tot_loss[loss=0.1679, simple_loss=0.257, pruned_loss=0.03939, over 2370745.42 frames. ], batch size: 55, lr: 3.70e-03, grad_scale: 8.0 2023-05-16 22:07:40,781 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=251362.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 22:07:43,381 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=251366.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 22:07:50,400 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.868e+02 2.632e+02 3.087e+02 3.733e+02 6.616e+02, threshold=6.173e+02, percent-clipped=1.0 2023-05-16 22:08:10,833 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=251404.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 22:08:14,073 INFO [finetune.py:992] (0/2) Epoch 13, batch 4600, loss[loss=0.1417, simple_loss=0.2208, pruned_loss=0.03126, over 11769.00 frames. ], tot_loss[loss=0.1672, simple_loss=0.2564, pruned_loss=0.03901, over 2370760.08 frames. ], batch size: 26, lr: 3.70e-03, grad_scale: 8.0 2023-05-16 22:08:15,802 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.4381, 4.8114, 3.0223, 2.7343, 4.0429, 2.6083, 3.9664, 3.3228], device='cuda:0'), covar=tensor([0.0676, 0.0500, 0.1055, 0.1483, 0.0319, 0.1314, 0.0471, 0.0792], device='cuda:0'), in_proj_covar=tensor([0.0190, 0.0258, 0.0178, 0.0202, 0.0142, 0.0181, 0.0198, 0.0177], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 22:08:34,908 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.6028, 3.6506, 3.3042, 3.1807, 3.0076, 2.8297, 3.6638, 2.3854], device='cuda:0'), covar=tensor([0.0348, 0.0117, 0.0192, 0.0202, 0.0338, 0.0329, 0.0128, 0.0442], device='cuda:0'), in_proj_covar=tensor([0.0193, 0.0160, 0.0164, 0.0189, 0.0203, 0.0198, 0.0171, 0.0202], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 22:08:50,343 INFO [finetune.py:992] (0/2) Epoch 13, batch 4650, loss[loss=0.1767, simple_loss=0.2695, pruned_loss=0.0419, over 12033.00 frames. ], tot_loss[loss=0.1671, simple_loss=0.2563, pruned_loss=0.03891, over 2377523.16 frames. ], batch size: 40, lr: 3.70e-03, grad_scale: 8.0 2023-05-16 22:08:50,655 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=2.78 vs. limit=5.0 2023-05-16 22:08:55,539 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=251465.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 22:08:57,175 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.85 vs. limit=2.0 2023-05-16 22:09:03,007 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.710e+02 2.631e+02 2.993e+02 3.545e+02 6.646e+02, threshold=5.986e+02, percent-clipped=1.0 2023-05-16 22:09:12,318 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=251489.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 22:09:17,809 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 2023-05-16 22:09:20,885 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=251501.0, num_to_drop=1, layers_to_drop={1} 2023-05-16 22:09:25,572 INFO [finetune.py:992] (0/2) Epoch 13, batch 4700, loss[loss=0.2052, simple_loss=0.2831, pruned_loss=0.0636, over 7766.00 frames. ], tot_loss[loss=0.167, simple_loss=0.256, pruned_loss=0.03896, over 2376401.81 frames. ], batch size: 97, lr: 3.70e-03, grad_scale: 8.0 2023-05-16 22:09:46,481 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=251537.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 22:09:55,186 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=251549.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 22:09:58,906 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=251554.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 22:10:02,299 INFO [finetune.py:992] (0/2) Epoch 13, batch 4750, loss[loss=0.2202, simple_loss=0.3011, pruned_loss=0.06963, over 8136.00 frames. ], tot_loss[loss=0.1661, simple_loss=0.2551, pruned_loss=0.03855, over 2378517.12 frames. ], batch size: 99, lr: 3.70e-03, grad_scale: 8.0 2023-05-16 22:10:15,783 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.985e+02 2.680e+02 3.120e+02 3.758e+02 6.071e+02, threshold=6.239e+02, percent-clipped=1.0 2023-05-16 22:10:24,680 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.37 vs. limit=2.0 2023-05-16 22:10:36,925 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.38 vs. limit=2.0 2023-05-16 22:10:38,635 INFO [finetune.py:992] (0/2) Epoch 13, batch 4800, loss[loss=0.1491, simple_loss=0.2345, pruned_loss=0.03183, over 12116.00 frames. ], tot_loss[loss=0.1679, simple_loss=0.2567, pruned_loss=0.03958, over 2366666.35 frames. ], batch size: 30, lr: 3.70e-03, grad_scale: 8.0 2023-05-16 22:10:40,657 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.31 vs. limit=2.0 2023-05-16 22:10:48,903 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.2306, 2.6087, 3.6417, 3.1677, 3.5149, 3.2698, 2.7176, 3.6222], device='cuda:0'), covar=tensor([0.0127, 0.0327, 0.0171, 0.0212, 0.0171, 0.0177, 0.0315, 0.0134], device='cuda:0'), in_proj_covar=tensor([0.0181, 0.0206, 0.0192, 0.0190, 0.0220, 0.0166, 0.0201, 0.0190], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 22:11:01,967 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=251641.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 22:11:13,325 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=251657.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 22:11:13,951 INFO [finetune.py:992] (0/2) Epoch 13, batch 4850, loss[loss=0.1527, simple_loss=0.2371, pruned_loss=0.03422, over 12093.00 frames. ], tot_loss[loss=0.1675, simple_loss=0.2565, pruned_loss=0.03923, over 2376590.43 frames. ], batch size: 32, lr: 3.70e-03, grad_scale: 8.0 2023-05-16 22:11:26,850 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.963e+02 2.707e+02 3.022e+02 3.497e+02 7.230e+02, threshold=6.045e+02, percent-clipped=1.0 2023-05-16 22:11:28,554 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.58 vs. limit=2.0 2023-05-16 22:11:46,194 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=251702.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 22:11:50,244 INFO [finetune.py:992] (0/2) Epoch 13, batch 4900, loss[loss=0.1705, simple_loss=0.2601, pruned_loss=0.04041, over 12146.00 frames. ], tot_loss[loss=0.1672, simple_loss=0.2557, pruned_loss=0.03933, over 2371221.97 frames. ], batch size: 34, lr: 3.70e-03, grad_scale: 8.0 2023-05-16 22:11:51,211 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=251709.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 22:12:21,380 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.7697, 3.4046, 5.1692, 2.6676, 2.9508, 3.8152, 3.3123, 3.8104], device='cuda:0'), covar=tensor([0.0388, 0.1088, 0.0281, 0.1131, 0.1811, 0.1615, 0.1283, 0.1164], device='cuda:0'), in_proj_covar=tensor([0.0238, 0.0237, 0.0254, 0.0184, 0.0239, 0.0297, 0.0226, 0.0266], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-16 22:12:26,817 INFO [finetune.py:992] (0/2) Epoch 13, batch 4950, loss[loss=0.1908, simple_loss=0.2861, pruned_loss=0.04774, over 11649.00 frames. ], tot_loss[loss=0.1674, simple_loss=0.2563, pruned_loss=0.03925, over 2364842.42 frames. ], batch size: 48, lr: 3.70e-03, grad_scale: 8.0 2023-05-16 22:12:28,367 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=251760.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 22:12:30,552 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.0764, 2.2463, 2.9281, 2.9651, 3.0170, 3.0763, 2.8984, 2.5059], device='cuda:0'), covar=tensor([0.0076, 0.0403, 0.0175, 0.0086, 0.0131, 0.0101, 0.0145, 0.0351], device='cuda:0'), in_proj_covar=tensor([0.0087, 0.0121, 0.0104, 0.0078, 0.0102, 0.0114, 0.0095, 0.0137], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 22:12:35,733 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=251770.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 22:12:39,749 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.714e+02 2.575e+02 3.026e+02 3.736e+02 6.475e+02, threshold=6.052e+02, percent-clipped=1.0 2023-05-16 22:13:02,873 INFO [finetune.py:992] (0/2) Epoch 13, batch 5000, loss[loss=0.1785, simple_loss=0.2674, pruned_loss=0.04482, over 12371.00 frames. ], tot_loss[loss=0.167, simple_loss=0.2561, pruned_loss=0.03901, over 2368474.47 frames. ], batch size: 38, lr: 3.70e-03, grad_scale: 8.0 2023-05-16 22:13:14,520 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.86 vs. limit=2.0 2023-05-16 22:13:21,434 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.5187, 2.4974, 3.0667, 4.3557, 2.5040, 4.3929, 4.5447, 4.4984], device='cuda:0'), covar=tensor([0.0119, 0.1307, 0.0541, 0.0139, 0.1316, 0.0252, 0.0135, 0.0093], device='cuda:0'), in_proj_covar=tensor([0.0121, 0.0201, 0.0182, 0.0119, 0.0190, 0.0179, 0.0173, 0.0121], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 22:13:35,941 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=251854.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 22:13:38,550 INFO [finetune.py:992] (0/2) Epoch 13, batch 5050, loss[loss=0.1794, simple_loss=0.2796, pruned_loss=0.03962, over 12082.00 frames. ], tot_loss[loss=0.1664, simple_loss=0.2553, pruned_loss=0.03872, over 2371362.83 frames. ], batch size: 42, lr: 3.69e-03, grad_scale: 8.0 2023-05-16 22:13:40,313 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.51 vs. limit=2.0 2023-05-16 22:13:51,911 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.990e+02 2.734e+02 3.058e+02 3.779e+02 7.472e+02, threshold=6.116e+02, percent-clipped=2.0 2023-05-16 22:14:10,055 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=3.37 vs. limit=5.0 2023-05-16 22:14:10,390 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=251902.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 22:14:14,635 INFO [finetune.py:992] (0/2) Epoch 13, batch 5100, loss[loss=0.1676, simple_loss=0.2437, pruned_loss=0.04578, over 12276.00 frames. ], tot_loss[loss=0.1667, simple_loss=0.2555, pruned_loss=0.03896, over 2363449.58 frames. ], batch size: 33, lr: 3.69e-03, grad_scale: 8.0 2023-05-16 22:14:49,365 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=251957.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 22:14:49,891 INFO [finetune.py:992] (0/2) Epoch 13, batch 5150, loss[loss=0.198, simple_loss=0.2759, pruned_loss=0.06006, over 7687.00 frames. ], tot_loss[loss=0.1665, simple_loss=0.2551, pruned_loss=0.03891, over 2370197.84 frames. ], batch size: 98, lr: 3.69e-03, grad_scale: 8.0 2023-05-16 22:15:02,696 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.779e+02 2.613e+02 3.156e+02 3.619e+02 6.427e+02, threshold=6.311e+02, percent-clipped=2.0 2023-05-16 22:15:18,341 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=251997.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 22:15:20,721 INFO [checkpoint.py:75] (0/2) Saving checkpoint to pruned_transducer_stateless7/exp_giga_finetune/checkpoint-152000.pt 2023-05-16 22:15:27,919 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=252005.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 22:15:30,026 INFO [finetune.py:992] (0/2) Epoch 13, batch 5200, loss[loss=0.1871, simple_loss=0.2754, pruned_loss=0.0494, over 12120.00 frames. ], tot_loss[loss=0.1662, simple_loss=0.255, pruned_loss=0.03873, over 2373598.85 frames. ], batch size: 38, lr: 3.69e-03, grad_scale: 8.0 2023-05-16 22:16:05,472 INFO [finetune.py:992] (0/2) Epoch 13, batch 5250, loss[loss=0.1521, simple_loss=0.2288, pruned_loss=0.03769, over 12033.00 frames. ], tot_loss[loss=0.1657, simple_loss=0.2545, pruned_loss=0.03848, over 2377549.62 frames. ], batch size: 31, lr: 3.69e-03, grad_scale: 8.0 2023-05-16 22:16:07,142 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=252060.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 22:16:10,383 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=252065.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 22:16:14,697 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.4708, 3.5598, 3.2906, 3.2310, 2.9733, 2.8369, 3.6543, 2.4064], device='cuda:0'), covar=tensor([0.0445, 0.0156, 0.0198, 0.0215, 0.0426, 0.0394, 0.0142, 0.0499], device='cuda:0'), in_proj_covar=tensor([0.0196, 0.0164, 0.0167, 0.0192, 0.0208, 0.0203, 0.0174, 0.0206], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 22:16:18,036 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.791e+02 2.604e+02 3.008e+02 3.541e+02 6.263e+02, threshold=6.016e+02, percent-clipped=0.0 2023-05-16 22:16:40,728 INFO [finetune.py:992] (0/2) Epoch 13, batch 5300, loss[loss=0.1787, simple_loss=0.273, pruned_loss=0.04222, over 11512.00 frames. ], tot_loss[loss=0.1662, simple_loss=0.2547, pruned_loss=0.03891, over 2368354.29 frames. ], batch size: 55, lr: 3.69e-03, grad_scale: 8.0 2023-05-16 22:16:40,795 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=252108.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 22:16:58,871 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-05-16 22:17:17,354 INFO [finetune.py:992] (0/2) Epoch 13, batch 5350, loss[loss=0.1597, simple_loss=0.2581, pruned_loss=0.03062, over 12167.00 frames. ], tot_loss[loss=0.1665, simple_loss=0.2551, pruned_loss=0.03898, over 2376264.76 frames. ], batch size: 36, lr: 3.69e-03, grad_scale: 8.0 2023-05-16 22:17:30,032 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.764e+02 2.749e+02 3.226e+02 3.819e+02 7.797e+02, threshold=6.451e+02, percent-clipped=2.0 2023-05-16 22:17:53,051 INFO [finetune.py:992] (0/2) Epoch 13, batch 5400, loss[loss=0.1843, simple_loss=0.2706, pruned_loss=0.04898, over 12160.00 frames. ], tot_loss[loss=0.1667, simple_loss=0.2554, pruned_loss=0.03899, over 2372920.43 frames. ], batch size: 34, lr: 3.69e-03, grad_scale: 8.0 2023-05-16 22:17:54,230 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=3.35 vs. limit=5.0 2023-05-16 22:18:28,920 INFO [finetune.py:992] (0/2) Epoch 13, batch 5450, loss[loss=0.1508, simple_loss=0.2255, pruned_loss=0.03807, over 11989.00 frames. ], tot_loss[loss=0.1658, simple_loss=0.2546, pruned_loss=0.03849, over 2379880.24 frames. ], batch size: 28, lr: 3.69e-03, grad_scale: 8.0 2023-05-16 22:18:42,283 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.691e+02 2.623e+02 2.970e+02 3.795e+02 7.378e+02, threshold=5.940e+02, percent-clipped=2.0 2023-05-16 22:18:57,178 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=252297.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 22:18:57,355 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=2.54 vs. limit=5.0 2023-05-16 22:19:05,257 INFO [finetune.py:992] (0/2) Epoch 13, batch 5500, loss[loss=0.1531, simple_loss=0.2422, pruned_loss=0.03203, over 12284.00 frames. ], tot_loss[loss=0.1665, simple_loss=0.2555, pruned_loss=0.03882, over 2370013.22 frames. ], batch size: 33, lr: 3.69e-03, grad_scale: 8.0 2023-05-16 22:19:23,406 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.8663, 3.3020, 2.4684, 2.2510, 2.8897, 2.3724, 3.1357, 2.7188], device='cuda:0'), covar=tensor([0.0595, 0.0642, 0.0925, 0.1258, 0.0293, 0.1023, 0.0464, 0.0721], device='cuda:0'), in_proj_covar=tensor([0.0189, 0.0259, 0.0177, 0.0202, 0.0142, 0.0181, 0.0199, 0.0176], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 22:19:31,840 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=252345.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 22:19:41,069 INFO [finetune.py:992] (0/2) Epoch 13, batch 5550, loss[loss=0.1706, simple_loss=0.2647, pruned_loss=0.03829, over 12284.00 frames. ], tot_loss[loss=0.1669, simple_loss=0.2557, pruned_loss=0.03904, over 2364871.40 frames. ], batch size: 37, lr: 3.69e-03, grad_scale: 8.0 2023-05-16 22:19:46,362 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=252365.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 22:19:47,894 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.8056, 3.0651, 4.7106, 4.9327, 3.1319, 2.6803, 3.1493, 2.2439], device='cuda:0'), covar=tensor([0.1512, 0.2850, 0.0425, 0.0386, 0.1197, 0.2366, 0.2566, 0.4022], device='cuda:0'), in_proj_covar=tensor([0.0294, 0.0384, 0.0272, 0.0297, 0.0270, 0.0304, 0.0382, 0.0371], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-16 22:19:54,194 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.998e+02 2.832e+02 3.242e+02 3.996e+02 1.149e+03, threshold=6.485e+02, percent-clipped=5.0 2023-05-16 22:20:03,026 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.74 vs. limit=2.0 2023-05-16 22:20:16,886 INFO [finetune.py:992] (0/2) Epoch 13, batch 5600, loss[loss=0.1472, simple_loss=0.2267, pruned_loss=0.03386, over 12013.00 frames. ], tot_loss[loss=0.167, simple_loss=0.2555, pruned_loss=0.03921, over 2358473.78 frames. ], batch size: 28, lr: 3.69e-03, grad_scale: 8.0 2023-05-16 22:20:20,599 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=252413.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 22:20:24,239 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=252417.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 22:20:36,369 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=252434.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 22:20:52,085 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.0585, 4.4010, 4.0387, 4.7965, 4.3975, 2.9058, 4.1795, 3.1414], device='cuda:0'), covar=tensor([0.0857, 0.0905, 0.1366, 0.0476, 0.1035, 0.1711, 0.0976, 0.2976], device='cuda:0'), in_proj_covar=tensor([0.0304, 0.0377, 0.0356, 0.0311, 0.0368, 0.0270, 0.0342, 0.0365], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 22:20:53,865 INFO [finetune.py:992] (0/2) Epoch 13, batch 5650, loss[loss=0.1681, simple_loss=0.2573, pruned_loss=0.03948, over 12136.00 frames. ], tot_loss[loss=0.1671, simple_loss=0.2554, pruned_loss=0.03938, over 2368273.58 frames. ], batch size: 39, lr: 3.69e-03, grad_scale: 8.0 2023-05-16 22:21:06,632 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.911e+02 2.706e+02 3.163e+02 3.828e+02 6.619e+02, threshold=6.327e+02, percent-clipped=2.0 2023-05-16 22:21:08,273 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=252478.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 22:21:20,521 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=252495.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 22:21:22,143 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2023-05-16 22:21:29,485 INFO [finetune.py:992] (0/2) Epoch 13, batch 5700, loss[loss=0.1729, simple_loss=0.2638, pruned_loss=0.04102, over 11819.00 frames. ], tot_loss[loss=0.1679, simple_loss=0.2566, pruned_loss=0.03954, over 2358921.39 frames. ], batch size: 44, lr: 3.69e-03, grad_scale: 16.0 2023-05-16 22:21:38,569 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.95 vs. limit=2.0 2023-05-16 22:21:41,158 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.75 vs. limit=2.0 2023-05-16 22:22:05,115 INFO [finetune.py:992] (0/2) Epoch 13, batch 5750, loss[loss=0.1688, simple_loss=0.2648, pruned_loss=0.0364, over 12364.00 frames. ], tot_loss[loss=0.1668, simple_loss=0.256, pruned_loss=0.03883, over 2366602.85 frames. ], batch size: 35, lr: 3.69e-03, grad_scale: 16.0 2023-05-16 22:22:18,525 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.751e+02 2.695e+02 3.008e+02 3.623e+02 6.938e+02, threshold=6.016e+02, percent-clipped=1.0 2023-05-16 22:22:29,913 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=252592.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 22:22:33,496 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.0624, 4.9622, 4.8688, 4.9590, 4.4815, 5.0887, 5.0607, 5.2242], device='cuda:0'), covar=tensor([0.0241, 0.0155, 0.0199, 0.0319, 0.0859, 0.0245, 0.0161, 0.0180], device='cuda:0'), in_proj_covar=tensor([0.0197, 0.0199, 0.0194, 0.0250, 0.0246, 0.0221, 0.0182, 0.0230], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0005, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:0') 2023-05-16 22:22:42,016 INFO [finetune.py:992] (0/2) Epoch 13, batch 5800, loss[loss=0.1471, simple_loss=0.2461, pruned_loss=0.02407, over 12359.00 frames. ], tot_loss[loss=0.1665, simple_loss=0.2556, pruned_loss=0.03865, over 2372398.20 frames. ], batch size: 35, lr: 3.69e-03, grad_scale: 16.0 2023-05-16 22:22:45,788 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=252613.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 22:22:49,509 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.5441, 2.5571, 3.1716, 4.3604, 2.5121, 4.5332, 4.6143, 4.5743], device='cuda:0'), covar=tensor([0.0135, 0.1308, 0.0533, 0.0180, 0.1316, 0.0203, 0.0130, 0.0101], device='cuda:0'), in_proj_covar=tensor([0.0122, 0.0203, 0.0184, 0.0121, 0.0191, 0.0180, 0.0175, 0.0122], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 22:23:14,381 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=252653.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 22:23:17,672 INFO [finetune.py:992] (0/2) Epoch 13, batch 5850, loss[loss=0.1581, simple_loss=0.2503, pruned_loss=0.03298, over 12330.00 frames. ], tot_loss[loss=0.1674, simple_loss=0.2568, pruned_loss=0.03897, over 2370668.61 frames. ], batch size: 30, lr: 3.69e-03, grad_scale: 16.0 2023-05-16 22:23:21,085 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.45 vs. limit=2.0 2023-05-16 22:23:25,057 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=252668.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 22:23:29,332 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=252674.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 22:23:30,483 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.053e+02 2.713e+02 3.174e+02 3.869e+02 5.476e+02, threshold=6.348e+02, percent-clipped=0.0 2023-05-16 22:23:36,493 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.4465, 4.2915, 4.2916, 4.6077, 3.2824, 4.2388, 2.8189, 4.3450], device='cuda:0'), covar=tensor([0.1450, 0.0611, 0.0793, 0.0553, 0.1073, 0.0511, 0.1619, 0.0956], device='cuda:0'), in_proj_covar=tensor([0.0228, 0.0261, 0.0294, 0.0355, 0.0239, 0.0239, 0.0257, 0.0371], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 22:23:52,366 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.68 vs. limit=2.0 2023-05-16 22:23:53,392 INFO [finetune.py:992] (0/2) Epoch 13, batch 5900, loss[loss=0.219, simple_loss=0.2937, pruned_loss=0.07213, over 7706.00 frames. ], tot_loss[loss=0.1672, simple_loss=0.2564, pruned_loss=0.03901, over 2371458.79 frames. ], batch size: 97, lr: 3.69e-03, grad_scale: 16.0 2023-05-16 22:24:09,192 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=252729.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 22:24:30,081 INFO [finetune.py:992] (0/2) Epoch 13, batch 5950, loss[loss=0.1443, simple_loss=0.2285, pruned_loss=0.03004, over 12251.00 frames. ], tot_loss[loss=0.1674, simple_loss=0.2566, pruned_loss=0.03911, over 2372362.78 frames. ], batch size: 32, lr: 3.69e-03, grad_scale: 16.0 2023-05-16 22:24:40,870 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=252773.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 22:24:42,853 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.776e+02 2.511e+02 2.995e+02 3.688e+02 9.486e+02, threshold=5.990e+02, percent-clipped=3.0 2023-05-16 22:24:52,787 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=252790.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 22:25:05,587 INFO [finetune.py:992] (0/2) Epoch 13, batch 6000, loss[loss=0.2238, simple_loss=0.3025, pruned_loss=0.07255, over 8528.00 frames. ], tot_loss[loss=0.1683, simple_loss=0.2574, pruned_loss=0.03959, over 2356081.99 frames. ], batch size: 98, lr: 3.69e-03, grad_scale: 16.0 2023-05-16 22:25:05,588 INFO [finetune.py:1017] (0/2) Computing validation loss 2023-05-16 22:25:14,501 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.9868, 4.8896, 5.0059, 5.0455, 4.6170, 4.6066, 4.5808, 4.8340], device='cuda:0'), covar=tensor([0.0743, 0.0564, 0.0685, 0.0517, 0.1769, 0.1419, 0.0498, 0.1184], device='cuda:0'), in_proj_covar=tensor([0.0550, 0.0712, 0.0630, 0.0645, 0.0870, 0.0762, 0.0568, 0.0491], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0005, 0.0004, 0.0004, 0.0005, 0.0005, 0.0004, 0.0003], device='cuda:0') 2023-05-16 22:25:21,403 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.5100, 3.1756, 3.3719, 3.5114, 3.1565, 3.4992, 3.5540, 3.6077], device='cuda:0'), covar=tensor([0.0294, 0.0298, 0.0250, 0.0464, 0.0589, 0.0522, 0.0228, 0.0294], device='cuda:0'), in_proj_covar=tensor([0.0197, 0.0201, 0.0195, 0.0251, 0.0247, 0.0222, 0.0182, 0.0231], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0005, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:0') 2023-05-16 22:25:24,020 INFO [finetune.py:1026] (0/2) Epoch 13, validation: loss=0.3121, simple_loss=0.3891, pruned_loss=0.1176, over 1020973.00 frames. 2023-05-16 22:25:24,021 INFO [finetune.py:1027] (0/2) Maximum memory allocated so far is 12827MB 2023-05-16 22:25:35,079 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.14 vs. limit=2.0 2023-05-16 22:26:00,011 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=252857.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 22:26:00,530 INFO [finetune.py:992] (0/2) Epoch 13, batch 6050, loss[loss=0.1642, simple_loss=0.2495, pruned_loss=0.03949, over 12366.00 frames. ], tot_loss[loss=0.1675, simple_loss=0.2567, pruned_loss=0.03918, over 2362445.98 frames. ], batch size: 30, lr: 3.69e-03, grad_scale: 16.0 2023-05-16 22:26:13,484 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.702e+02 2.667e+02 3.146e+02 3.999e+02 6.169e+02, threshold=6.293e+02, percent-clipped=2.0 2023-05-16 22:26:18,075 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.6396, 2.1021, 3.1942, 2.6886, 3.1487, 2.8796, 2.2296, 3.2022], device='cuda:0'), covar=tensor([0.0229, 0.0534, 0.0304, 0.0337, 0.0254, 0.0244, 0.0478, 0.0191], device='cuda:0'), in_proj_covar=tensor([0.0181, 0.0207, 0.0193, 0.0190, 0.0222, 0.0166, 0.0201, 0.0190], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 22:26:34,117 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.41 vs. limit=2.0 2023-05-16 22:26:36,468 INFO [finetune.py:992] (0/2) Epoch 13, batch 6100, loss[loss=0.1619, simple_loss=0.2549, pruned_loss=0.03443, over 12094.00 frames. ], tot_loss[loss=0.1674, simple_loss=0.2569, pruned_loss=0.03898, over 2366396.65 frames. ], batch size: 33, lr: 3.69e-03, grad_scale: 16.0 2023-05-16 22:26:40,224 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.3819, 4.7852, 2.7753, 2.7430, 4.0518, 2.5030, 3.9903, 3.2763], device='cuda:0'), covar=tensor([0.0662, 0.0404, 0.1183, 0.1457, 0.0277, 0.1385, 0.0439, 0.0741], device='cuda:0'), in_proj_covar=tensor([0.0192, 0.0261, 0.0180, 0.0204, 0.0143, 0.0183, 0.0201, 0.0178], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 22:26:43,682 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=252918.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 22:27:04,755 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=252948.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 22:27:11,729 INFO [finetune.py:992] (0/2) Epoch 13, batch 6150, loss[loss=0.1569, simple_loss=0.2505, pruned_loss=0.03166, over 12142.00 frames. ], tot_loss[loss=0.1678, simple_loss=0.2575, pruned_loss=0.03906, over 2375198.26 frames. ], batch size: 39, lr: 3.69e-03, grad_scale: 16.0 2023-05-16 22:27:20,294 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=252969.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 22:27:25,175 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.798e+02 2.819e+02 3.351e+02 3.995e+02 6.387e+02, threshold=6.701e+02, percent-clipped=1.0 2023-05-16 22:27:38,288 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=252993.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 22:27:49,132 INFO [finetune.py:992] (0/2) Epoch 13, batch 6200, loss[loss=0.152, simple_loss=0.2418, pruned_loss=0.03108, over 12123.00 frames. ], tot_loss[loss=0.167, simple_loss=0.2563, pruned_loss=0.03887, over 2375773.77 frames. ], batch size: 33, lr: 3.69e-03, grad_scale: 16.0 2023-05-16 22:28:00,645 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=253024.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 22:28:22,198 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=253054.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 22:28:24,735 INFO [finetune.py:992] (0/2) Epoch 13, batch 6250, loss[loss=0.136, simple_loss=0.2184, pruned_loss=0.02684, over 12379.00 frames. ], tot_loss[loss=0.1664, simple_loss=0.2556, pruned_loss=0.03863, over 2379222.37 frames. ], batch size: 30, lr: 3.68e-03, grad_scale: 16.0 2023-05-16 22:28:35,763 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=253073.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 22:28:37,686 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.727e+02 2.515e+02 3.088e+02 3.662e+02 5.589e+02, threshold=6.175e+02, percent-clipped=0.0 2023-05-16 22:28:47,984 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=253090.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 22:28:55,227 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.14 vs. limit=2.0 2023-05-16 22:29:01,270 INFO [finetune.py:992] (0/2) Epoch 13, batch 6300, loss[loss=0.1811, simple_loss=0.2725, pruned_loss=0.04479, over 12134.00 frames. ], tot_loss[loss=0.1672, simple_loss=0.2564, pruned_loss=0.03905, over 2373268.56 frames. ], batch size: 38, lr: 3.68e-03, grad_scale: 16.0 2023-05-16 22:29:09,336 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.5656, 2.7580, 3.9166, 4.5433, 3.9496, 4.5938, 3.9188, 3.5083], device='cuda:0'), covar=tensor([0.0046, 0.0346, 0.0119, 0.0039, 0.0109, 0.0060, 0.0108, 0.0294], device='cuda:0'), in_proj_covar=tensor([0.0088, 0.0122, 0.0104, 0.0078, 0.0103, 0.0115, 0.0095, 0.0136], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 22:29:10,666 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=253121.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 22:29:23,159 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=253138.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 22:29:27,142 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.90 vs. limit=2.0 2023-05-16 22:29:37,371 INFO [finetune.py:992] (0/2) Epoch 13, batch 6350, loss[loss=0.1701, simple_loss=0.2541, pruned_loss=0.04302, over 12184.00 frames. ], tot_loss[loss=0.1667, simple_loss=0.2559, pruned_loss=0.03874, over 2372672.72 frames. ], batch size: 31, lr: 3.68e-03, grad_scale: 16.0 2023-05-16 22:29:41,936 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.0342, 4.9744, 4.8042, 4.8801, 4.5612, 5.0489, 5.0803, 5.2581], device='cuda:0'), covar=tensor([0.0227, 0.0136, 0.0189, 0.0343, 0.0736, 0.0234, 0.0139, 0.0168], device='cuda:0'), in_proj_covar=tensor([0.0196, 0.0199, 0.0194, 0.0249, 0.0245, 0.0220, 0.0180, 0.0229], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0005, 0.0004, 0.0004, 0.0003, 0.0005], device='cuda:0') 2023-05-16 22:29:42,095 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2023-05-16 22:29:50,168 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.713e+02 2.572e+02 2.972e+02 3.425e+02 6.380e+02, threshold=5.944e+02, percent-clipped=2.0 2023-05-16 22:30:13,169 INFO [finetune.py:992] (0/2) Epoch 13, batch 6400, loss[loss=0.1605, simple_loss=0.2551, pruned_loss=0.033, over 12157.00 frames. ], tot_loss[loss=0.1667, simple_loss=0.2562, pruned_loss=0.03858, over 2378259.89 frames. ], batch size: 36, lr: 3.68e-03, grad_scale: 16.0 2023-05-16 22:30:16,829 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=253213.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 22:30:42,105 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=253248.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 22:30:49,254 INFO [finetune.py:992] (0/2) Epoch 13, batch 6450, loss[loss=0.1586, simple_loss=0.2521, pruned_loss=0.0325, over 12274.00 frames. ], tot_loss[loss=0.1673, simple_loss=0.2566, pruned_loss=0.039, over 2371433.05 frames. ], batch size: 37, lr: 3.68e-03, grad_scale: 16.0 2023-05-16 22:30:57,224 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=253269.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 22:31:02,829 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.668e+02 2.693e+02 3.152e+02 3.812e+02 6.020e+02, threshold=6.304e+02, percent-clipped=1.0 2023-05-16 22:31:17,221 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=253296.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 22:31:25,741 INFO [finetune.py:992] (0/2) Epoch 13, batch 6500, loss[loss=0.1614, simple_loss=0.2517, pruned_loss=0.03556, over 12302.00 frames. ], tot_loss[loss=0.1675, simple_loss=0.257, pruned_loss=0.03899, over 2372987.79 frames. ], batch size: 34, lr: 3.68e-03, grad_scale: 16.0 2023-05-16 22:31:32,090 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=253317.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 22:31:36,539 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.4710, 4.1549, 4.1668, 4.4621, 3.1712, 4.0189, 2.5660, 4.1249], device='cuda:0'), covar=tensor([0.1522, 0.0659, 0.0860, 0.0694, 0.1128, 0.0602, 0.1900, 0.1069], device='cuda:0'), in_proj_covar=tensor([0.0228, 0.0261, 0.0292, 0.0355, 0.0239, 0.0240, 0.0256, 0.0369], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 22:31:37,141 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=253324.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 22:31:40,794 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=253329.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 22:31:55,031 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=253349.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 22:32:01,404 INFO [finetune.py:992] (0/2) Epoch 13, batch 6550, loss[loss=0.1795, simple_loss=0.2693, pruned_loss=0.04481, over 11827.00 frames. ], tot_loss[loss=0.1676, simple_loss=0.2573, pruned_loss=0.03893, over 2373868.42 frames. ], batch size: 44, lr: 3.68e-03, grad_scale: 16.0 2023-05-16 22:32:11,543 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=253372.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 22:32:12,457 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.3614, 4.0587, 4.2198, 4.6291, 3.1232, 3.9190, 2.5375, 4.2597], device='cuda:0'), covar=tensor([0.1542, 0.0704, 0.0865, 0.0581, 0.1161, 0.0660, 0.1841, 0.1111], device='cuda:0'), in_proj_covar=tensor([0.0229, 0.0263, 0.0295, 0.0358, 0.0242, 0.0242, 0.0259, 0.0373], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 22:32:14,394 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.672e+02 2.653e+02 3.031e+02 3.652e+02 7.428e+02, threshold=6.061e+02, percent-clipped=2.0 2023-05-16 22:32:17,387 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=253380.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 22:32:25,017 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=253390.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 22:32:37,844 INFO [finetune.py:992] (0/2) Epoch 13, batch 6600, loss[loss=0.1522, simple_loss=0.2394, pruned_loss=0.03246, over 12004.00 frames. ], tot_loss[loss=0.1671, simple_loss=0.2566, pruned_loss=0.0388, over 2374385.64 frames. ], batch size: 31, lr: 3.68e-03, grad_scale: 16.0 2023-05-16 22:32:46,508 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=253419.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 22:33:02,104 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=253441.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 22:33:14,102 INFO [finetune.py:992] (0/2) Epoch 13, batch 6650, loss[loss=0.1699, simple_loss=0.2637, pruned_loss=0.03802, over 12039.00 frames. ], tot_loss[loss=0.1663, simple_loss=0.2556, pruned_loss=0.03849, over 2374818.02 frames. ], batch size: 42, lr: 3.68e-03, grad_scale: 16.0 2023-05-16 22:33:14,347 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.0390, 3.7625, 3.8735, 4.0291, 2.5097, 3.7676, 2.6384, 3.6967], device='cuda:0'), covar=tensor([0.1619, 0.0670, 0.0682, 0.0537, 0.1278, 0.0611, 0.1645, 0.0788], device='cuda:0'), in_proj_covar=tensor([0.0231, 0.0265, 0.0297, 0.0361, 0.0243, 0.0244, 0.0261, 0.0376], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 22:33:26,878 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.868e+02 2.716e+02 3.254e+02 3.771e+02 1.116e+03, threshold=6.508e+02, percent-clipped=6.0 2023-05-16 22:33:29,994 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=253480.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 22:33:36,650 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=3.09 vs. limit=5.0 2023-05-16 22:33:49,768 INFO [finetune.py:992] (0/2) Epoch 13, batch 6700, loss[loss=0.1668, simple_loss=0.2594, pruned_loss=0.0371, over 12101.00 frames. ], tot_loss[loss=0.1667, simple_loss=0.256, pruned_loss=0.03876, over 2370731.11 frames. ], batch size: 42, lr: 3.68e-03, grad_scale: 16.0 2023-05-16 22:33:53,439 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=253513.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 22:33:59,861 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=253521.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 22:34:26,704 INFO [finetune.py:992] (0/2) Epoch 13, batch 6750, loss[loss=0.1712, simple_loss=0.2727, pruned_loss=0.03485, over 12066.00 frames. ], tot_loss[loss=0.1662, simple_loss=0.2553, pruned_loss=0.03857, over 2368080.17 frames. ], batch size: 42, lr: 3.68e-03, grad_scale: 16.0 2023-05-16 22:34:28,882 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=253561.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 22:34:33,297 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=253567.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 22:34:39,474 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.668e+02 2.798e+02 3.178e+02 3.722e+02 7.084e+02, threshold=6.356e+02, percent-clipped=0.0 2023-05-16 22:34:43,967 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=253582.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 22:35:02,851 INFO [finetune.py:992] (0/2) Epoch 13, batch 6800, loss[loss=0.2163, simple_loss=0.3016, pruned_loss=0.06549, over 11604.00 frames. ], tot_loss[loss=0.1668, simple_loss=0.2557, pruned_loss=0.03893, over 2357650.69 frames. ], batch size: 48, lr: 3.68e-03, grad_scale: 16.0 2023-05-16 22:35:17,246 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=253628.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 22:35:24,684 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 2023-05-16 22:35:30,228 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.7714, 3.7740, 3.3605, 3.2448, 3.0367, 2.9420, 3.7865, 2.5225], device='cuda:0'), covar=tensor([0.0359, 0.0153, 0.0196, 0.0248, 0.0396, 0.0383, 0.0138, 0.0496], device='cuda:0'), in_proj_covar=tensor([0.0197, 0.0164, 0.0166, 0.0193, 0.0207, 0.0205, 0.0173, 0.0207], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 22:35:32,222 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=253649.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 22:35:33,009 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=253650.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 22:35:38,494 INFO [finetune.py:992] (0/2) Epoch 13, batch 6850, loss[loss=0.147, simple_loss=0.2245, pruned_loss=0.03477, over 11787.00 frames. ], tot_loss[loss=0.1659, simple_loss=0.255, pruned_loss=0.03845, over 2361932.65 frames. ], batch size: 26, lr: 3.68e-03, grad_scale: 16.0 2023-05-16 22:35:52,194 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.819e+02 2.532e+02 3.056e+02 3.723e+02 6.224e+02, threshold=6.112e+02, percent-clipped=1.0 2023-05-16 22:35:58,650 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=253685.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 22:36:07,860 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=253697.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 22:36:15,504 INFO [finetune.py:992] (0/2) Epoch 13, batch 6900, loss[loss=0.1878, simple_loss=0.2735, pruned_loss=0.05107, over 12150.00 frames. ], tot_loss[loss=0.1666, simple_loss=0.2559, pruned_loss=0.03863, over 2368447.36 frames. ], batch size: 39, lr: 3.68e-03, grad_scale: 16.0 2023-05-16 22:36:17,715 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=253711.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 22:36:31,272 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.0743, 4.9420, 4.8799, 4.8932, 4.5969, 5.0626, 5.0648, 5.1938], device='cuda:0'), covar=tensor([0.0226, 0.0162, 0.0175, 0.0389, 0.0769, 0.0348, 0.0164, 0.0202], device='cuda:0'), in_proj_covar=tensor([0.0197, 0.0199, 0.0193, 0.0250, 0.0246, 0.0221, 0.0180, 0.0230], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0005, 0.0004, 0.0004, 0.0003, 0.0005], device='cuda:0') 2023-05-16 22:36:35,351 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=253736.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 22:36:49,259 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.4047, 3.8705, 4.0526, 4.3429, 2.9177, 3.8176, 2.5269, 4.0382], device='cuda:0'), covar=tensor([0.1598, 0.0817, 0.0942, 0.0747, 0.1409, 0.0720, 0.1925, 0.1072], device='cuda:0'), in_proj_covar=tensor([0.0229, 0.0263, 0.0296, 0.0359, 0.0242, 0.0243, 0.0260, 0.0375], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 22:36:51,108 INFO [finetune.py:992] (0/2) Epoch 13, batch 6950, loss[loss=0.175, simple_loss=0.2716, pruned_loss=0.03918, over 11436.00 frames. ], tot_loss[loss=0.1668, simple_loss=0.2559, pruned_loss=0.03884, over 2355940.11 frames. ], batch size: 55, lr: 3.68e-03, grad_scale: 16.0 2023-05-16 22:37:03,238 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=253775.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 22:37:03,824 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.860e+02 2.804e+02 3.198e+02 4.094e+02 1.309e+03, threshold=6.395e+02, percent-clipped=4.0 2023-05-16 22:37:04,371 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.88 vs. limit=2.0 2023-05-16 22:37:27,390 INFO [finetune.py:992] (0/2) Epoch 13, batch 7000, loss[loss=0.1842, simple_loss=0.276, pruned_loss=0.04624, over 11287.00 frames. ], tot_loss[loss=0.166, simple_loss=0.2554, pruned_loss=0.0383, over 2372189.52 frames. ], batch size: 55, lr: 3.68e-03, grad_scale: 16.0 2023-05-16 22:38:03,341 INFO [finetune.py:992] (0/2) Epoch 13, batch 7050, loss[loss=0.1608, simple_loss=0.2519, pruned_loss=0.03489, over 12286.00 frames. ], tot_loss[loss=0.1663, simple_loss=0.2557, pruned_loss=0.03841, over 2379260.27 frames. ], batch size: 33, lr: 3.68e-03, grad_scale: 16.0 2023-05-16 22:38:16,179 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.767e+02 2.719e+02 3.082e+02 3.666e+02 6.671e+02, threshold=6.164e+02, percent-clipped=1.0 2023-05-16 22:38:16,983 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=253877.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 22:38:39,272 INFO [finetune.py:992] (0/2) Epoch 13, batch 7100, loss[loss=0.1578, simple_loss=0.2477, pruned_loss=0.03391, over 12167.00 frames. ], tot_loss[loss=0.1665, simple_loss=0.2559, pruned_loss=0.03857, over 2374851.69 frames. ], batch size: 31, lr: 3.68e-03, grad_scale: 16.0 2023-05-16 22:38:49,901 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=253923.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 22:39:15,637 INFO [finetune.py:992] (0/2) Epoch 13, batch 7150, loss[loss=0.1562, simple_loss=0.2382, pruned_loss=0.03708, over 11798.00 frames. ], tot_loss[loss=0.167, simple_loss=0.2565, pruned_loss=0.03878, over 2369715.43 frames. ], batch size: 26, lr: 3.68e-03, grad_scale: 16.0 2023-05-16 22:39:28,434 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.548e+02 2.775e+02 3.126e+02 3.691e+02 5.581e+02, threshold=6.252e+02, percent-clipped=0.0 2023-05-16 22:39:34,959 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=253985.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 22:39:46,615 INFO [checkpoint.py:75] (0/2) Saving checkpoint to pruned_transducer_stateless7/exp_giga_finetune/checkpoint-154000.pt 2023-05-16 22:39:53,930 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=254006.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 22:39:55,382 INFO [finetune.py:992] (0/2) Epoch 13, batch 7200, loss[loss=0.1474, simple_loss=0.2225, pruned_loss=0.03617, over 11764.00 frames. ], tot_loss[loss=0.1672, simple_loss=0.2562, pruned_loss=0.03905, over 2356277.34 frames. ], batch size: 26, lr: 3.68e-03, grad_scale: 16.0 2023-05-16 22:39:57,357 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.95 vs. limit=2.0 2023-05-16 22:39:58,821 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.75 vs. limit=2.0 2023-05-16 22:40:13,108 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=254033.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 22:40:15,441 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=254036.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 22:40:17,792 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.95 vs. limit=2.0 2023-05-16 22:40:30,710 INFO [finetune.py:992] (0/2) Epoch 13, batch 7250, loss[loss=0.1571, simple_loss=0.2541, pruned_loss=0.03, over 12106.00 frames. ], tot_loss[loss=0.1678, simple_loss=0.2567, pruned_loss=0.03943, over 2360807.25 frames. ], batch size: 33, lr: 3.68e-03, grad_scale: 16.0 2023-05-16 22:40:41,496 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.3357, 6.1474, 5.6907, 5.7555, 6.2189, 5.5574, 5.7714, 5.7273], device='cuda:0'), covar=tensor([0.1459, 0.0970, 0.1140, 0.1890, 0.0907, 0.2106, 0.1719, 0.1055], device='cuda:0'), in_proj_covar=tensor([0.0367, 0.0506, 0.0401, 0.0456, 0.0476, 0.0445, 0.0400, 0.0400], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 22:40:42,438 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.5850, 2.4558, 4.5266, 4.7952, 3.0951, 2.4481, 2.8087, 1.9521], device='cuda:0'), covar=tensor([0.1812, 0.3549, 0.0478, 0.0344, 0.1186, 0.2583, 0.3132, 0.5200], device='cuda:0'), in_proj_covar=tensor([0.0299, 0.0391, 0.0278, 0.0301, 0.0275, 0.0310, 0.0388, 0.0377], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-16 22:40:42,988 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=254075.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 22:40:43,612 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.970e+02 2.641e+02 3.030e+02 3.732e+02 8.374e+02, threshold=6.060e+02, percent-clipped=2.0 2023-05-16 22:40:45,306 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.13 vs. limit=2.0 2023-05-16 22:40:48,273 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=3.27 vs. limit=5.0 2023-05-16 22:40:49,332 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=254084.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 22:41:06,357 INFO [finetune.py:992] (0/2) Epoch 13, batch 7300, loss[loss=0.129, simple_loss=0.2124, pruned_loss=0.02283, over 12296.00 frames. ], tot_loss[loss=0.1675, simple_loss=0.2565, pruned_loss=0.03919, over 2372217.20 frames. ], batch size: 28, lr: 3.68e-03, grad_scale: 16.0 2023-05-16 22:41:17,568 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=254123.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 22:41:43,113 INFO [finetune.py:992] (0/2) Epoch 13, batch 7350, loss[loss=0.1572, simple_loss=0.2501, pruned_loss=0.03214, over 11624.00 frames. ], tot_loss[loss=0.1676, simple_loss=0.2565, pruned_loss=0.03935, over 2368440.00 frames. ], batch size: 48, lr: 3.68e-03, grad_scale: 16.0 2023-05-16 22:41:52,010 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.92 vs. limit=2.0 2023-05-16 22:41:55,833 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.915e+02 2.809e+02 3.297e+02 3.980e+02 6.738e+02, threshold=6.593e+02, percent-clipped=2.0 2023-05-16 22:41:56,626 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=254177.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 22:42:18,907 INFO [finetune.py:992] (0/2) Epoch 13, batch 7400, loss[loss=0.1971, simple_loss=0.288, pruned_loss=0.05314, over 10560.00 frames. ], tot_loss[loss=0.1676, simple_loss=0.2567, pruned_loss=0.03928, over 2374734.22 frames. ], batch size: 68, lr: 3.67e-03, grad_scale: 16.0 2023-05-16 22:42:24,064 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.6989, 4.4818, 4.5231, 4.6731, 4.5665, 4.5927, 4.4602, 2.3899], device='cuda:0'), covar=tensor([0.0243, 0.0107, 0.0150, 0.0095, 0.0079, 0.0208, 0.0153, 0.1108], device='cuda:0'), in_proj_covar=tensor([0.0070, 0.0081, 0.0084, 0.0075, 0.0062, 0.0095, 0.0084, 0.0098], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 22:42:29,735 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=254223.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 22:42:31,077 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=254225.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 22:42:55,247 INFO [finetune.py:992] (0/2) Epoch 13, batch 7450, loss[loss=0.1531, simple_loss=0.2404, pruned_loss=0.03288, over 12342.00 frames. ], tot_loss[loss=0.1671, simple_loss=0.2563, pruned_loss=0.03896, over 2381950.01 frames. ], batch size: 31, lr: 3.67e-03, grad_scale: 16.0 2023-05-16 22:43:04,660 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=254271.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 22:43:08,183 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.982e+02 2.663e+02 2.998e+02 3.648e+02 6.249e+02, threshold=5.996e+02, percent-clipped=0.0 2023-05-16 22:43:17,641 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.37 vs. limit=2.0 2023-05-16 22:43:30,013 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=254306.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 22:43:31,345 INFO [finetune.py:992] (0/2) Epoch 13, batch 7500, loss[loss=0.1441, simple_loss=0.2311, pruned_loss=0.02859, over 12357.00 frames. ], tot_loss[loss=0.1671, simple_loss=0.2564, pruned_loss=0.03892, over 2379812.88 frames. ], batch size: 31, lr: 3.67e-03, grad_scale: 16.0 2023-05-16 22:43:53,958 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.8033, 2.7971, 4.6166, 4.6005, 2.8488, 2.5734, 2.9135, 2.1769], device='cuda:0'), covar=tensor([0.1602, 0.2936, 0.0434, 0.0453, 0.1348, 0.2469, 0.2786, 0.4044], device='cuda:0'), in_proj_covar=tensor([0.0298, 0.0390, 0.0277, 0.0301, 0.0273, 0.0309, 0.0386, 0.0375], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-16 22:44:04,339 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=254354.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 22:44:07,113 INFO [finetune.py:992] (0/2) Epoch 13, batch 7550, loss[loss=0.2145, simple_loss=0.2954, pruned_loss=0.06681, over 7865.00 frames. ], tot_loss[loss=0.1671, simple_loss=0.2562, pruned_loss=0.03898, over 2374800.18 frames. ], batch size: 98, lr: 3.67e-03, grad_scale: 16.0 2023-05-16 22:44:11,028 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.2093, 2.6497, 3.7935, 3.1335, 3.6256, 3.3206, 2.6222, 3.6393], device='cuda:0'), covar=tensor([0.0163, 0.0352, 0.0150, 0.0250, 0.0145, 0.0175, 0.0389, 0.0131], device='cuda:0'), in_proj_covar=tensor([0.0182, 0.0208, 0.0194, 0.0190, 0.0220, 0.0166, 0.0201, 0.0191], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 22:44:19,948 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.867e+02 2.642e+02 3.202e+02 3.788e+02 1.184e+03, threshold=6.403e+02, percent-clipped=7.0 2023-05-16 22:44:30,093 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.3050, 4.9500, 5.2031, 5.1804, 4.8779, 5.1185, 5.0355, 3.0243], device='cuda:0'), covar=tensor([0.0070, 0.0061, 0.0058, 0.0046, 0.0044, 0.0091, 0.0063, 0.0642], device='cuda:0'), in_proj_covar=tensor([0.0070, 0.0081, 0.0085, 0.0075, 0.0062, 0.0095, 0.0084, 0.0099], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 22:44:43,826 INFO [finetune.py:992] (0/2) Epoch 13, batch 7600, loss[loss=0.2211, simple_loss=0.2948, pruned_loss=0.07367, over 7956.00 frames. ], tot_loss[loss=0.167, simple_loss=0.2561, pruned_loss=0.03896, over 2374723.15 frames. ], batch size: 98, lr: 3.67e-03, grad_scale: 16.0 2023-05-16 22:45:13,381 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.5011, 2.6764, 3.3337, 4.3195, 2.5459, 4.3810, 4.4882, 4.5066], device='cuda:0'), covar=tensor([0.0106, 0.1248, 0.0462, 0.0160, 0.1253, 0.0271, 0.0139, 0.0104], device='cuda:0'), in_proj_covar=tensor([0.0122, 0.0201, 0.0183, 0.0120, 0.0188, 0.0180, 0.0174, 0.0121], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 22:45:19,552 INFO [finetune.py:992] (0/2) Epoch 13, batch 7650, loss[loss=0.222, simple_loss=0.2912, pruned_loss=0.07642, over 7955.00 frames. ], tot_loss[loss=0.1682, simple_loss=0.2574, pruned_loss=0.03952, over 2369407.30 frames. ], batch size: 98, lr: 3.67e-03, grad_scale: 16.0 2023-05-16 22:45:32,319 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.837e+02 2.660e+02 3.239e+02 3.711e+02 7.626e+02, threshold=6.478e+02, percent-clipped=3.0 2023-05-16 22:45:55,168 INFO [finetune.py:992] (0/2) Epoch 13, batch 7700, loss[loss=0.1476, simple_loss=0.2313, pruned_loss=0.03196, over 12186.00 frames. ], tot_loss[loss=0.1691, simple_loss=0.2581, pruned_loss=0.04008, over 2362590.78 frames. ], batch size: 31, lr: 3.67e-03, grad_scale: 32.0 2023-05-16 22:46:15,359 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=254536.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 22:46:31,511 INFO [finetune.py:992] (0/2) Epoch 13, batch 7750, loss[loss=0.1527, simple_loss=0.2346, pruned_loss=0.03534, over 12042.00 frames. ], tot_loss[loss=0.1689, simple_loss=0.2579, pruned_loss=0.03997, over 2363519.00 frames. ], batch size: 31, lr: 3.67e-03, grad_scale: 32.0 2023-05-16 22:46:43,342 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.1626, 2.5293, 3.7653, 3.1745, 3.5724, 3.3096, 2.6595, 3.6131], device='cuda:0'), covar=tensor([0.0154, 0.0362, 0.0150, 0.0238, 0.0142, 0.0176, 0.0354, 0.0141], device='cuda:0'), in_proj_covar=tensor([0.0181, 0.0207, 0.0193, 0.0189, 0.0220, 0.0165, 0.0200, 0.0191], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 22:46:44,050 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=254574.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 22:46:45,220 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.963e+02 2.670e+02 3.101e+02 3.752e+02 9.239e+02, threshold=6.202e+02, percent-clipped=1.0 2023-05-16 22:46:57,391 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=4.96 vs. limit=5.0 2023-05-16 22:47:00,576 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=254597.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 22:47:08,420 INFO [finetune.py:992] (0/2) Epoch 13, batch 7800, loss[loss=0.186, simple_loss=0.28, pruned_loss=0.04604, over 12070.00 frames. ], tot_loss[loss=0.1688, simple_loss=0.2577, pruned_loss=0.03995, over 2364303.53 frames. ], batch size: 42, lr: 3.67e-03, grad_scale: 32.0 2023-05-16 22:47:22,306 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.5552, 2.5417, 3.3920, 4.4577, 2.5081, 4.4261, 4.5771, 4.5872], device='cuda:0'), covar=tensor([0.0131, 0.1292, 0.0479, 0.0139, 0.1321, 0.0281, 0.0128, 0.0091], device='cuda:0'), in_proj_covar=tensor([0.0123, 0.0204, 0.0185, 0.0121, 0.0190, 0.0182, 0.0176, 0.0123], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 22:47:28,090 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=254635.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 22:47:29,538 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.6936, 3.6622, 3.2972, 3.2164, 2.8094, 2.7130, 3.6959, 2.3318], device='cuda:0'), covar=tensor([0.0363, 0.0162, 0.0183, 0.0222, 0.0451, 0.0410, 0.0147, 0.0490], device='cuda:0'), in_proj_covar=tensor([0.0198, 0.0168, 0.0169, 0.0195, 0.0209, 0.0207, 0.0176, 0.0210], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 22:47:44,195 INFO [finetune.py:992] (0/2) Epoch 13, batch 7850, loss[loss=0.1712, simple_loss=0.2714, pruned_loss=0.0355, over 11535.00 frames. ], tot_loss[loss=0.1697, simple_loss=0.2588, pruned_loss=0.04034, over 2351075.50 frames. ], batch size: 48, lr: 3.67e-03, grad_scale: 32.0 2023-05-16 22:47:45,456 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=4.79 vs. limit=5.0 2023-05-16 22:47:57,226 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.890e+02 2.782e+02 3.189e+02 3.657e+02 7.453e+02, threshold=6.378e+02, percent-clipped=3.0 2023-05-16 22:48:07,161 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.6592, 4.0661, 4.2162, 4.6439, 3.4420, 3.8664, 2.6947, 4.1419], device='cuda:0'), covar=tensor([0.1503, 0.0746, 0.0824, 0.0589, 0.1092, 0.0710, 0.1797, 0.1071], device='cuda:0'), in_proj_covar=tensor([0.0224, 0.0259, 0.0290, 0.0352, 0.0240, 0.0238, 0.0256, 0.0368], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 22:48:21,220 INFO [finetune.py:992] (0/2) Epoch 13, batch 7900, loss[loss=0.1836, simple_loss=0.2789, pruned_loss=0.04416, over 11195.00 frames. ], tot_loss[loss=0.1707, simple_loss=0.2593, pruned_loss=0.04106, over 2344969.44 frames. ], batch size: 55, lr: 3.67e-03, grad_scale: 32.0 2023-05-16 22:48:54,213 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=254755.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 22:48:56,128 INFO [finetune.py:992] (0/2) Epoch 13, batch 7950, loss[loss=0.1323, simple_loss=0.2138, pruned_loss=0.0254, over 12120.00 frames. ], tot_loss[loss=0.1712, simple_loss=0.2595, pruned_loss=0.04143, over 2343322.56 frames. ], batch size: 30, lr: 3.67e-03, grad_scale: 32.0 2023-05-16 22:49:05,626 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.4396, 2.5352, 3.6391, 4.3318, 3.8228, 4.3036, 3.7776, 2.9377], device='cuda:0'), covar=tensor([0.0035, 0.0403, 0.0136, 0.0046, 0.0127, 0.0080, 0.0116, 0.0394], device='cuda:0'), in_proj_covar=tensor([0.0087, 0.0123, 0.0105, 0.0078, 0.0103, 0.0115, 0.0097, 0.0137], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 22:49:08,892 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.958e+02 2.549e+02 3.114e+02 3.760e+02 7.466e+02, threshold=6.228e+02, percent-clipped=3.0 2023-05-16 22:49:23,979 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=254797.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 22:49:25,442 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.0858, 5.0239, 4.9134, 4.9975, 4.6239, 5.1093, 5.0937, 5.2675], device='cuda:0'), covar=tensor([0.0211, 0.0138, 0.0199, 0.0312, 0.0756, 0.0317, 0.0141, 0.0163], device='cuda:0'), in_proj_covar=tensor([0.0198, 0.0199, 0.0193, 0.0248, 0.0247, 0.0222, 0.0179, 0.0231], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0005, 0.0004, 0.0004, 0.0003, 0.0005], device='cuda:0') 2023-05-16 22:49:31,914 INFO [finetune.py:992] (0/2) Epoch 13, batch 8000, loss[loss=0.1422, simple_loss=0.2294, pruned_loss=0.02744, over 12246.00 frames. ], tot_loss[loss=0.1694, simple_loss=0.2583, pruned_loss=0.04024, over 2353797.01 frames. ], batch size: 32, lr: 3.67e-03, grad_scale: 32.0 2023-05-16 22:49:37,756 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=254816.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 22:49:38,202 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=2.01 vs. limit=2.0 2023-05-16 22:49:47,583 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.8473, 4.6730, 4.7571, 4.8934, 4.6774, 4.8580, 4.7158, 2.5939], device='cuda:0'), covar=tensor([0.0109, 0.0075, 0.0092, 0.0057, 0.0047, 0.0103, 0.0091, 0.0795], device='cuda:0'), in_proj_covar=tensor([0.0070, 0.0081, 0.0084, 0.0075, 0.0061, 0.0095, 0.0084, 0.0098], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 22:49:54,677 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.5510, 5.3339, 5.4528, 5.5221, 5.1105, 5.1544, 4.8699, 5.4249], device='cuda:0'), covar=tensor([0.0666, 0.0580, 0.0854, 0.0585, 0.1954, 0.1234, 0.0532, 0.1035], device='cuda:0'), in_proj_covar=tensor([0.0544, 0.0702, 0.0624, 0.0640, 0.0861, 0.0751, 0.0567, 0.0485], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0005, 0.0005, 0.0003, 0.0003], device='cuda:0') 2023-05-16 22:50:03,293 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=254851.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 22:50:08,793 INFO [finetune.py:992] (0/2) Epoch 13, batch 8050, loss[loss=0.1544, simple_loss=0.2486, pruned_loss=0.03015, over 11864.00 frames. ], tot_loss[loss=0.169, simple_loss=0.2581, pruned_loss=0.03991, over 2358628.83 frames. ], batch size: 44, lr: 3.67e-03, grad_scale: 32.0 2023-05-16 22:50:09,027 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=254858.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 22:50:15,402 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.6794, 5.6861, 5.4154, 5.0450, 5.0470, 5.6229, 5.2342, 5.0611], device='cuda:0'), covar=tensor([0.0808, 0.0862, 0.0756, 0.1591, 0.0761, 0.0780, 0.1617, 0.1048], device='cuda:0'), in_proj_covar=tensor([0.0630, 0.0568, 0.0531, 0.0645, 0.0422, 0.0732, 0.0796, 0.0582], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0004, 0.0004, 0.0003], device='cuda:0') 2023-05-16 22:50:21,576 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.855e+02 2.618e+02 3.061e+02 3.754e+02 7.407e+02, threshold=6.122e+02, percent-clipped=2.0 2023-05-16 22:50:32,926 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=254892.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 22:50:44,218 INFO [finetune.py:992] (0/2) Epoch 13, batch 8100, loss[loss=0.1582, simple_loss=0.2515, pruned_loss=0.03247, over 12125.00 frames. ], tot_loss[loss=0.1693, simple_loss=0.2588, pruned_loss=0.03992, over 2368027.51 frames. ], batch size: 33, lr: 3.67e-03, grad_scale: 32.0 2023-05-16 22:50:47,195 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=254912.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 22:50:59,715 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=254930.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 22:51:19,565 INFO [finetune.py:992] (0/2) Epoch 13, batch 8150, loss[loss=0.188, simple_loss=0.2769, pruned_loss=0.04955, over 11695.00 frames. ], tot_loss[loss=0.1682, simple_loss=0.2577, pruned_loss=0.03937, over 2373471.42 frames. ], batch size: 48, lr: 3.67e-03, grad_scale: 32.0 2023-05-16 22:51:25,029 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-05-16 22:51:33,172 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.698e+02 2.598e+02 3.051e+02 3.699e+02 6.433e+02, threshold=6.102e+02, percent-clipped=1.0 2023-05-16 22:51:43,294 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.9981, 4.7299, 4.8960, 4.9564, 4.8061, 4.9426, 4.8410, 2.6340], device='cuda:0'), covar=tensor([0.0113, 0.0073, 0.0098, 0.0072, 0.0057, 0.0105, 0.0134, 0.0834], device='cuda:0'), in_proj_covar=tensor([0.0070, 0.0080, 0.0083, 0.0075, 0.0061, 0.0094, 0.0083, 0.0097], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 22:51:56,904 INFO [finetune.py:992] (0/2) Epoch 13, batch 8200, loss[loss=0.1571, simple_loss=0.2444, pruned_loss=0.03495, over 12293.00 frames. ], tot_loss[loss=0.1688, simple_loss=0.2583, pruned_loss=0.03962, over 2371754.62 frames. ], batch size: 33, lr: 3.67e-03, grad_scale: 32.0 2023-05-16 22:52:32,410 INFO [finetune.py:992] (0/2) Epoch 13, batch 8250, loss[loss=0.1823, simple_loss=0.2716, pruned_loss=0.04643, over 12128.00 frames. ], tot_loss[loss=0.1692, simple_loss=0.2587, pruned_loss=0.03988, over 2369817.66 frames. ], batch size: 38, lr: 3.67e-03, grad_scale: 32.0 2023-05-16 22:52:45,115 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.396e+02 2.808e+02 3.143e+02 3.527e+02 5.785e+02, threshold=6.286e+02, percent-clipped=0.0 2023-05-16 22:52:55,326 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.2320, 5.1323, 5.0407, 5.0664, 4.7761, 5.2235, 5.2965, 5.3626], device='cuda:0'), covar=tensor([0.0155, 0.0130, 0.0179, 0.0329, 0.0674, 0.0258, 0.0116, 0.0149], device='cuda:0'), in_proj_covar=tensor([0.0199, 0.0200, 0.0194, 0.0251, 0.0248, 0.0224, 0.0180, 0.0232], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0005, 0.0004, 0.0005, 0.0003, 0.0005], device='cuda:0') 2023-05-16 22:53:08,029 INFO [finetune.py:992] (0/2) Epoch 13, batch 8300, loss[loss=0.1734, simple_loss=0.2687, pruned_loss=0.0391, over 11960.00 frames. ], tot_loss[loss=0.1686, simple_loss=0.2582, pruned_loss=0.03947, over 2369953.60 frames. ], batch size: 42, lr: 3.67e-03, grad_scale: 32.0 2023-05-16 22:53:08,997 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.2287, 3.9590, 4.1331, 4.4447, 2.9469, 3.8430, 2.6939, 3.9524], device='cuda:0'), covar=tensor([0.1663, 0.0738, 0.0734, 0.0594, 0.1256, 0.0646, 0.1828, 0.1215], device='cuda:0'), in_proj_covar=tensor([0.0228, 0.0263, 0.0293, 0.0356, 0.0242, 0.0240, 0.0259, 0.0372], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 22:53:10,216 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=255111.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 22:53:11,055 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.4309, 5.0454, 5.3631, 4.7157, 5.0510, 4.8469, 5.3790, 5.0008], device='cuda:0'), covar=tensor([0.0221, 0.0314, 0.0246, 0.0264, 0.0366, 0.0322, 0.0196, 0.0351], device='cuda:0'), in_proj_covar=tensor([0.0266, 0.0271, 0.0293, 0.0266, 0.0269, 0.0269, 0.0240, 0.0221], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 22:53:16,009 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.7466, 2.5133, 4.8456, 5.1606, 3.1076, 2.5980, 2.9466, 2.0152], device='cuda:0'), covar=tensor([0.1699, 0.3917, 0.0423, 0.0294, 0.1213, 0.2505, 0.2979, 0.5043], device='cuda:0'), in_proj_covar=tensor([0.0296, 0.0387, 0.0274, 0.0300, 0.0272, 0.0306, 0.0385, 0.0373], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-16 22:53:25,843 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.6350, 2.7218, 3.8577, 4.6121, 3.9459, 4.6662, 3.9824, 3.3488], device='cuda:0'), covar=tensor([0.0045, 0.0387, 0.0129, 0.0035, 0.0150, 0.0068, 0.0109, 0.0328], device='cuda:0'), in_proj_covar=tensor([0.0087, 0.0122, 0.0103, 0.0077, 0.0103, 0.0114, 0.0096, 0.0136], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 22:53:41,736 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=255153.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 22:53:45,158 INFO [finetune.py:992] (0/2) Epoch 13, batch 8350, loss[loss=0.1496, simple_loss=0.2404, pruned_loss=0.02938, over 12152.00 frames. ], tot_loss[loss=0.1674, simple_loss=0.2572, pruned_loss=0.03878, over 2375224.78 frames. ], batch size: 34, lr: 3.67e-03, grad_scale: 32.0 2023-05-16 22:53:57,868 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.038e+02 2.635e+02 2.966e+02 3.892e+02 6.227e+02, threshold=5.932e+02, percent-clipped=0.0 2023-05-16 22:54:09,516 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=255192.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 22:54:20,369 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=255207.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 22:54:20,965 INFO [finetune.py:992] (0/2) Epoch 13, batch 8400, loss[loss=0.2007, simple_loss=0.2802, pruned_loss=0.06057, over 12051.00 frames. ], tot_loss[loss=0.1679, simple_loss=0.2573, pruned_loss=0.03923, over 2372934.42 frames. ], batch size: 42, lr: 3.67e-03, grad_scale: 32.0 2023-05-16 22:54:36,673 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=255230.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 22:54:43,623 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=255240.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 22:54:57,119 INFO [finetune.py:992] (0/2) Epoch 13, batch 8450, loss[loss=0.1422, simple_loss=0.2186, pruned_loss=0.03295, over 11988.00 frames. ], tot_loss[loss=0.1676, simple_loss=0.2569, pruned_loss=0.03912, over 2376359.14 frames. ], batch size: 28, lr: 3.67e-03, grad_scale: 32.0 2023-05-16 22:55:10,080 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.739e+02 2.590e+02 2.974e+02 3.616e+02 6.901e+02, threshold=5.947e+02, percent-clipped=2.0 2023-05-16 22:55:11,570 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=255278.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 22:55:33,488 INFO [finetune.py:992] (0/2) Epoch 13, batch 8500, loss[loss=0.139, simple_loss=0.2171, pruned_loss=0.03042, over 12182.00 frames. ], tot_loss[loss=0.1674, simple_loss=0.257, pruned_loss=0.03889, over 2380936.64 frames. ], batch size: 29, lr: 3.67e-03, grad_scale: 32.0 2023-05-16 22:55:33,841 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2023-05-16 22:55:43,148 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2023-05-16 22:56:09,128 INFO [finetune.py:992] (0/2) Epoch 13, batch 8550, loss[loss=0.2375, simple_loss=0.3129, pruned_loss=0.08106, over 7521.00 frames. ], tot_loss[loss=0.1669, simple_loss=0.2563, pruned_loss=0.03879, over 2374652.76 frames. ], batch size: 98, lr: 3.67e-03, grad_scale: 32.0 2023-05-16 22:56:19,242 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.3451, 2.0817, 3.7798, 4.3862, 3.9358, 4.2727, 3.8636, 2.7942], device='cuda:0'), covar=tensor([0.0065, 0.0554, 0.0128, 0.0049, 0.0116, 0.0100, 0.0133, 0.0439], device='cuda:0'), in_proj_covar=tensor([0.0087, 0.0123, 0.0103, 0.0077, 0.0103, 0.0114, 0.0096, 0.0136], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 22:56:21,831 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.563e+02 2.802e+02 3.178e+02 3.829e+02 1.090e+03, threshold=6.355e+02, percent-clipped=3.0 2023-05-16 22:56:30,570 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=255388.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 22:56:45,527 INFO [finetune.py:992] (0/2) Epoch 13, batch 8600, loss[loss=0.1973, simple_loss=0.2863, pruned_loss=0.05418, over 12107.00 frames. ], tot_loss[loss=0.1677, simple_loss=0.2572, pruned_loss=0.03914, over 2379255.74 frames. ], batch size: 38, lr: 3.66e-03, grad_scale: 32.0 2023-05-16 22:56:47,789 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=255411.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 22:57:15,769 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=255449.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 22:57:16,450 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.0276, 4.6611, 5.0024, 4.4463, 4.6636, 4.4980, 5.0137, 4.7093], device='cuda:0'), covar=tensor([0.0321, 0.0436, 0.0321, 0.0269, 0.0412, 0.0363, 0.0225, 0.0397], device='cuda:0'), in_proj_covar=tensor([0.0269, 0.0273, 0.0295, 0.0268, 0.0270, 0.0270, 0.0242, 0.0222], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 22:57:18,472 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=255453.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 22:57:21,899 INFO [finetune.py:992] (0/2) Epoch 13, batch 8650, loss[loss=0.1762, simple_loss=0.2728, pruned_loss=0.03981, over 12016.00 frames. ], tot_loss[loss=0.1674, simple_loss=0.2569, pruned_loss=0.03898, over 2377418.36 frames. ], batch size: 40, lr: 3.66e-03, grad_scale: 32.0 2023-05-16 22:57:22,658 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=255459.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 22:57:34,825 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.816e+02 2.718e+02 3.189e+02 3.747e+02 9.115e+02, threshold=6.379e+02, percent-clipped=2.0 2023-05-16 22:57:40,632 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.3724, 5.2118, 5.3407, 5.3645, 4.9744, 5.0173, 4.7380, 5.2757], device='cuda:0'), covar=tensor([0.0740, 0.0553, 0.0749, 0.0581, 0.1821, 0.1364, 0.0551, 0.1100], device='cuda:0'), in_proj_covar=tensor([0.0536, 0.0698, 0.0617, 0.0632, 0.0851, 0.0744, 0.0562, 0.0479], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0005, 0.0005, 0.0003, 0.0003], device='cuda:0') 2023-05-16 22:57:52,693 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=255501.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 22:57:54,851 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.7687, 5.4643, 5.0459, 5.0694, 5.5462, 4.8827, 5.0500, 4.9971], device='cuda:0'), covar=tensor([0.1442, 0.0934, 0.1123, 0.1846, 0.0919, 0.2100, 0.1751, 0.1458], device='cuda:0'), in_proj_covar=tensor([0.0367, 0.0500, 0.0399, 0.0456, 0.0473, 0.0440, 0.0400, 0.0397], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 22:57:57,053 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=255507.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 22:57:57,626 INFO [finetune.py:992] (0/2) Epoch 13, batch 8700, loss[loss=0.1827, simple_loss=0.2754, pruned_loss=0.04495, over 11190.00 frames. ], tot_loss[loss=0.1676, simple_loss=0.2571, pruned_loss=0.03906, over 2378520.35 frames. ], batch size: 55, lr: 3.66e-03, grad_scale: 32.0 2023-05-16 22:58:31,698 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=255555.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 22:58:33,632 INFO [finetune.py:992] (0/2) Epoch 13, batch 8750, loss[loss=0.1715, simple_loss=0.2545, pruned_loss=0.04423, over 12083.00 frames. ], tot_loss[loss=0.1675, simple_loss=0.2566, pruned_loss=0.03914, over 2379507.91 frames. ], batch size: 32, lr: 3.66e-03, grad_scale: 32.0 2023-05-16 22:58:46,348 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.940e+02 2.685e+02 3.096e+02 3.779e+02 7.101e+02, threshold=6.191e+02, percent-clipped=1.0 2023-05-16 22:58:47,626 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.66 vs. limit=2.0 2023-05-16 22:58:49,331 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.2970, 6.2308, 5.7220, 5.8408, 6.2404, 5.5837, 5.8037, 5.7807], device='cuda:0'), covar=tensor([0.1556, 0.0859, 0.1150, 0.1690, 0.0905, 0.2175, 0.1860, 0.1196], device='cuda:0'), in_proj_covar=tensor([0.0365, 0.0500, 0.0399, 0.0456, 0.0473, 0.0439, 0.0399, 0.0397], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 22:59:09,923 INFO [finetune.py:992] (0/2) Epoch 13, batch 8800, loss[loss=0.1977, simple_loss=0.2762, pruned_loss=0.05957, over 12368.00 frames. ], tot_loss[loss=0.1679, simple_loss=0.2566, pruned_loss=0.03959, over 2375136.19 frames. ], batch size: 38, lr: 3.66e-03, grad_scale: 32.0 2023-05-16 22:59:45,640 INFO [finetune.py:992] (0/2) Epoch 13, batch 8850, loss[loss=0.1579, simple_loss=0.2447, pruned_loss=0.03552, over 12104.00 frames. ], tot_loss[loss=0.168, simple_loss=0.2567, pruned_loss=0.03962, over 2379882.17 frames. ], batch size: 33, lr: 3.66e-03, grad_scale: 32.0 2023-05-16 22:59:58,470 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.969e+02 2.704e+02 3.244e+02 3.773e+02 9.691e+02, threshold=6.489e+02, percent-clipped=3.0 2023-05-16 23:00:11,622 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.88 vs. limit=2.0 2023-05-16 23:00:16,429 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.5611, 3.9904, 3.8634, 4.2883, 3.1642, 3.9647, 2.3615, 4.2219], device='cuda:0'), covar=tensor([0.1189, 0.0683, 0.1241, 0.1037, 0.0990, 0.0522, 0.1816, 0.1078], device='cuda:0'), in_proj_covar=tensor([0.0228, 0.0263, 0.0295, 0.0356, 0.0243, 0.0241, 0.0260, 0.0374], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 23:00:21,998 INFO [finetune.py:992] (0/2) Epoch 13, batch 8900, loss[loss=0.1577, simple_loss=0.2549, pruned_loss=0.03029, over 12352.00 frames. ], tot_loss[loss=0.1681, simple_loss=0.2571, pruned_loss=0.03951, over 2379960.36 frames. ], batch size: 36, lr: 3.66e-03, grad_scale: 32.0 2023-05-16 23:00:48,221 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=255744.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 23:00:56,749 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.0646, 4.9389, 4.8577, 4.8778, 4.5369, 4.9751, 5.0888, 5.2201], device='cuda:0'), covar=tensor([0.0241, 0.0149, 0.0199, 0.0392, 0.0778, 0.0351, 0.0140, 0.0183], device='cuda:0'), in_proj_covar=tensor([0.0198, 0.0200, 0.0193, 0.0249, 0.0246, 0.0223, 0.0180, 0.0231], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0005, 0.0004, 0.0004, 0.0003, 0.0005], device='cuda:0') 2023-05-16 23:00:58,024 INFO [finetune.py:992] (0/2) Epoch 13, batch 8950, loss[loss=0.1649, simple_loss=0.2583, pruned_loss=0.03575, over 10660.00 frames. ], tot_loss[loss=0.1684, simple_loss=0.2572, pruned_loss=0.03977, over 2372029.18 frames. ], batch size: 68, lr: 3.66e-03, grad_scale: 32.0 2023-05-16 23:01:10,950 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.075e+02 2.765e+02 3.126e+02 3.642e+02 6.981e+02, threshold=6.253e+02, percent-clipped=2.0 2023-05-16 23:01:30,830 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.4134, 2.3158, 3.6402, 4.2711, 3.8098, 4.2853, 3.8836, 2.8805], device='cuda:0'), covar=tensor([0.0049, 0.0490, 0.0144, 0.0060, 0.0128, 0.0085, 0.0115, 0.0432], device='cuda:0'), in_proj_covar=tensor([0.0087, 0.0123, 0.0104, 0.0078, 0.0103, 0.0114, 0.0096, 0.0136], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 23:01:34,141 INFO [finetune.py:992] (0/2) Epoch 13, batch 9000, loss[loss=0.1833, simple_loss=0.2732, pruned_loss=0.04665, over 12147.00 frames. ], tot_loss[loss=0.1676, simple_loss=0.2564, pruned_loss=0.03941, over 2372149.47 frames. ], batch size: 34, lr: 3.66e-03, grad_scale: 32.0 2023-05-16 23:01:34,141 INFO [finetune.py:1017] (0/2) Computing validation loss 2023-05-16 23:01:52,586 INFO [finetune.py:1026] (0/2) Epoch 13, validation: loss=0.3238, simple_loss=0.3958, pruned_loss=0.1259, over 1020973.00 frames. 2023-05-16 23:01:52,586 INFO [finetune.py:1027] (0/2) Maximum memory allocated so far is 12827MB 2023-05-16 23:01:59,548 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=4.57 vs. limit=5.0 2023-05-16 23:02:28,939 INFO [finetune.py:992] (0/2) Epoch 13, batch 9050, loss[loss=0.1925, simple_loss=0.2952, pruned_loss=0.04493, over 12341.00 frames. ], tot_loss[loss=0.168, simple_loss=0.257, pruned_loss=0.03949, over 2377789.42 frames. ], batch size: 35, lr: 3.66e-03, grad_scale: 32.0 2023-05-16 23:02:37,405 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.2496, 4.3652, 2.7720, 2.3979, 3.7650, 2.4443, 3.8255, 3.0406], device='cuda:0'), covar=tensor([0.0665, 0.0636, 0.1134, 0.1550, 0.0322, 0.1317, 0.0496, 0.0873], device='cuda:0'), in_proj_covar=tensor([0.0190, 0.0260, 0.0177, 0.0202, 0.0144, 0.0180, 0.0201, 0.0177], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 23:02:41,566 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.786e+02 2.848e+02 3.184e+02 4.017e+02 7.938e+02, threshold=6.367e+02, percent-clipped=2.0 2023-05-16 23:03:04,324 INFO [finetune.py:992] (0/2) Epoch 13, batch 9100, loss[loss=0.1667, simple_loss=0.2561, pruned_loss=0.0387, over 12297.00 frames. ], tot_loss[loss=0.1674, simple_loss=0.2566, pruned_loss=0.03914, over 2384689.42 frames. ], batch size: 34, lr: 3.66e-03, grad_scale: 32.0 2023-05-16 23:03:39,425 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=255956.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 23:03:40,614 INFO [finetune.py:992] (0/2) Epoch 13, batch 9150, loss[loss=0.1646, simple_loss=0.2639, pruned_loss=0.03268, over 12355.00 frames. ], tot_loss[loss=0.1682, simple_loss=0.257, pruned_loss=0.03969, over 2374127.52 frames. ], batch size: 36, lr: 3.66e-03, grad_scale: 32.0 2023-05-16 23:03:53,467 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.219e+02 2.849e+02 3.314e+02 4.165e+02 8.053e+02, threshold=6.628e+02, percent-clipped=4.0 2023-05-16 23:04:00,106 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([6.1191, 6.0564, 5.8516, 5.4283, 5.2573, 6.0410, 5.6534, 5.4272], device='cuda:0'), covar=tensor([0.0770, 0.1004, 0.0698, 0.1665, 0.0739, 0.0827, 0.1588, 0.0987], device='cuda:0'), in_proj_covar=tensor([0.0629, 0.0571, 0.0530, 0.0646, 0.0426, 0.0740, 0.0792, 0.0579], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0004, 0.0004, 0.0003], device='cuda:0') 2023-05-16 23:04:11,648 INFO [checkpoint.py:75] (0/2) Saving checkpoint to pruned_transducer_stateless7/exp_giga_finetune/checkpoint-156000.pt 2023-05-16 23:04:20,283 INFO [finetune.py:992] (0/2) Epoch 13, batch 9200, loss[loss=0.1405, simple_loss=0.2314, pruned_loss=0.02483, over 12342.00 frames. ], tot_loss[loss=0.1673, simple_loss=0.2564, pruned_loss=0.03916, over 2373185.17 frames. ], batch size: 30, lr: 3.66e-03, grad_scale: 32.0 2023-05-16 23:04:26,797 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=256017.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 23:04:32,860 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=3.05 vs. limit=5.0 2023-05-16 23:04:42,093 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=256038.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 23:04:46,208 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=256044.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 23:04:56,096 INFO [finetune.py:992] (0/2) Epoch 13, batch 9250, loss[loss=0.1735, simple_loss=0.2722, pruned_loss=0.03739, over 10547.00 frames. ], tot_loss[loss=0.1674, simple_loss=0.2568, pruned_loss=0.03896, over 2377817.58 frames. ], batch size: 68, lr: 3.66e-03, grad_scale: 32.0 2023-05-16 23:05:08,911 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.029e+02 2.567e+02 3.035e+02 3.603e+02 9.664e+02, threshold=6.070e+02, percent-clipped=1.0 2023-05-16 23:05:12,613 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=256081.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 23:05:15,386 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=256084.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 23:05:20,901 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=256092.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 23:05:26,093 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=256099.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 23:05:32,136 INFO [finetune.py:992] (0/2) Epoch 13, batch 9300, loss[loss=0.1558, simple_loss=0.2331, pruned_loss=0.03925, over 12189.00 frames. ], tot_loss[loss=0.1671, simple_loss=0.2568, pruned_loss=0.03875, over 2376953.75 frames. ], batch size: 31, lr: 3.66e-03, grad_scale: 32.0 2023-05-16 23:05:49,869 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=256132.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 23:05:56,809 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=256142.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 23:05:58,971 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=256145.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 23:06:07,902 INFO [finetune.py:992] (0/2) Epoch 13, batch 9350, loss[loss=0.1936, simple_loss=0.2846, pruned_loss=0.0513, over 11368.00 frames. ], tot_loss[loss=0.1674, simple_loss=0.2567, pruned_loss=0.039, over 2375975.82 frames. ], batch size: 55, lr: 3.66e-03, grad_scale: 32.0 2023-05-16 23:06:20,556 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.954e+02 2.667e+02 3.151e+02 3.931e+02 7.899e+02, threshold=6.301e+02, percent-clipped=2.0 2023-05-16 23:06:32,749 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=256193.0, num_to_drop=1, layers_to_drop={2} 2023-05-16 23:06:43,622 INFO [finetune.py:992] (0/2) Epoch 13, batch 9400, loss[loss=0.1521, simple_loss=0.2415, pruned_loss=0.03133, over 12346.00 frames. ], tot_loss[loss=0.1667, simple_loss=0.2557, pruned_loss=0.03883, over 2376771.45 frames. ], batch size: 31, lr: 3.66e-03, grad_scale: 32.0 2023-05-16 23:07:19,741 INFO [finetune.py:992] (0/2) Epoch 13, batch 9450, loss[loss=0.1682, simple_loss=0.2623, pruned_loss=0.03707, over 12118.00 frames. ], tot_loss[loss=0.1672, simple_loss=0.2561, pruned_loss=0.03913, over 2374444.09 frames. ], batch size: 33, lr: 3.66e-03, grad_scale: 32.0 2023-05-16 23:07:33,398 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.869e+02 2.790e+02 3.151e+02 3.840e+02 8.265e+02, threshold=6.302e+02, percent-clipped=1.0 2023-05-16 23:07:39,950 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=256285.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 23:07:43,566 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.6671, 2.7812, 3.2505, 4.5389, 2.3683, 4.3179, 4.6264, 4.7099], device='cuda:0'), covar=tensor([0.0127, 0.1236, 0.0507, 0.0183, 0.1370, 0.0307, 0.0134, 0.0093], device='cuda:0'), in_proj_covar=tensor([0.0124, 0.0205, 0.0186, 0.0121, 0.0191, 0.0183, 0.0178, 0.0124], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 23:07:55,899 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.6514, 5.2673, 5.6245, 5.0145, 5.2804, 5.1126, 5.6304, 5.1970], device='cuda:0'), covar=tensor([0.0216, 0.0313, 0.0244, 0.0210, 0.0342, 0.0299, 0.0204, 0.0235], device='cuda:0'), in_proj_covar=tensor([0.0266, 0.0271, 0.0294, 0.0266, 0.0268, 0.0269, 0.0239, 0.0220], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 23:07:56,469 INFO [finetune.py:992] (0/2) Epoch 13, batch 9500, loss[loss=0.1355, simple_loss=0.2158, pruned_loss=0.02758, over 11780.00 frames. ], tot_loss[loss=0.1678, simple_loss=0.2563, pruned_loss=0.03969, over 2364707.37 frames. ], batch size: 26, lr: 3.66e-03, grad_scale: 32.0 2023-05-16 23:07:59,416 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=256312.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 23:08:23,771 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=256346.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 23:08:32,194 INFO [finetune.py:992] (0/2) Epoch 13, batch 9550, loss[loss=0.1794, simple_loss=0.2588, pruned_loss=0.04996, over 12146.00 frames. ], tot_loss[loss=0.1678, simple_loss=0.2564, pruned_loss=0.03962, over 2365404.05 frames. ], batch size: 36, lr: 3.66e-03, grad_scale: 32.0 2023-05-16 23:08:42,939 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=256372.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 23:08:45,574 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.846e+02 2.580e+02 3.048e+02 3.870e+02 8.344e+02, threshold=6.096e+02, percent-clipped=3.0 2023-05-16 23:08:53,676 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.7487, 2.8290, 4.4779, 4.6410, 2.8955, 2.5885, 2.9547, 2.1116], device='cuda:0'), covar=tensor([0.1542, 0.2907, 0.0464, 0.0431, 0.1288, 0.2299, 0.2717, 0.4044], device='cuda:0'), in_proj_covar=tensor([0.0300, 0.0389, 0.0276, 0.0302, 0.0273, 0.0310, 0.0387, 0.0376], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-16 23:08:58,363 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=256394.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 23:09:08,882 INFO [finetune.py:992] (0/2) Epoch 13, batch 9600, loss[loss=0.1883, simple_loss=0.2788, pruned_loss=0.04891, over 12149.00 frames. ], tot_loss[loss=0.1679, simple_loss=0.2565, pruned_loss=0.03962, over 2366760.88 frames. ], batch size: 34, lr: 3.66e-03, grad_scale: 32.0 2023-05-16 23:09:26,984 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=256433.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 23:09:29,672 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=256437.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 23:09:31,803 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=256440.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 23:09:44,585 INFO [finetune.py:992] (0/2) Epoch 13, batch 9650, loss[loss=0.1823, simple_loss=0.2696, pruned_loss=0.0475, over 12124.00 frames. ], tot_loss[loss=0.1683, simple_loss=0.2573, pruned_loss=0.03965, over 2367744.29 frames. ], batch size: 38, lr: 3.66e-03, grad_scale: 32.0 2023-05-16 23:09:54,097 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=256471.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 23:09:57,456 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.896e+02 2.607e+02 3.105e+02 3.769e+02 8.630e+02, threshold=6.211e+02, percent-clipped=3.0 2023-05-16 23:10:06,184 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=256488.0, num_to_drop=1, layers_to_drop={3} 2023-05-16 23:10:20,829 INFO [finetune.py:992] (0/2) Epoch 13, batch 9700, loss[loss=0.1558, simple_loss=0.2468, pruned_loss=0.03242, over 12349.00 frames. ], tot_loss[loss=0.1674, simple_loss=0.2569, pruned_loss=0.03894, over 2377285.73 frames. ], batch size: 36, lr: 3.66e-03, grad_scale: 64.0 2023-05-16 23:10:38,058 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=256532.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 23:10:45,338 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.3655, 4.9117, 4.2466, 5.1344, 4.5220, 2.4533, 4.0742, 2.9915], device='cuda:0'), covar=tensor([0.0746, 0.0610, 0.1226, 0.0419, 0.1049, 0.2026, 0.1269, 0.3143], device='cuda:0'), in_proj_covar=tensor([0.0306, 0.0381, 0.0362, 0.0322, 0.0372, 0.0274, 0.0348, 0.0371], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 23:10:56,424 INFO [finetune.py:992] (0/2) Epoch 13, batch 9750, loss[loss=0.1803, simple_loss=0.2674, pruned_loss=0.04657, over 12050.00 frames. ], tot_loss[loss=0.167, simple_loss=0.2566, pruned_loss=0.03874, over 2381412.41 frames. ], batch size: 42, lr: 3.66e-03, grad_scale: 32.0 2023-05-16 23:11:10,585 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.927e+02 2.672e+02 3.266e+02 4.073e+02 7.510e+02, threshold=6.532e+02, percent-clipped=3.0 2023-05-16 23:11:26,584 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.8301, 3.0439, 4.7422, 4.8814, 3.1293, 2.7217, 3.1480, 2.1977], device='cuda:0'), covar=tensor([0.1555, 0.2728, 0.0405, 0.0406, 0.1196, 0.2363, 0.2525, 0.4015], device='cuda:0'), in_proj_covar=tensor([0.0301, 0.0391, 0.0277, 0.0304, 0.0275, 0.0312, 0.0389, 0.0379], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-16 23:11:32,947 INFO [finetune.py:992] (0/2) Epoch 13, batch 9800, loss[loss=0.1795, simple_loss=0.2657, pruned_loss=0.0466, over 12311.00 frames. ], tot_loss[loss=0.1663, simple_loss=0.2557, pruned_loss=0.03848, over 2390176.94 frames. ], batch size: 34, lr: 3.65e-03, grad_scale: 32.0 2023-05-16 23:11:36,056 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=256612.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 23:11:56,521 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=256641.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 23:12:08,144 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.86 vs. limit=2.0 2023-05-16 23:12:08,505 INFO [finetune.py:992] (0/2) Epoch 13, batch 9850, loss[loss=0.1942, simple_loss=0.2922, pruned_loss=0.04808, over 10494.00 frames. ], tot_loss[loss=0.1668, simple_loss=0.2564, pruned_loss=0.03858, over 2395490.48 frames. ], batch size: 68, lr: 3.65e-03, grad_scale: 32.0 2023-05-16 23:12:09,968 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=256660.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 23:12:20,044 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.14 vs. limit=2.0 2023-05-16 23:12:22,417 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.684e+02 2.747e+02 3.196e+02 3.714e+02 8.972e+02, threshold=6.391e+02, percent-clipped=3.0 2023-05-16 23:12:34,597 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=256694.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 23:12:44,395 INFO [finetune.py:992] (0/2) Epoch 13, batch 9900, loss[loss=0.1677, simple_loss=0.2609, pruned_loss=0.03723, over 12301.00 frames. ], tot_loss[loss=0.1677, simple_loss=0.257, pruned_loss=0.03927, over 2382331.24 frames. ], batch size: 37, lr: 3.65e-03, grad_scale: 32.0 2023-05-16 23:12:59,670 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=256728.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 23:13:05,886 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=256737.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 23:13:08,025 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=256740.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 23:13:09,278 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=256742.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 23:13:20,319 INFO [finetune.py:992] (0/2) Epoch 13, batch 9950, loss[loss=0.1444, simple_loss=0.225, pruned_loss=0.03191, over 12192.00 frames. ], tot_loss[loss=0.1685, simple_loss=0.2576, pruned_loss=0.03968, over 2375200.48 frames. ], batch size: 29, lr: 3.65e-03, grad_scale: 32.0 2023-05-16 23:13:24,979 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.12 vs. limit=2.0 2023-05-16 23:13:34,444 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.861e+02 2.747e+02 3.108e+02 3.691e+02 6.204e+02, threshold=6.215e+02, percent-clipped=0.0 2023-05-16 23:13:38,368 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.36 vs. limit=2.0 2023-05-16 23:13:39,342 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=256785.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 23:13:41,489 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=256788.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 23:13:41,630 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=256788.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 23:13:56,568 INFO [finetune.py:992] (0/2) Epoch 13, batch 10000, loss[loss=0.1673, simple_loss=0.2626, pruned_loss=0.03602, over 12092.00 frames. ], tot_loss[loss=0.1689, simple_loss=0.2582, pruned_loss=0.03976, over 2370445.29 frames. ], batch size: 32, lr: 3.65e-03, grad_scale: 16.0 2023-05-16 23:14:10,194 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=256827.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 23:14:10,323 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=256827.0, num_to_drop=1, layers_to_drop={0} 2023-05-16 23:14:16,516 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=256836.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 23:14:32,909 INFO [finetune.py:992] (0/2) Epoch 13, batch 10050, loss[loss=0.1661, simple_loss=0.2621, pruned_loss=0.03506, over 12142.00 frames. ], tot_loss[loss=0.1702, simple_loss=0.2594, pruned_loss=0.04047, over 2363142.73 frames. ], batch size: 36, lr: 3.65e-03, grad_scale: 16.0 2023-05-16 23:14:46,915 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.829e+02 2.663e+02 3.175e+02 3.942e+02 8.687e+02, threshold=6.350e+02, percent-clipped=3.0 2023-05-16 23:14:50,669 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=256883.0, num_to_drop=1, layers_to_drop={1} 2023-05-16 23:14:54,250 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=256888.0, num_to_drop=1, layers_to_drop={3} 2023-05-16 23:14:57,134 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=2.33 vs. limit=5.0 2023-05-16 23:15:08,018 INFO [finetune.py:992] (0/2) Epoch 13, batch 10100, loss[loss=0.1314, simple_loss=0.2193, pruned_loss=0.02173, over 11832.00 frames. ], tot_loss[loss=0.1687, simple_loss=0.2578, pruned_loss=0.03982, over 2370071.82 frames. ], batch size: 26, lr: 3.65e-03, grad_scale: 16.0 2023-05-16 23:15:31,683 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=256941.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 23:15:33,905 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=256944.0, num_to_drop=1, layers_to_drop={3} 2023-05-16 23:15:44,637 INFO [finetune.py:992] (0/2) Epoch 13, batch 10150, loss[loss=0.1581, simple_loss=0.242, pruned_loss=0.03714, over 11994.00 frames. ], tot_loss[loss=0.1677, simple_loss=0.2569, pruned_loss=0.0393, over 2366655.71 frames. ], batch size: 28, lr: 3.65e-03, grad_scale: 16.0 2023-05-16 23:15:44,813 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.4172, 5.0267, 5.4062, 4.6676, 5.0215, 4.7652, 5.4249, 5.1162], device='cuda:0'), covar=tensor([0.0275, 0.0388, 0.0259, 0.0241, 0.0381, 0.0349, 0.0196, 0.0260], device='cuda:0'), in_proj_covar=tensor([0.0266, 0.0270, 0.0293, 0.0265, 0.0267, 0.0269, 0.0239, 0.0220], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 23:15:58,633 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.749e+02 2.565e+02 3.029e+02 3.805e+02 7.887e+02, threshold=6.058e+02, percent-clipped=0.0 2023-05-16 23:16:01,054 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=2.26 vs. limit=5.0 2023-05-16 23:16:06,494 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=256989.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 23:16:08,698 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.9455, 5.7054, 5.3172, 5.2305, 5.7576, 5.0692, 5.2381, 5.2016], device='cuda:0'), covar=tensor([0.1394, 0.0939, 0.1060, 0.1988, 0.0940, 0.1885, 0.1629, 0.1307], device='cuda:0'), in_proj_covar=tensor([0.0357, 0.0494, 0.0392, 0.0446, 0.0464, 0.0430, 0.0390, 0.0391], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 23:16:20,978 INFO [finetune.py:992] (0/2) Epoch 13, batch 10200, loss[loss=0.1949, simple_loss=0.2833, pruned_loss=0.05326, over 11783.00 frames. ], tot_loss[loss=0.1671, simple_loss=0.2561, pruned_loss=0.03905, over 2368939.43 frames. ], batch size: 44, lr: 3.65e-03, grad_scale: 16.0 2023-05-16 23:16:24,498 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=257013.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 23:16:34,944 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=257028.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 23:16:55,843 INFO [finetune.py:992] (0/2) Epoch 13, batch 10250, loss[loss=0.1623, simple_loss=0.256, pruned_loss=0.03431, over 12153.00 frames. ], tot_loss[loss=0.1669, simple_loss=0.2559, pruned_loss=0.03894, over 2378293.02 frames. ], batch size: 36, lr: 3.65e-03, grad_scale: 16.0 2023-05-16 23:16:57,488 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.2824, 4.7212, 3.0119, 2.8896, 3.9153, 2.6081, 3.9616, 3.3006], device='cuda:0'), covar=tensor([0.0691, 0.0523, 0.1159, 0.1447, 0.0366, 0.1392, 0.0514, 0.0837], device='cuda:0'), in_proj_covar=tensor([0.0189, 0.0260, 0.0178, 0.0202, 0.0145, 0.0181, 0.0200, 0.0176], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 23:17:07,380 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=257074.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 23:17:08,708 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=257076.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 23:17:10,108 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.095e+02 2.697e+02 3.195e+02 3.744e+02 8.707e+02, threshold=6.389e+02, percent-clipped=4.0 2023-05-16 23:17:22,218 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.2945, 4.7856, 2.9443, 2.7379, 4.0081, 2.5827, 4.0066, 3.2710], device='cuda:0'), covar=tensor([0.0690, 0.0509, 0.1054, 0.1467, 0.0308, 0.1318, 0.0453, 0.0838], device='cuda:0'), in_proj_covar=tensor([0.0189, 0.0259, 0.0177, 0.0201, 0.0144, 0.0180, 0.0199, 0.0176], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 23:17:22,896 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.3673, 4.7836, 3.0013, 2.8663, 3.9990, 2.6966, 4.0258, 3.4525], device='cuda:0'), covar=tensor([0.0611, 0.0411, 0.1009, 0.1258, 0.0331, 0.1089, 0.0413, 0.0659], device='cuda:0'), in_proj_covar=tensor([0.0189, 0.0259, 0.0177, 0.0201, 0.0144, 0.0180, 0.0199, 0.0176], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 23:17:32,062 INFO [finetune.py:992] (0/2) Epoch 13, batch 10300, loss[loss=0.1722, simple_loss=0.268, pruned_loss=0.03823, over 12066.00 frames. ], tot_loss[loss=0.1674, simple_loss=0.2568, pruned_loss=0.03899, over 2371774.33 frames. ], batch size: 40, lr: 3.65e-03, grad_scale: 16.0 2023-05-16 23:17:45,889 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=257127.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 23:17:56,505 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.2082, 3.7049, 3.8009, 4.1671, 2.8737, 3.5759, 2.5689, 3.6656], device='cuda:0'), covar=tensor([0.1565, 0.0770, 0.0942, 0.0706, 0.1161, 0.0694, 0.1861, 0.1175], device='cuda:0'), in_proj_covar=tensor([0.0229, 0.0265, 0.0298, 0.0359, 0.0244, 0.0243, 0.0262, 0.0378], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 23:18:08,398 INFO [finetune.py:992] (0/2) Epoch 13, batch 10350, loss[loss=0.1414, simple_loss=0.2364, pruned_loss=0.02324, over 12155.00 frames. ], tot_loss[loss=0.1674, simple_loss=0.2569, pruned_loss=0.03897, over 2375073.62 frames. ], batch size: 34, lr: 3.65e-03, grad_scale: 16.0 2023-05-16 23:18:13,622 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=257165.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 23:18:17,198 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.2207, 5.2376, 5.0045, 4.5718, 4.6889, 5.1723, 4.8341, 4.6250], device='cuda:0'), covar=tensor([0.0786, 0.0843, 0.0700, 0.1528, 0.1147, 0.0755, 0.1521, 0.1119], device='cuda:0'), in_proj_covar=tensor([0.0632, 0.0565, 0.0529, 0.0645, 0.0425, 0.0739, 0.0788, 0.0581], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0004, 0.0004, 0.0003], device='cuda:0') 2023-05-16 23:18:20,720 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=257175.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 23:18:22,863 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.999e+02 2.630e+02 3.099e+02 3.655e+02 9.989e+02, threshold=6.198e+02, percent-clipped=1.0 2023-05-16 23:18:26,407 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=257183.0, num_to_drop=1, layers_to_drop={2} 2023-05-16 23:18:40,917 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.4682, 5.2960, 5.4783, 5.4893, 5.0879, 5.1310, 4.9063, 5.4030], device='cuda:0'), covar=tensor([0.0757, 0.0603, 0.0803, 0.0536, 0.1909, 0.1378, 0.0554, 0.1036], device='cuda:0'), in_proj_covar=tensor([0.0549, 0.0707, 0.0624, 0.0639, 0.0864, 0.0758, 0.0568, 0.0483], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0005, 0.0004, 0.0004, 0.0005, 0.0005, 0.0003, 0.0003], device='cuda:0') 2023-05-16 23:18:44,320 INFO [finetune.py:992] (0/2) Epoch 13, batch 10400, loss[loss=0.159, simple_loss=0.2534, pruned_loss=0.03231, over 12076.00 frames. ], tot_loss[loss=0.167, simple_loss=0.2567, pruned_loss=0.03865, over 2374562.62 frames. ], batch size: 42, lr: 3.65e-03, grad_scale: 16.0 2023-05-16 23:18:54,523 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=257222.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 23:18:56,863 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=4.85 vs. limit=5.0 2023-05-16 23:18:57,186 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=257226.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 23:19:07,148 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=257239.0, num_to_drop=1, layers_to_drop={3} 2023-05-16 23:19:08,585 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=257241.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 23:19:20,101 INFO [finetune.py:992] (0/2) Epoch 13, batch 10450, loss[loss=0.1886, simple_loss=0.2744, pruned_loss=0.05138, over 12112.00 frames. ], tot_loss[loss=0.1677, simple_loss=0.2572, pruned_loss=0.03908, over 2372249.01 frames. ], batch size: 38, lr: 3.65e-03, grad_scale: 16.0 2023-05-16 23:19:34,113 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.000e+02 2.638e+02 3.175e+02 3.855e+02 7.127e+02, threshold=6.350e+02, percent-clipped=2.0 2023-05-16 23:19:34,987 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.3280, 4.9513, 5.3383, 4.6458, 4.9883, 4.7041, 5.3125, 4.9547], device='cuda:0'), covar=tensor([0.0263, 0.0380, 0.0265, 0.0275, 0.0362, 0.0382, 0.0242, 0.0298], device='cuda:0'), in_proj_covar=tensor([0.0266, 0.0273, 0.0294, 0.0268, 0.0268, 0.0271, 0.0242, 0.0222], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 23:19:38,607 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=257283.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 23:19:52,322 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=257302.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 23:19:55,981 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.6454, 3.2307, 5.1184, 2.5185, 2.9386, 3.8085, 3.1570, 3.8007], device='cuda:0'), covar=tensor([0.0427, 0.1166, 0.0271, 0.1126, 0.1736, 0.1452, 0.1335, 0.1091], device='cuda:0'), in_proj_covar=tensor([0.0236, 0.0235, 0.0256, 0.0182, 0.0236, 0.0294, 0.0226, 0.0264], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-16 23:19:56,449 INFO [finetune.py:992] (0/2) Epoch 13, batch 10500, loss[loss=0.1716, simple_loss=0.2595, pruned_loss=0.04185, over 12057.00 frames. ], tot_loss[loss=0.1666, simple_loss=0.2559, pruned_loss=0.03865, over 2377229.94 frames. ], batch size: 37, lr: 3.65e-03, grad_scale: 16.0 2023-05-16 23:20:25,270 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.80 vs. limit=2.0 2023-05-16 23:20:32,021 INFO [finetune.py:992] (0/2) Epoch 13, batch 10550, loss[loss=0.1707, simple_loss=0.2707, pruned_loss=0.03537, over 12355.00 frames. ], tot_loss[loss=0.1658, simple_loss=0.2549, pruned_loss=0.03833, over 2374739.75 frames. ], batch size: 35, lr: 3.65e-03, grad_scale: 16.0 2023-05-16 23:20:39,927 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=257369.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 23:20:42,568 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=4.82 vs. limit=5.0 2023-05-16 23:20:46,426 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.553e+02 2.617e+02 2.916e+02 3.833e+02 7.684e+02, threshold=5.832e+02, percent-clipped=1.0 2023-05-16 23:21:08,918 INFO [finetune.py:992] (0/2) Epoch 13, batch 10600, loss[loss=0.1497, simple_loss=0.2313, pruned_loss=0.03407, over 12193.00 frames. ], tot_loss[loss=0.1656, simple_loss=0.2544, pruned_loss=0.03839, over 2377196.81 frames. ], batch size: 29, lr: 3.65e-03, grad_scale: 16.0 2023-05-16 23:21:13,407 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.0179, 4.7153, 4.8366, 4.8510, 4.6791, 4.9148, 4.8001, 2.5020], device='cuda:0'), covar=tensor([0.0094, 0.0067, 0.0081, 0.0057, 0.0049, 0.0092, 0.0075, 0.0854], device='cuda:0'), in_proj_covar=tensor([0.0070, 0.0080, 0.0083, 0.0074, 0.0061, 0.0093, 0.0083, 0.0097], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-16 23:21:36,222 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.47 vs. limit=2.0 2023-05-16 23:21:45,037 INFO [finetune.py:992] (0/2) Epoch 13, batch 10650, loss[loss=0.1667, simple_loss=0.2445, pruned_loss=0.04442, over 12354.00 frames. ], tot_loss[loss=0.1657, simple_loss=0.2546, pruned_loss=0.03839, over 2380665.49 frames. ], batch size: 30, lr: 3.65e-03, grad_scale: 16.0 2023-05-16 23:21:51,558 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=257467.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 23:21:59,302 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.949e+02 2.794e+02 3.204e+02 3.744e+02 1.956e+03, threshold=6.408e+02, percent-clipped=5.0 2023-05-16 23:22:00,970 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.4961, 5.0811, 5.4923, 4.8312, 5.0924, 4.8662, 5.5068, 5.1631], device='cuda:0'), covar=tensor([0.0244, 0.0411, 0.0275, 0.0255, 0.0412, 0.0410, 0.0193, 0.0274], device='cuda:0'), in_proj_covar=tensor([0.0269, 0.0275, 0.0298, 0.0270, 0.0271, 0.0272, 0.0244, 0.0225], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 23:22:03,114 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=257483.0, num_to_drop=1, layers_to_drop={0} 2023-05-16 23:22:20,608 INFO [finetune.py:992] (0/2) Epoch 13, batch 10700, loss[loss=0.183, simple_loss=0.2654, pruned_loss=0.05029, over 12142.00 frames. ], tot_loss[loss=0.1664, simple_loss=0.2552, pruned_loss=0.03879, over 2372460.03 frames. ], batch size: 36, lr: 3.65e-03, grad_scale: 16.0 2023-05-16 23:22:25,900 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.1920, 4.6132, 4.1045, 4.9910, 4.4257, 3.0761, 4.2923, 3.0409], device='cuda:0'), covar=tensor([0.0899, 0.0866, 0.1440, 0.0481, 0.1155, 0.1623, 0.1090, 0.3534], device='cuda:0'), in_proj_covar=tensor([0.0309, 0.0383, 0.0363, 0.0322, 0.0374, 0.0274, 0.0350, 0.0372], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 23:22:29,856 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=257521.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 23:22:34,913 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=257528.0, num_to_drop=1, layers_to_drop={2} 2023-05-16 23:22:37,532 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=257531.0, num_to_drop=1, layers_to_drop={1} 2023-05-16 23:22:43,313 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=257539.0, num_to_drop=1, layers_to_drop={0} 2023-05-16 23:22:56,850 INFO [finetune.py:992] (0/2) Epoch 13, batch 10750, loss[loss=0.1392, simple_loss=0.2155, pruned_loss=0.03141, over 11993.00 frames. ], tot_loss[loss=0.1666, simple_loss=0.2551, pruned_loss=0.03901, over 2370134.97 frames. ], batch size: 28, lr: 3.65e-03, grad_scale: 16.0 2023-05-16 23:23:11,968 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.827e+02 2.658e+02 3.256e+02 3.758e+02 5.619e+02, threshold=6.512e+02, percent-clipped=0.0 2023-05-16 23:23:12,081 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=257578.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 23:23:12,130 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=257578.0, num_to_drop=1, layers_to_drop={1} 2023-05-16 23:23:14,672 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.34 vs. limit=2.0 2023-05-16 23:23:15,733 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.7940, 2.7371, 3.8711, 4.5919, 4.0201, 4.6792, 3.9812, 3.1984], device='cuda:0'), covar=tensor([0.0037, 0.0420, 0.0148, 0.0044, 0.0126, 0.0093, 0.0152, 0.0390], device='cuda:0'), in_proj_covar=tensor([0.0090, 0.0127, 0.0108, 0.0081, 0.0108, 0.0119, 0.0100, 0.0142], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-16 23:23:18,424 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=257587.0, num_to_drop=1, layers_to_drop={1} 2023-05-16 23:23:25,381 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=257597.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 23:23:33,278 INFO [finetune.py:992] (0/2) Epoch 13, batch 10800, loss[loss=0.1631, simple_loss=0.2604, pruned_loss=0.03287, over 11357.00 frames. ], tot_loss[loss=0.1667, simple_loss=0.2551, pruned_loss=0.03919, over 2371433.98 frames. ], batch size: 55, lr: 3.65e-03, grad_scale: 16.0 2023-05-16 23:23:36,467 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.14 vs. limit=2.0 2023-05-16 23:23:37,746 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.1885, 2.5926, 3.8182, 3.1836, 3.5366, 3.3015, 2.6544, 3.6159], device='cuda:0'), covar=tensor([0.0131, 0.0376, 0.0129, 0.0236, 0.0160, 0.0199, 0.0391, 0.0139], device='cuda:0'), in_proj_covar=tensor([0.0183, 0.0208, 0.0195, 0.0191, 0.0224, 0.0169, 0.0201, 0.0193], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 23:23:55,452 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=257639.0, num_to_drop=1, layers_to_drop={0} 2023-05-16 23:24:09,016 INFO [finetune.py:992] (0/2) Epoch 13, batch 10850, loss[loss=0.1657, simple_loss=0.2563, pruned_loss=0.03754, over 12154.00 frames. ], tot_loss[loss=0.1678, simple_loss=0.2563, pruned_loss=0.03967, over 2366544.13 frames. ], batch size: 34, lr: 3.65e-03, grad_scale: 16.0 2023-05-16 23:24:10,207 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=4.75 vs. limit=5.0 2023-05-16 23:24:17,002 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=257669.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 23:24:23,598 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.952e+02 2.635e+02 3.079e+02 3.670e+02 7.869e+02, threshold=6.158e+02, percent-clipped=2.0 2023-05-16 23:24:46,261 INFO [finetune.py:992] (0/2) Epoch 13, batch 10900, loss[loss=0.1416, simple_loss=0.2245, pruned_loss=0.02937, over 12270.00 frames. ], tot_loss[loss=0.1672, simple_loss=0.2559, pruned_loss=0.03918, over 2366366.88 frames. ], batch size: 28, lr: 3.65e-03, grad_scale: 16.0 2023-05-16 23:24:52,554 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=257717.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 23:25:14,862 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.9796, 5.9620, 5.7474, 5.2971, 5.2190, 5.9274, 5.5000, 5.3623], device='cuda:0'), covar=tensor([0.0768, 0.0946, 0.0647, 0.1597, 0.0676, 0.0670, 0.1595, 0.0964], device='cuda:0'), in_proj_covar=tensor([0.0633, 0.0567, 0.0529, 0.0645, 0.0426, 0.0741, 0.0793, 0.0580], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0004, 0.0004, 0.0003], device='cuda:0') 2023-05-16 23:25:18,589 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.4896, 2.8336, 3.2173, 4.3955, 2.4054, 4.4280, 4.5174, 4.5779], device='cuda:0'), covar=tensor([0.0158, 0.1153, 0.0528, 0.0163, 0.1318, 0.0236, 0.0165, 0.0095], device='cuda:0'), in_proj_covar=tensor([0.0126, 0.0208, 0.0188, 0.0122, 0.0192, 0.0184, 0.0180, 0.0125], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 23:25:22,537 INFO [finetune.py:992] (0/2) Epoch 13, batch 10950, loss[loss=0.1737, simple_loss=0.2652, pruned_loss=0.0411, over 12115.00 frames. ], tot_loss[loss=0.1676, simple_loss=0.2566, pruned_loss=0.0393, over 2362632.20 frames. ], batch size: 33, lr: 3.65e-03, grad_scale: 16.0 2023-05-16 23:25:36,593 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.858e+02 2.625e+02 3.127e+02 3.912e+02 6.509e+02, threshold=6.254e+02, percent-clipped=2.0 2023-05-16 23:25:43,537 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.3044, 4.8594, 5.2981, 4.6352, 4.9628, 4.6848, 5.3230, 5.0421], device='cuda:0'), covar=tensor([0.0315, 0.0464, 0.0322, 0.0285, 0.0438, 0.0374, 0.0223, 0.0290], device='cuda:0'), in_proj_covar=tensor([0.0266, 0.0271, 0.0295, 0.0267, 0.0268, 0.0269, 0.0242, 0.0222], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 23:25:58,129 INFO [finetune.py:992] (0/2) Epoch 13, batch 11000, loss[loss=0.1832, simple_loss=0.2763, pruned_loss=0.04511, over 12359.00 frames. ], tot_loss[loss=0.1689, simple_loss=0.2579, pruned_loss=0.03992, over 2353570.47 frames. ], batch size: 36, lr: 3.65e-03, grad_scale: 16.0 2023-05-16 23:26:08,359 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=257821.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 23:26:09,763 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=257823.0, num_to_drop=1, layers_to_drop={1} 2023-05-16 23:26:34,544 INFO [finetune.py:992] (0/2) Epoch 13, batch 11050, loss[loss=0.211, simple_loss=0.2871, pruned_loss=0.06739, over 8396.00 frames. ], tot_loss[loss=0.1736, simple_loss=0.2624, pruned_loss=0.04242, over 2307174.71 frames. ], batch size: 97, lr: 3.64e-03, grad_scale: 16.0 2023-05-16 23:26:35,469 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=257859.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 23:26:42,246 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=257869.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 23:26:45,078 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([6.2241, 6.1464, 5.9629, 5.4430, 5.3488, 6.1564, 5.7582, 5.5914], device='cuda:0'), covar=tensor([0.0704, 0.1077, 0.0635, 0.1777, 0.0598, 0.0689, 0.1455, 0.0889], device='cuda:0'), in_proj_covar=tensor([0.0633, 0.0568, 0.0528, 0.0646, 0.0426, 0.0740, 0.0790, 0.0579], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0004, 0.0004, 0.0003], device='cuda:0') 2023-05-16 23:26:48,436 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.191e+02 3.085e+02 3.639e+02 4.467e+02 8.268e+02, threshold=7.277e+02, percent-clipped=4.0 2023-05-16 23:26:48,590 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=257878.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 23:26:52,263 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=257883.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 23:27:01,881 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=257897.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 23:27:06,287 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2023-05-16 23:27:09,260 INFO [finetune.py:992] (0/2) Epoch 13, batch 11100, loss[loss=0.2821, simple_loss=0.3414, pruned_loss=0.1114, over 8104.00 frames. ], tot_loss[loss=0.1791, simple_loss=0.2673, pruned_loss=0.04542, over 2254334.03 frames. ], batch size: 97, lr: 3.64e-03, grad_scale: 16.0 2023-05-16 23:27:18,754 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=257920.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 23:27:22,670 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=257926.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 23:27:24,120 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.7249, 3.1012, 3.4279, 3.6348, 3.5865, 3.5913, 3.4284, 2.6637], device='cuda:0'), covar=tensor([0.0088, 0.0188, 0.0155, 0.0083, 0.0073, 0.0133, 0.0095, 0.0692], device='cuda:0'), in_proj_covar=tensor([0.0069, 0.0080, 0.0082, 0.0074, 0.0061, 0.0093, 0.0083, 0.0096], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-16 23:27:28,908 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=257934.0, num_to_drop=1, layers_to_drop={1} 2023-05-16 23:27:35,828 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=257944.0, num_to_drop=1, layers_to_drop={3} 2023-05-16 23:27:36,350 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=257945.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 23:27:40,038 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=257950.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 23:27:45,406 INFO [finetune.py:992] (0/2) Epoch 13, batch 11150, loss[loss=0.3217, simple_loss=0.3746, pruned_loss=0.1344, over 6812.00 frames. ], tot_loss[loss=0.185, simple_loss=0.2726, pruned_loss=0.04872, over 2198330.26 frames. ], batch size: 98, lr: 3.64e-03, grad_scale: 16.0 2023-05-16 23:27:59,419 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.323e+02 3.230e+02 4.017e+02 4.830e+02 1.321e+03, threshold=8.034e+02, percent-clipped=3.0 2023-05-16 23:28:15,648 INFO [checkpoint.py:75] (0/2) Saving checkpoint to pruned_transducer_stateless7/exp_giga_finetune/checkpoint-158000.pt 2023-05-16 23:28:24,879 INFO [finetune.py:992] (0/2) Epoch 13, batch 11200, loss[loss=0.2921, simple_loss=0.3511, pruned_loss=0.1166, over 7461.00 frames. ], tot_loss[loss=0.193, simple_loss=0.2795, pruned_loss=0.05328, over 2129678.00 frames. ], batch size: 98, lr: 3.64e-03, grad_scale: 16.0 2023-05-16 23:28:27,205 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=258011.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 23:28:49,436 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.5942, 4.5074, 4.6076, 4.6424, 4.3268, 4.4009, 4.2689, 4.4838], device='cuda:0'), covar=tensor([0.0836, 0.0622, 0.0930, 0.0643, 0.1889, 0.1335, 0.0551, 0.1245], device='cuda:0'), in_proj_covar=tensor([0.0533, 0.0692, 0.0605, 0.0624, 0.0840, 0.0736, 0.0553, 0.0473], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0005, 0.0004, 0.0003, 0.0003], device='cuda:0') 2023-05-16 23:28:58,072 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=4.93 vs. limit=5.0 2023-05-16 23:28:59,653 INFO [finetune.py:992] (0/2) Epoch 13, batch 11250, loss[loss=0.2728, simple_loss=0.3299, pruned_loss=0.1078, over 6350.00 frames. ], tot_loss[loss=0.2008, simple_loss=0.2861, pruned_loss=0.05773, over 2071400.69 frames. ], batch size: 97, lr: 3.64e-03, grad_scale: 16.0 2023-05-16 23:29:14,065 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.627e+02 3.593e+02 4.439e+02 5.299e+02 8.794e+02, threshold=8.879e+02, percent-clipped=2.0 2023-05-16 23:29:23,950 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=4.78 vs. limit=5.0 2023-05-16 23:29:35,210 INFO [finetune.py:992] (0/2) Epoch 13, batch 11300, loss[loss=0.276, simple_loss=0.3305, pruned_loss=0.1108, over 6475.00 frames. ], tot_loss[loss=0.2083, simple_loss=0.2925, pruned_loss=0.06203, over 1993876.62 frames. ], batch size: 97, lr: 3.64e-03, grad_scale: 16.0 2023-05-16 23:29:45,509 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=258123.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 23:29:52,560 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 2023-05-16 23:30:00,083 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=2.92 vs. limit=5.0 2023-05-16 23:30:10,569 INFO [finetune.py:992] (0/2) Epoch 13, batch 11350, loss[loss=0.2668, simple_loss=0.3383, pruned_loss=0.09769, over 6348.00 frames. ], tot_loss[loss=0.2148, simple_loss=0.2982, pruned_loss=0.06576, over 1932290.25 frames. ], batch size: 101, lr: 3.64e-03, grad_scale: 16.0 2023-05-16 23:30:19,363 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=258171.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 23:30:23,982 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.439e+02 3.322e+02 4.233e+02 4.928e+02 1.163e+03, threshold=8.466e+02, percent-clipped=1.0 2023-05-16 23:30:33,611 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([1.9264, 2.2364, 2.3193, 2.1879, 1.9506, 1.9052, 2.0834, 1.5663], device='cuda:0'), covar=tensor([0.0342, 0.0177, 0.0160, 0.0193, 0.0311, 0.0238, 0.0187, 0.0400], device='cuda:0'), in_proj_covar=tensor([0.0193, 0.0163, 0.0163, 0.0188, 0.0200, 0.0200, 0.0170, 0.0202], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 23:30:35,098 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.51 vs. limit=2.0 2023-05-16 23:30:37,039 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-05-16 23:30:44,440 INFO [finetune.py:992] (0/2) Epoch 13, batch 11400, loss[loss=0.277, simple_loss=0.338, pruned_loss=0.1079, over 6888.00 frames. ], tot_loss[loss=0.2185, simple_loss=0.3016, pruned_loss=0.06772, over 1909413.22 frames. ], batch size: 98, lr: 3.64e-03, grad_scale: 16.0 2023-05-16 23:30:50,555 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=258215.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 23:31:03,664 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=258234.0, num_to_drop=1, layers_to_drop={0} 2023-05-16 23:31:06,912 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=258239.0, num_to_drop=1, layers_to_drop={0} 2023-05-16 23:31:09,685 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.4501, 3.9835, 4.2165, 4.3409, 4.2612, 4.2776, 4.2572, 2.6367], device='cuda:0'), covar=tensor([0.0075, 0.0136, 0.0115, 0.0064, 0.0052, 0.0122, 0.0073, 0.0855], device='cuda:0'), in_proj_covar=tensor([0.0068, 0.0078, 0.0081, 0.0073, 0.0060, 0.0091, 0.0081, 0.0095], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-16 23:31:19,665 INFO [finetune.py:992] (0/2) Epoch 13, batch 11450, loss[loss=0.2105, simple_loss=0.3061, pruned_loss=0.05746, over 10149.00 frames. ], tot_loss[loss=0.2218, simple_loss=0.304, pruned_loss=0.06984, over 1880394.30 frames. ], batch size: 68, lr: 3.64e-03, grad_scale: 16.0 2023-05-16 23:31:34,364 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.211e+02 3.543e+02 4.076e+02 4.810e+02 8.439e+02, threshold=8.152e+02, percent-clipped=0.0 2023-05-16 23:31:37,114 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=258282.0, num_to_drop=1, layers_to_drop={1} 2023-05-16 23:31:53,344 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=258306.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 23:31:54,652 INFO [finetune.py:992] (0/2) Epoch 13, batch 11500, loss[loss=0.2391, simple_loss=0.3251, pruned_loss=0.07652, over 10387.00 frames. ], tot_loss[loss=0.2248, simple_loss=0.3063, pruned_loss=0.07163, over 1858223.80 frames. ], batch size: 69, lr: 3.64e-03, grad_scale: 16.0 2023-05-16 23:32:30,656 INFO [finetune.py:992] (0/2) Epoch 13, batch 11550, loss[loss=0.2557, simple_loss=0.3256, pruned_loss=0.09286, over 7014.00 frames. ], tot_loss[loss=0.2267, simple_loss=0.3074, pruned_loss=0.07302, over 1823908.36 frames. ], batch size: 101, lr: 3.64e-03, grad_scale: 16.0 2023-05-16 23:32:44,069 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.695e+02 3.630e+02 4.010e+02 4.837e+02 8.435e+02, threshold=8.019e+02, percent-clipped=2.0 2023-05-16 23:33:06,075 INFO [finetune.py:992] (0/2) Epoch 13, batch 11600, loss[loss=0.3006, simple_loss=0.3522, pruned_loss=0.1245, over 6866.00 frames. ], tot_loss[loss=0.229, simple_loss=0.3089, pruned_loss=0.07459, over 1794664.63 frames. ], batch size: 104, lr: 3.64e-03, grad_scale: 16.0 2023-05-16 23:33:39,085 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=258454.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 23:33:41,831 INFO [finetune.py:992] (0/2) Epoch 13, batch 11650, loss[loss=0.2759, simple_loss=0.339, pruned_loss=0.1064, over 7169.00 frames. ], tot_loss[loss=0.2288, simple_loss=0.308, pruned_loss=0.07476, over 1787338.15 frames. ], batch size: 99, lr: 3.64e-03, grad_scale: 16.0 2023-05-16 23:33:57,117 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.268e+02 3.530e+02 3.883e+02 4.638e+02 7.401e+02, threshold=7.767e+02, percent-clipped=0.0 2023-05-16 23:33:57,591 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=4.07 vs. limit=5.0 2023-05-16 23:34:09,169 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.2667, 3.2340, 3.1525, 2.9834, 2.7035, 2.6146, 2.9504, 2.0086], device='cuda:0'), covar=tensor([0.0427, 0.0126, 0.0127, 0.0180, 0.0307, 0.0304, 0.0197, 0.0556], device='cuda:0'), in_proj_covar=tensor([0.0190, 0.0159, 0.0160, 0.0184, 0.0196, 0.0196, 0.0167, 0.0200], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 23:34:16,995 INFO [finetune.py:992] (0/2) Epoch 13, batch 11700, loss[loss=0.1916, simple_loss=0.2773, pruned_loss=0.05293, over 10911.00 frames. ], tot_loss[loss=0.2301, simple_loss=0.3085, pruned_loss=0.07583, over 1761223.03 frames. ], batch size: 55, lr: 3.64e-03, grad_scale: 16.0 2023-05-16 23:34:19,075 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=258511.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 23:34:21,727 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=258515.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 23:34:21,777 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=258515.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 23:34:39,262 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=258539.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 23:34:51,847 INFO [finetune.py:992] (0/2) Epoch 13, batch 11750, loss[loss=0.2478, simple_loss=0.3124, pruned_loss=0.09158, over 6905.00 frames. ], tot_loss[loss=0.2318, simple_loss=0.3098, pruned_loss=0.07692, over 1739104.65 frames. ], batch size: 98, lr: 3.64e-03, grad_scale: 16.0 2023-05-16 23:34:55,251 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=258563.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 23:35:01,714 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=258572.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 23:35:05,453 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.510e+02 3.510e+02 4.130e+02 4.908e+02 7.908e+02, threshold=8.259e+02, percent-clipped=1.0 2023-05-16 23:35:12,900 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=258587.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 23:35:20,280 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.5466, 5.5132, 5.3521, 4.9335, 4.8863, 5.5342, 5.2023, 5.0786], device='cuda:0'), covar=tensor([0.0773, 0.1020, 0.0667, 0.1594, 0.0955, 0.0672, 0.1479, 0.0970], device='cuda:0'), in_proj_covar=tensor([0.0605, 0.0541, 0.0503, 0.0613, 0.0406, 0.0698, 0.0744, 0.0549], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002], device='cuda:0') 2023-05-16 23:35:26,110 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=258606.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 23:35:27,307 INFO [finetune.py:992] (0/2) Epoch 13, batch 11800, loss[loss=0.2632, simple_loss=0.3293, pruned_loss=0.09855, over 6793.00 frames. ], tot_loss[loss=0.2336, simple_loss=0.3114, pruned_loss=0.0779, over 1728088.96 frames. ], batch size: 101, lr: 3.64e-03, grad_scale: 16.0 2023-05-16 23:35:28,781 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.9981, 5.7241, 5.3856, 5.3722, 5.7897, 5.0710, 5.3010, 5.3586], device='cuda:0'), covar=tensor([0.1230, 0.0903, 0.0944, 0.1627, 0.0850, 0.2133, 0.1816, 0.0972], device='cuda:0'), in_proj_covar=tensor([0.0346, 0.0482, 0.0384, 0.0435, 0.0453, 0.0420, 0.0379, 0.0380], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 23:35:47,760 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.14 vs. limit=2.0 2023-05-16 23:36:00,266 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=258654.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 23:36:02,918 INFO [finetune.py:992] (0/2) Epoch 13, batch 11850, loss[loss=0.2453, simple_loss=0.3118, pruned_loss=0.08936, over 6642.00 frames. ], tot_loss[loss=0.2329, simple_loss=0.3114, pruned_loss=0.07721, over 1729630.73 frames. ], batch size: 99, lr: 3.64e-03, grad_scale: 16.0 2023-05-16 23:36:11,946 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.2220, 2.0721, 3.0396, 4.1133, 2.0855, 4.1622, 4.1680, 4.2961], device='cuda:0'), covar=tensor([0.0142, 0.1538, 0.0536, 0.0150, 0.1572, 0.0230, 0.0254, 0.0106], device='cuda:0'), in_proj_covar=tensor([0.0119, 0.0200, 0.0180, 0.0116, 0.0185, 0.0174, 0.0170, 0.0119], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 23:36:16,147 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.3290, 3.4921, 3.1617, 3.5666, 3.3216, 2.5427, 3.1735, 2.8711], device='cuda:0'), covar=tensor([0.0879, 0.1121, 0.1694, 0.0790, 0.1682, 0.1817, 0.1325, 0.3001], device='cuda:0'), in_proj_covar=tensor([0.0291, 0.0360, 0.0342, 0.0299, 0.0353, 0.0261, 0.0330, 0.0351], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-16 23:36:16,523 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.736e+02 3.563e+02 4.265e+02 5.101e+02 7.301e+02, threshold=8.530e+02, percent-clipped=0.0 2023-05-16 23:36:37,164 INFO [finetune.py:992] (0/2) Epoch 13, batch 11900, loss[loss=0.2266, simple_loss=0.3029, pruned_loss=0.07513, over 6679.00 frames. ], tot_loss[loss=0.2312, simple_loss=0.3108, pruned_loss=0.07579, over 1715070.65 frames. ], batch size: 101, lr: 3.64e-03, grad_scale: 16.0 2023-05-16 23:36:41,142 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.39 vs. limit=2.0 2023-05-16 23:36:45,565 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([1.9915, 2.2758, 2.3102, 2.2196, 2.0228, 2.0105, 2.1686, 1.7025], device='cuda:0'), covar=tensor([0.0301, 0.0186, 0.0182, 0.0202, 0.0326, 0.0255, 0.0193, 0.0397], device='cuda:0'), in_proj_covar=tensor([0.0186, 0.0157, 0.0157, 0.0182, 0.0192, 0.0193, 0.0164, 0.0197], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 23:37:10,667 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 2023-05-16 23:37:11,600 INFO [finetune.py:992] (0/2) Epoch 13, batch 11950, loss[loss=0.254, simple_loss=0.3233, pruned_loss=0.09231, over 6973.00 frames. ], tot_loss[loss=0.2269, simple_loss=0.3075, pruned_loss=0.07321, over 1697653.46 frames. ], batch size: 99, lr: 3.64e-03, grad_scale: 16.0 2023-05-16 23:37:26,347 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.223e+02 3.058e+02 3.451e+02 4.238e+02 1.115e+03, threshold=6.902e+02, percent-clipped=1.0 2023-05-16 23:37:46,148 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 2023-05-16 23:37:47,177 INFO [finetune.py:992] (0/2) Epoch 13, batch 12000, loss[loss=0.214, simple_loss=0.3033, pruned_loss=0.06233, over 11127.00 frames. ], tot_loss[loss=0.2213, simple_loss=0.3029, pruned_loss=0.06985, over 1697889.43 frames. ], batch size: 55, lr: 3.64e-03, grad_scale: 32.0 2023-05-16 23:37:47,178 INFO [finetune.py:1017] (0/2) Computing validation loss 2023-05-16 23:38:04,269 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.9854, 3.0823, 2.9881, 3.3004, 3.1223, 2.3466, 2.9835, 2.6464], device='cuda:0'), covar=tensor([0.1043, 0.1216, 0.1659, 0.0775, 0.1385, 0.1762, 0.1153, 0.2946], device='cuda:0'), in_proj_covar=tensor([0.0290, 0.0358, 0.0340, 0.0296, 0.0352, 0.0260, 0.0329, 0.0350], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-16 23:38:05,259 INFO [finetune.py:1026] (0/2) Epoch 13, validation: loss=0.2836, simple_loss=0.3609, pruned_loss=0.1031, over 1020973.00 frames. 2023-05-16 23:38:05,260 INFO [finetune.py:1027] (0/2) Maximum memory allocated so far is 12827MB 2023-05-16 23:38:07,431 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=258810.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 23:38:14,790 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.7855, 3.7929, 3.7973, 3.8653, 3.6709, 3.7131, 3.6453, 3.7766], device='cuda:0'), covar=tensor([0.0982, 0.0706, 0.1210, 0.0674, 0.1670, 0.1118, 0.0541, 0.1009], device='cuda:0'), in_proj_covar=tensor([0.0500, 0.0644, 0.0568, 0.0580, 0.0778, 0.0687, 0.0516, 0.0445], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0003], device='cuda:0') 2023-05-16 23:38:25,610 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-05-16 23:38:39,386 INFO [finetune.py:992] (0/2) Epoch 13, batch 12050, loss[loss=0.2125, simple_loss=0.2967, pruned_loss=0.06415, over 7366.00 frames. ], tot_loss[loss=0.2167, simple_loss=0.2993, pruned_loss=0.06702, over 1697224.46 frames. ], batch size: 99, lr: 3.64e-03, grad_scale: 32.0 2023-05-16 23:38:46,324 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=258867.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 23:38:53,927 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.922e+02 3.018e+02 3.545e+02 4.227e+02 7.592e+02, threshold=7.089e+02, percent-clipped=1.0 2023-05-16 23:39:10,040 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.4534, 2.9803, 3.7694, 2.2747, 2.6637, 2.9962, 2.9481, 3.0145], device='cuda:0'), covar=tensor([0.0464, 0.1088, 0.0340, 0.1347, 0.1786, 0.1576, 0.1254, 0.1409], device='cuda:0'), in_proj_covar=tensor([0.0222, 0.0224, 0.0233, 0.0174, 0.0223, 0.0274, 0.0214, 0.0247], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:0') 2023-05-16 23:39:12,979 INFO [finetune.py:992] (0/2) Epoch 13, batch 12100, loss[loss=0.2117, simple_loss=0.3084, pruned_loss=0.05748, over 11008.00 frames. ], tot_loss[loss=0.2162, simple_loss=0.2991, pruned_loss=0.06665, over 1699924.88 frames. ], batch size: 55, lr: 3.64e-03, grad_scale: 32.0 2023-05-16 23:39:37,117 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.9678, 2.0911, 3.7363, 2.9857, 3.5595, 3.2100, 2.2142, 3.6245], device='cuda:0'), covar=tensor([0.0180, 0.0544, 0.0132, 0.0301, 0.0142, 0.0189, 0.0498, 0.0141], device='cuda:0'), in_proj_covar=tensor([0.0169, 0.0193, 0.0177, 0.0175, 0.0204, 0.0156, 0.0188, 0.0176], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 23:39:45,258 INFO [finetune.py:992] (0/2) Epoch 13, batch 12150, loss[loss=0.1942, simple_loss=0.2834, pruned_loss=0.05246, over 10609.00 frames. ], tot_loss[loss=0.2162, simple_loss=0.299, pruned_loss=0.06672, over 1709850.82 frames. ], batch size: 69, lr: 3.64e-03, grad_scale: 16.0 2023-05-16 23:39:49,843 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=258965.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 23:39:58,685 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.990e+02 3.395e+02 3.927e+02 4.540e+02 8.822e+02, threshold=7.853e+02, percent-clipped=2.0 2023-05-16 23:40:10,040 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=4.51 vs. limit=5.0 2023-05-16 23:40:17,267 INFO [finetune.py:992] (0/2) Epoch 13, batch 12200, loss[loss=0.2504, simple_loss=0.3161, pruned_loss=0.0923, over 6578.00 frames. ], tot_loss[loss=0.2174, simple_loss=0.2998, pruned_loss=0.06754, over 1680554.07 frames. ], batch size: 98, lr: 3.64e-03, grad_scale: 16.0 2023-05-16 23:40:28,898 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=259026.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 23:40:30,468 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.34 vs. limit=2.0 2023-05-16 23:40:39,664 INFO [checkpoint.py:75] (0/2) Saving checkpoint to pruned_transducer_stateless7/exp_giga_finetune/epoch-13.pt 2023-05-16 23:41:03,242 INFO [finetune.py:992] (0/2) Epoch 14, batch 0, loss[loss=0.2012, simple_loss=0.2855, pruned_loss=0.05849, over 12107.00 frames. ], tot_loss[loss=0.2012, simple_loss=0.2855, pruned_loss=0.05849, over 12107.00 frames. ], batch size: 38, lr: 3.63e-03, grad_scale: 16.0 2023-05-16 23:41:03,243 INFO [finetune.py:1017] (0/2) Computing validation loss 2023-05-16 23:41:18,178 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.7756, 2.0033, 2.4192, 2.8475, 2.1379, 2.8834, 2.6885, 2.9322], device='cuda:0'), covar=tensor([0.0225, 0.1155, 0.0489, 0.0243, 0.1082, 0.0371, 0.0389, 0.0184], device='cuda:0'), in_proj_covar=tensor([0.0116, 0.0197, 0.0176, 0.0114, 0.0183, 0.0170, 0.0166, 0.0117], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 23:41:20,625 INFO [finetune.py:1026] (0/2) Epoch 14, validation: loss=0.2862, simple_loss=0.3615, pruned_loss=0.1055, over 1020973.00 frames. 2023-05-16 23:41:20,625 INFO [finetune.py:1027] (0/2) Maximum memory allocated so far is 12827MB 2023-05-16 23:41:47,003 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.004e+02 2.906e+02 3.519e+02 4.209e+02 7.031e+02, threshold=7.039e+02, percent-clipped=0.0 2023-05-16 23:41:57,084 INFO [finetune.py:992] (0/2) Epoch 14, batch 50, loss[loss=0.1743, simple_loss=0.2722, pruned_loss=0.03816, over 12157.00 frames. ], tot_loss[loss=0.1754, simple_loss=0.2658, pruned_loss=0.0425, over 547305.52 frames. ], batch size: 34, lr: 3.63e-03, grad_scale: 16.0 2023-05-16 23:42:10,632 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=259110.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 23:42:15,143 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 2023-05-16 23:42:33,372 INFO [finetune.py:992] (0/2) Epoch 14, batch 100, loss[loss=0.1941, simple_loss=0.2813, pruned_loss=0.05343, over 12040.00 frames. ], tot_loss[loss=0.1735, simple_loss=0.264, pruned_loss=0.04147, over 958511.42 frames. ], batch size: 37, lr: 3.63e-03, grad_scale: 16.0 2023-05-16 23:42:44,742 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=259158.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 23:42:51,243 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=259167.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 23:42:59,571 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.838e+02 2.720e+02 3.079e+02 3.578e+02 6.303e+02, threshold=6.158e+02, percent-clipped=0.0 2023-05-16 23:43:08,596 INFO [finetune.py:992] (0/2) Epoch 14, batch 150, loss[loss=0.2096, simple_loss=0.3043, pruned_loss=0.05745, over 12345.00 frames. ], tot_loss[loss=0.1732, simple_loss=0.264, pruned_loss=0.04123, over 1270304.34 frames. ], batch size: 35, lr: 3.63e-03, grad_scale: 16.0 2023-05-16 23:43:25,194 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=259215.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 23:43:33,190 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=259226.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 23:43:40,977 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.7428, 3.0375, 3.7641, 4.6660, 3.8863, 4.7331, 4.0139, 3.4992], device='cuda:0'), covar=tensor([0.0030, 0.0356, 0.0139, 0.0032, 0.0136, 0.0066, 0.0127, 0.0335], device='cuda:0'), in_proj_covar=tensor([0.0087, 0.0124, 0.0104, 0.0078, 0.0105, 0.0116, 0.0097, 0.0139], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 23:43:41,725 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.5710, 2.6541, 3.3160, 4.4500, 2.4341, 4.3863, 4.5866, 4.6264], device='cuda:0'), covar=tensor([0.0141, 0.1281, 0.0498, 0.0176, 0.1275, 0.0223, 0.0122, 0.0102], device='cuda:0'), in_proj_covar=tensor([0.0118, 0.0200, 0.0179, 0.0116, 0.0186, 0.0173, 0.0169, 0.0119], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 23:43:44,975 INFO [finetune.py:992] (0/2) Epoch 14, batch 200, loss[loss=0.1841, simple_loss=0.2794, pruned_loss=0.04438, over 12056.00 frames. ], tot_loss[loss=0.1724, simple_loss=0.2627, pruned_loss=0.04109, over 1514538.28 frames. ], batch size: 37, lr: 3.63e-03, grad_scale: 16.0 2023-05-16 23:44:01,372 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=259264.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 23:44:11,575 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.938e+02 2.966e+02 3.574e+02 4.503e+02 1.050e+03, threshold=7.149e+02, percent-clipped=8.0 2023-05-16 23:44:17,621 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.0399, 2.1878, 2.6039, 3.0753, 2.2056, 3.1141, 3.0369, 3.1634], device='cuda:0'), covar=tensor([0.0201, 0.1118, 0.0540, 0.0211, 0.1073, 0.0379, 0.0343, 0.0174], device='cuda:0'), in_proj_covar=tensor([0.0118, 0.0200, 0.0179, 0.0115, 0.0186, 0.0173, 0.0169, 0.0119], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 23:44:17,650 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=259287.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 23:44:20,930 INFO [finetune.py:992] (0/2) Epoch 14, batch 250, loss[loss=0.1606, simple_loss=0.2444, pruned_loss=0.03836, over 12259.00 frames. ], tot_loss[loss=0.1727, simple_loss=0.2623, pruned_loss=0.04155, over 1701674.56 frames. ], batch size: 32, lr: 3.63e-03, grad_scale: 16.0 2023-05-16 23:44:41,534 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=259321.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 23:44:44,458 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=259325.0, num_to_drop=1, layers_to_drop={1} 2023-05-16 23:44:56,103 INFO [finetune.py:992] (0/2) Epoch 14, batch 300, loss[loss=0.1784, simple_loss=0.2752, pruned_loss=0.04082, over 12275.00 frames. ], tot_loss[loss=0.1714, simple_loss=0.2609, pruned_loss=0.041, over 1853187.87 frames. ], batch size: 37, lr: 3.63e-03, grad_scale: 16.0 2023-05-16 23:45:22,642 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.780e+02 2.710e+02 3.145e+02 3.769e+02 6.486e+02, threshold=6.290e+02, percent-clipped=0.0 2023-05-16 23:45:24,477 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.37 vs. limit=2.0 2023-05-16 23:45:31,830 INFO [finetune.py:992] (0/2) Epoch 14, batch 350, loss[loss=0.1492, simple_loss=0.229, pruned_loss=0.03463, over 11847.00 frames. ], tot_loss[loss=0.1708, simple_loss=0.2601, pruned_loss=0.04079, over 1972636.46 frames. ], batch size: 26, lr: 3.63e-03, grad_scale: 16.0 2023-05-16 23:45:36,954 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.5913, 4.4524, 4.4302, 4.4875, 4.2013, 4.5357, 4.5409, 4.7992], device='cuda:0'), covar=tensor([0.0227, 0.0183, 0.0238, 0.0436, 0.0728, 0.0416, 0.0201, 0.0203], device='cuda:0'), in_proj_covar=tensor([0.0184, 0.0186, 0.0180, 0.0231, 0.0228, 0.0206, 0.0167, 0.0216], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:0') 2023-05-16 23:45:48,637 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=259414.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 23:46:08,713 INFO [finetune.py:992] (0/2) Epoch 14, batch 400, loss[loss=0.1481, simple_loss=0.2301, pruned_loss=0.03304, over 12358.00 frames. ], tot_loss[loss=0.1693, simple_loss=0.2583, pruned_loss=0.04019, over 2066492.86 frames. ], batch size: 30, lr: 3.63e-03, grad_scale: 16.0 2023-05-16 23:46:32,200 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=259475.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 23:46:34,827 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.823e+02 2.661e+02 3.375e+02 3.946e+02 9.512e+02, threshold=6.750e+02, percent-clipped=4.0 2023-05-16 23:46:44,069 INFO [finetune.py:992] (0/2) Epoch 14, batch 450, loss[loss=0.1469, simple_loss=0.2329, pruned_loss=0.03041, over 12006.00 frames. ], tot_loss[loss=0.1689, simple_loss=0.2579, pruned_loss=0.03994, over 2135070.18 frames. ], batch size: 28, lr: 3.63e-03, grad_scale: 16.0 2023-05-16 23:47:20,177 INFO [finetune.py:992] (0/2) Epoch 14, batch 500, loss[loss=0.1537, simple_loss=0.2356, pruned_loss=0.03586, over 12031.00 frames. ], tot_loss[loss=0.168, simple_loss=0.2567, pruned_loss=0.03959, over 2195655.83 frames. ], batch size: 31, lr: 3.63e-03, grad_scale: 8.0 2023-05-16 23:47:47,785 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.689e+02 2.764e+02 3.286e+02 4.293e+02 4.287e+03, threshold=6.572e+02, percent-clipped=4.0 2023-05-16 23:47:49,407 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=259582.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 23:47:56,270 INFO [finetune.py:992] (0/2) Epoch 14, batch 550, loss[loss=0.1836, simple_loss=0.2788, pruned_loss=0.04418, over 12156.00 frames. ], tot_loss[loss=0.1694, simple_loss=0.2584, pruned_loss=0.04019, over 2225526.19 frames. ], batch size: 34, lr: 3.63e-03, grad_scale: 8.0 2023-05-16 23:48:16,383 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=259620.0, num_to_drop=1, layers_to_drop={0} 2023-05-16 23:48:17,159 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=259621.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 23:48:32,066 INFO [finetune.py:992] (0/2) Epoch 14, batch 600, loss[loss=0.1345, simple_loss=0.2181, pruned_loss=0.02541, over 11777.00 frames. ], tot_loss[loss=0.1691, simple_loss=0.2582, pruned_loss=0.03999, over 2259358.96 frames. ], batch size: 26, lr: 3.63e-03, grad_scale: 8.0 2023-05-16 23:48:51,459 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=259669.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 23:48:59,223 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.757e+02 2.593e+02 3.062e+02 3.666e+02 6.023e+02, threshold=6.124e+02, percent-clipped=0.0 2023-05-16 23:49:07,542 INFO [finetune.py:992] (0/2) Epoch 14, batch 650, loss[loss=0.1812, simple_loss=0.2672, pruned_loss=0.04758, over 12034.00 frames. ], tot_loss[loss=0.1692, simple_loss=0.2589, pruned_loss=0.03972, over 2288655.83 frames. ], batch size: 40, lr: 3.63e-03, grad_scale: 8.0 2023-05-16 23:49:43,305 INFO [finetune.py:992] (0/2) Epoch 14, batch 700, loss[loss=0.1614, simple_loss=0.2498, pruned_loss=0.03653, over 12305.00 frames. ], tot_loss[loss=0.1693, simple_loss=0.2589, pruned_loss=0.03989, over 2306151.32 frames. ], batch size: 34, lr: 3.63e-03, grad_scale: 8.0 2023-05-16 23:49:47,609 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.3173, 2.2603, 2.9877, 4.1953, 2.1764, 4.1042, 4.2326, 4.3506], device='cuda:0'), covar=tensor([0.0137, 0.1461, 0.0589, 0.0154, 0.1421, 0.0330, 0.0186, 0.0096], device='cuda:0'), in_proj_covar=tensor([0.0117, 0.0197, 0.0176, 0.0115, 0.0184, 0.0172, 0.0168, 0.0117], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 23:49:48,258 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=259749.0, num_to_drop=1, layers_to_drop={1} 2023-05-16 23:49:54,321 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=3.07 vs. limit=5.0 2023-05-16 23:50:03,066 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=259770.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 23:50:10,193 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.903e+02 2.727e+02 3.180e+02 3.776e+02 6.703e+02, threshold=6.360e+02, percent-clipped=3.0 2023-05-16 23:50:11,778 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=259782.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 23:50:18,753 INFO [finetune.py:992] (0/2) Epoch 14, batch 750, loss[loss=0.1427, simple_loss=0.2342, pruned_loss=0.02556, over 12176.00 frames. ], tot_loss[loss=0.1681, simple_loss=0.2575, pruned_loss=0.03932, over 2323579.71 frames. ], batch size: 31, lr: 3.63e-03, grad_scale: 8.0 2023-05-16 23:50:32,040 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=259810.0, num_to_drop=1, layers_to_drop={1} 2023-05-16 23:50:52,002 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.62 vs. limit=2.0 2023-05-16 23:50:55,695 INFO [finetune.py:992] (0/2) Epoch 14, batch 800, loss[loss=0.1464, simple_loss=0.2294, pruned_loss=0.03175, over 12245.00 frames. ], tot_loss[loss=0.1685, simple_loss=0.2581, pruned_loss=0.03942, over 2338232.32 frames. ], batch size: 32, lr: 3.63e-03, grad_scale: 8.0 2023-05-16 23:50:56,590 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=259843.0, num_to_drop=1, layers_to_drop={1} 2023-05-16 23:51:06,305 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.3298, 4.9113, 5.3375, 4.6174, 5.0216, 4.8001, 5.3663, 5.1116], device='cuda:0'), covar=tensor([0.0248, 0.0387, 0.0288, 0.0268, 0.0350, 0.0318, 0.0199, 0.0239], device='cuda:0'), in_proj_covar=tensor([0.0257, 0.0260, 0.0283, 0.0259, 0.0259, 0.0259, 0.0234, 0.0211], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-16 23:51:12,749 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=259866.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 23:51:14,137 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.9043, 5.9400, 5.6584, 5.2564, 5.0870, 5.8329, 5.4476, 5.1659], device='cuda:0'), covar=tensor([0.0827, 0.0893, 0.0756, 0.1635, 0.0826, 0.0707, 0.1492, 0.1097], device='cuda:0'), in_proj_covar=tensor([0.0614, 0.0557, 0.0513, 0.0627, 0.0411, 0.0714, 0.0769, 0.0561], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002], device='cuda:0') 2023-05-16 23:51:22,557 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.015e+02 2.584e+02 2.997e+02 3.509e+02 7.772e+02, threshold=5.993e+02, percent-clipped=1.0 2023-05-16 23:51:24,130 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=259882.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 23:51:31,071 INFO [finetune.py:992] (0/2) Epoch 14, batch 850, loss[loss=0.1376, simple_loss=0.2224, pruned_loss=0.02639, over 11987.00 frames. ], tot_loss[loss=0.1676, simple_loss=0.2571, pruned_loss=0.03903, over 2352873.01 frames. ], batch size: 28, lr: 3.63e-03, grad_scale: 8.0 2023-05-16 23:51:37,046 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.6366, 4.8464, 4.2663, 5.2185, 4.7887, 3.3404, 4.5345, 3.1961], device='cuda:0'), covar=tensor([0.0717, 0.0802, 0.1360, 0.0463, 0.1028, 0.1516, 0.0910, 0.3384], device='cuda:0'), in_proj_covar=tensor([0.0304, 0.0376, 0.0356, 0.0311, 0.0366, 0.0272, 0.0342, 0.0366], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 23:51:50,749 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=259920.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 23:51:55,607 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=259927.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 23:51:57,578 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=259930.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 23:52:06,157 INFO [finetune.py:992] (0/2) Epoch 14, batch 900, loss[loss=0.1833, simple_loss=0.2751, pruned_loss=0.04579, over 12135.00 frames. ], tot_loss[loss=0.1673, simple_loss=0.257, pruned_loss=0.03884, over 2360809.67 frames. ], batch size: 39, lr: 3.63e-03, grad_scale: 8.0 2023-05-16 23:52:24,349 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=259968.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 23:52:33,418 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.861e+02 2.713e+02 3.274e+02 3.740e+02 6.362e+02, threshold=6.548e+02, percent-clipped=1.0 2023-05-16 23:52:37,727 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.4973, 5.2857, 5.4239, 5.4543, 5.1032, 5.1474, 4.8653, 5.4271], device='cuda:0'), covar=tensor([0.0680, 0.0654, 0.0736, 0.0578, 0.1994, 0.1402, 0.0523, 0.0882], device='cuda:0'), in_proj_covar=tensor([0.0525, 0.0677, 0.0596, 0.0610, 0.0816, 0.0723, 0.0542, 0.0463], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0005, 0.0004, 0.0003, 0.0003], device='cuda:0') 2023-05-16 23:52:41,906 INFO [finetune.py:992] (0/2) Epoch 14, batch 950, loss[loss=0.152, simple_loss=0.2429, pruned_loss=0.0306, over 12238.00 frames. ], tot_loss[loss=0.1677, simple_loss=0.2574, pruned_loss=0.03899, over 2360555.34 frames. ], batch size: 32, lr: 3.62e-03, grad_scale: 8.0 2023-05-16 23:52:46,989 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.1175, 2.3855, 3.5996, 3.0452, 3.3974, 3.1595, 2.4451, 3.5180], device='cuda:0'), covar=tensor([0.0122, 0.0389, 0.0124, 0.0220, 0.0158, 0.0177, 0.0340, 0.0131], device='cuda:0'), in_proj_covar=tensor([0.0176, 0.0203, 0.0188, 0.0185, 0.0214, 0.0164, 0.0197, 0.0185], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 23:52:48,482 INFO [checkpoint.py:75] (0/2) Saving checkpoint to pruned_transducer_stateless7/exp_giga_finetune/checkpoint-160000.pt 2023-05-16 23:53:05,639 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 2023-05-16 23:53:13,203 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=260030.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 23:53:21,471 INFO [finetune.py:992] (0/2) Epoch 14, batch 1000, loss[loss=0.1877, simple_loss=0.2766, pruned_loss=0.0494, over 11766.00 frames. ], tot_loss[loss=0.1691, simple_loss=0.2589, pruned_loss=0.03967, over 2361703.68 frames. ], batch size: 44, lr: 3.62e-03, grad_scale: 8.0 2023-05-16 23:53:40,876 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.2911, 2.7359, 3.7263, 3.2199, 3.6102, 3.2348, 2.7211, 3.6715], device='cuda:0'), covar=tensor([0.0144, 0.0339, 0.0169, 0.0224, 0.0160, 0.0184, 0.0331, 0.0122], device='cuda:0'), in_proj_covar=tensor([0.0177, 0.0203, 0.0188, 0.0185, 0.0214, 0.0164, 0.0197, 0.0185], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 23:53:41,489 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=260070.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 23:53:48,607 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.902e+02 2.539e+02 3.055e+02 3.631e+02 6.681e+02, threshold=6.111e+02, percent-clipped=1.0 2023-05-16 23:53:50,462 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.61 vs. limit=2.0 2023-05-16 23:53:54,693 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.3143, 3.2877, 4.6624, 2.5684, 2.7303, 3.5331, 3.0720, 3.5413], device='cuda:0'), covar=tensor([0.0482, 0.1114, 0.0380, 0.1185, 0.1919, 0.1464, 0.1384, 0.1311], device='cuda:0'), in_proj_covar=tensor([0.0234, 0.0234, 0.0248, 0.0183, 0.0235, 0.0289, 0.0224, 0.0260], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-16 23:53:56,641 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=260091.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 23:53:57,144 INFO [finetune.py:992] (0/2) Epoch 14, batch 1050, loss[loss=0.1698, simple_loss=0.2669, pruned_loss=0.03633, over 12364.00 frames. ], tot_loss[loss=0.1683, simple_loss=0.2579, pruned_loss=0.03931, over 2370822.83 frames. ], batch size: 35, lr: 3.62e-03, grad_scale: 8.0 2023-05-16 23:53:57,349 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.3382, 3.5281, 3.2705, 3.1659, 2.7477, 2.6388, 3.6507, 2.0971], device='cuda:0'), covar=tensor([0.0443, 0.0149, 0.0183, 0.0191, 0.0426, 0.0404, 0.0133, 0.0549], device='cuda:0'), in_proj_covar=tensor([0.0193, 0.0161, 0.0162, 0.0186, 0.0199, 0.0198, 0.0169, 0.0204], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 23:54:06,280 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=260105.0, num_to_drop=1, layers_to_drop={2} 2023-05-16 23:54:15,382 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=260118.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 23:54:20,523 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.3464, 5.1401, 5.3198, 5.3319, 4.9653, 5.0009, 4.7171, 5.1986], device='cuda:0'), covar=tensor([0.0737, 0.0659, 0.0933, 0.0579, 0.1954, 0.1652, 0.0627, 0.1243], device='cuda:0'), in_proj_covar=tensor([0.0532, 0.0684, 0.0602, 0.0617, 0.0824, 0.0730, 0.0547, 0.0467], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0005, 0.0004, 0.0003, 0.0003], device='cuda:0') 2023-05-16 23:54:31,309 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=260138.0, num_to_drop=1, layers_to_drop={0} 2023-05-16 23:54:33,971 INFO [finetune.py:992] (0/2) Epoch 14, batch 1100, loss[loss=0.1531, simple_loss=0.2464, pruned_loss=0.02994, over 12367.00 frames. ], tot_loss[loss=0.168, simple_loss=0.2577, pruned_loss=0.03919, over 2370392.83 frames. ], batch size: 35, lr: 3.62e-03, grad_scale: 8.0 2023-05-16 23:55:01,289 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.964e+02 2.617e+02 3.016e+02 3.739e+02 7.886e+02, threshold=6.033e+02, percent-clipped=2.0 2023-05-16 23:55:07,118 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.3471, 4.1290, 4.0462, 4.3217, 2.8077, 3.8488, 2.5192, 4.1730], device='cuda:0'), covar=tensor([0.1579, 0.0671, 0.0930, 0.0714, 0.1280, 0.0646, 0.1973, 0.1051], device='cuda:0'), in_proj_covar=tensor([0.0230, 0.0264, 0.0295, 0.0358, 0.0242, 0.0241, 0.0262, 0.0375], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 23:55:09,763 INFO [finetune.py:992] (0/2) Epoch 14, batch 1150, loss[loss=0.1565, simple_loss=0.2465, pruned_loss=0.03319, over 12073.00 frames. ], tot_loss[loss=0.167, simple_loss=0.2565, pruned_loss=0.03873, over 2378411.39 frames. ], batch size: 32, lr: 3.62e-03, grad_scale: 8.0 2023-05-16 23:55:15,052 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.2687, 2.7044, 3.7613, 3.1796, 3.5411, 3.2817, 2.6185, 3.6414], device='cuda:0'), covar=tensor([0.0138, 0.0349, 0.0131, 0.0281, 0.0173, 0.0186, 0.0359, 0.0133], device='cuda:0'), in_proj_covar=tensor([0.0177, 0.0203, 0.0188, 0.0185, 0.0214, 0.0163, 0.0197, 0.0185], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 23:55:31,780 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=260222.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 23:55:33,375 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=260224.0, num_to_drop=1, layers_to_drop={1} 2023-05-16 23:55:45,482 INFO [finetune.py:992] (0/2) Epoch 14, batch 1200, loss[loss=0.2035, simple_loss=0.2805, pruned_loss=0.06328, over 12110.00 frames. ], tot_loss[loss=0.1673, simple_loss=0.257, pruned_loss=0.0388, over 2384520.39 frames. ], batch size: 38, lr: 3.62e-03, grad_scale: 8.0 2023-05-16 23:56:12,508 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.935e+02 2.690e+02 3.229e+02 3.657e+02 1.170e+03, threshold=6.458e+02, percent-clipped=4.0 2023-05-16 23:56:17,506 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=260285.0, num_to_drop=1, layers_to_drop={2} 2023-05-16 23:56:22,321 INFO [finetune.py:992] (0/2) Epoch 14, batch 1250, loss[loss=0.1626, simple_loss=0.252, pruned_loss=0.03663, over 10536.00 frames. ], tot_loss[loss=0.1679, simple_loss=0.2576, pruned_loss=0.03908, over 2373521.60 frames. ], batch size: 68, lr: 3.62e-03, grad_scale: 8.0 2023-05-16 23:56:50,088 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=2.85 vs. limit=5.0 2023-05-16 23:56:57,483 INFO [finetune.py:992] (0/2) Epoch 14, batch 1300, loss[loss=0.1886, simple_loss=0.277, pruned_loss=0.05015, over 12058.00 frames. ], tot_loss[loss=0.1675, simple_loss=0.2569, pruned_loss=0.03908, over 2379799.90 frames. ], batch size: 42, lr: 3.62e-03, grad_scale: 8.0 2023-05-16 23:57:24,494 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.846e+02 2.744e+02 3.203e+02 3.844e+02 2.936e+03, threshold=6.405e+02, percent-clipped=4.0 2023-05-16 23:57:28,890 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=260386.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 23:57:29,023 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.6136, 3.0895, 3.5809, 4.5246, 4.0516, 4.6440, 3.9671, 3.3277], device='cuda:0'), covar=tensor([0.0039, 0.0317, 0.0141, 0.0041, 0.0103, 0.0060, 0.0120, 0.0337], device='cuda:0'), in_proj_covar=tensor([0.0088, 0.0124, 0.0105, 0.0079, 0.0104, 0.0116, 0.0097, 0.0139], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-16 23:57:33,185 INFO [finetune.py:992] (0/2) Epoch 14, batch 1350, loss[loss=0.1717, simple_loss=0.2616, pruned_loss=0.04094, over 12150.00 frames. ], tot_loss[loss=0.1678, simple_loss=0.257, pruned_loss=0.03929, over 2381933.75 frames. ], batch size: 34, lr: 3.62e-03, grad_scale: 8.0 2023-05-16 23:57:42,857 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=260405.0, num_to_drop=1, layers_to_drop={1} 2023-05-16 23:57:49,381 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.8826, 4.8216, 4.6632, 4.8265, 4.5065, 4.9805, 4.9531, 5.1210], device='cuda:0'), covar=tensor([0.0303, 0.0161, 0.0208, 0.0330, 0.0724, 0.0303, 0.0176, 0.0187], device='cuda:0'), in_proj_covar=tensor([0.0193, 0.0195, 0.0189, 0.0242, 0.0239, 0.0217, 0.0175, 0.0225], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:0') 2023-05-16 23:58:04,398 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=4.71 vs. limit=5.0 2023-05-16 23:58:08,046 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=260438.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 23:58:10,851 INFO [finetune.py:992] (0/2) Epoch 14, batch 1400, loss[loss=0.2154, simple_loss=0.2867, pruned_loss=0.07204, over 8049.00 frames. ], tot_loss[loss=0.1674, simple_loss=0.2567, pruned_loss=0.03902, over 2384596.98 frames. ], batch size: 98, lr: 3.62e-03, grad_scale: 8.0 2023-05-16 23:58:15,914 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.8678, 3.4841, 5.2321, 2.9342, 2.9741, 4.0888, 3.3134, 3.8344], device='cuda:0'), covar=tensor([0.0441, 0.1112, 0.0421, 0.1104, 0.1863, 0.1298, 0.1343, 0.1286], device='cuda:0'), in_proj_covar=tensor([0.0237, 0.0237, 0.0253, 0.0185, 0.0238, 0.0292, 0.0227, 0.0264], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-16 23:58:18,568 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=260453.0, num_to_drop=1, layers_to_drop={0} 2023-05-16 23:58:22,420 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=3.00 vs. limit=5.0 2023-05-16 23:58:32,643 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.5224, 2.6056, 3.1959, 4.4112, 2.3491, 4.3529, 4.5309, 4.5767], device='cuda:0'), covar=tensor([0.0151, 0.1231, 0.0504, 0.0165, 0.1305, 0.0247, 0.0132, 0.0087], device='cuda:0'), in_proj_covar=tensor([0.0122, 0.0203, 0.0181, 0.0118, 0.0189, 0.0177, 0.0173, 0.0121], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-16 23:58:38,138 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.850e+02 2.620e+02 3.097e+02 3.656e+02 5.501e+02, threshold=6.194e+02, percent-clipped=0.0 2023-05-16 23:58:42,433 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=260486.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 23:58:46,521 INFO [finetune.py:992] (0/2) Epoch 14, batch 1450, loss[loss=0.1631, simple_loss=0.265, pruned_loss=0.03062, over 12287.00 frames. ], tot_loss[loss=0.1677, simple_loss=0.2574, pruned_loss=0.03899, over 2382298.89 frames. ], batch size: 37, lr: 3.62e-03, grad_scale: 8.0 2023-05-16 23:59:08,134 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=260522.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 23:59:22,355 INFO [finetune.py:992] (0/2) Epoch 14, batch 1500, loss[loss=0.1692, simple_loss=0.2696, pruned_loss=0.03444, over 12146.00 frames. ], tot_loss[loss=0.1667, simple_loss=0.2565, pruned_loss=0.03843, over 2382214.67 frames. ], batch size: 36, lr: 3.62e-03, grad_scale: 8.0 2023-05-16 23:59:42,598 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=260570.0, num_to_drop=0, layers_to_drop=set() 2023-05-16 23:59:49,692 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.748e+02 2.561e+02 3.079e+02 3.709e+02 5.375e+02, threshold=6.157e+02, percent-clipped=0.0 2023-05-16 23:59:49,790 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=260580.0, num_to_drop=1, layers_to_drop={2} 2023-05-16 23:59:59,537 INFO [finetune.py:992] (0/2) Epoch 14, batch 1550, loss[loss=0.1438, simple_loss=0.2394, pruned_loss=0.0241, over 12365.00 frames. ], tot_loss[loss=0.1662, simple_loss=0.2557, pruned_loss=0.03836, over 2388587.64 frames. ], batch size: 36, lr: 3.62e-03, grad_scale: 8.0 2023-05-17 00:00:34,928 INFO [finetune.py:992] (0/2) Epoch 14, batch 1600, loss[loss=0.163, simple_loss=0.2571, pruned_loss=0.03445, over 12095.00 frames. ], tot_loss[loss=0.1663, simple_loss=0.2558, pruned_loss=0.03841, over 2394719.26 frames. ], batch size: 39, lr: 3.62e-03, grad_scale: 8.0 2023-05-17 00:00:37,705 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.84 vs. limit=2.0 2023-05-17 00:00:54,771 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.43 vs. limit=2.0 2023-05-17 00:00:58,603 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.6109, 4.3541, 4.5064, 4.5020, 4.3819, 4.5646, 4.4290, 2.5955], device='cuda:0'), covar=tensor([0.0112, 0.0072, 0.0076, 0.0062, 0.0057, 0.0100, 0.0084, 0.0752], device='cuda:0'), in_proj_covar=tensor([0.0068, 0.0078, 0.0081, 0.0072, 0.0060, 0.0091, 0.0080, 0.0095], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-17 00:01:01,892 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.604e+02 2.613e+02 3.003e+02 3.667e+02 8.089e+02, threshold=6.007e+02, percent-clipped=3.0 2023-05-17 00:01:06,309 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=260686.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 00:01:10,555 INFO [finetune.py:992] (0/2) Epoch 14, batch 1650, loss[loss=0.1792, simple_loss=0.2785, pruned_loss=0.03997, over 12333.00 frames. ], tot_loss[loss=0.1658, simple_loss=0.2557, pruned_loss=0.03802, over 2390122.02 frames. ], batch size: 36, lr: 3.62e-03, grad_scale: 8.0 2023-05-17 00:01:37,934 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=260730.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 00:01:40,726 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=260734.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 00:01:46,408 INFO [finetune.py:992] (0/2) Epoch 14, batch 1700, loss[loss=0.2025, simple_loss=0.2844, pruned_loss=0.06034, over 7948.00 frames. ], tot_loss[loss=0.1647, simple_loss=0.2543, pruned_loss=0.0376, over 2391236.00 frames. ], batch size: 97, lr: 3.62e-03, grad_scale: 8.0 2023-05-17 00:01:54,294 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=260751.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 00:02:03,409 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.7497, 2.3533, 3.2546, 3.6404, 3.4259, 3.6816, 3.3984, 2.6031], device='cuda:0'), covar=tensor([0.0058, 0.0450, 0.0166, 0.0059, 0.0128, 0.0117, 0.0124, 0.0439], device='cuda:0'), in_proj_covar=tensor([0.0090, 0.0126, 0.0107, 0.0080, 0.0106, 0.0118, 0.0099, 0.0142], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-17 00:02:14,603 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.669e+02 2.628e+02 3.189e+02 3.651e+02 5.680e+02, threshold=6.378e+02, percent-clipped=0.0 2023-05-17 00:02:22,634 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=260791.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 00:02:23,150 INFO [finetune.py:992] (0/2) Epoch 14, batch 1750, loss[loss=0.1721, simple_loss=0.2615, pruned_loss=0.04134, over 12282.00 frames. ], tot_loss[loss=0.1655, simple_loss=0.255, pruned_loss=0.03798, over 2387948.99 frames. ], batch size: 37, lr: 3.62e-03, grad_scale: 8.0 2023-05-17 00:02:37,859 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=260812.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 00:02:58,976 INFO [finetune.py:992] (0/2) Epoch 14, batch 1800, loss[loss=0.18, simple_loss=0.267, pruned_loss=0.04654, over 12107.00 frames. ], tot_loss[loss=0.166, simple_loss=0.2555, pruned_loss=0.03826, over 2383719.08 frames. ], batch size: 33, lr: 3.62e-03, grad_scale: 8.0 2023-05-17 00:03:25,987 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.979e+02 2.814e+02 3.306e+02 4.275e+02 7.612e+02, threshold=6.612e+02, percent-clipped=3.0 2023-05-17 00:03:26,116 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=260880.0, num_to_drop=1, layers_to_drop={0} 2023-05-17 00:03:26,773 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=260881.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 00:03:35,660 INFO [finetune.py:992] (0/2) Epoch 14, batch 1850, loss[loss=0.1503, simple_loss=0.2396, pruned_loss=0.03051, over 12088.00 frames. ], tot_loss[loss=0.1657, simple_loss=0.2554, pruned_loss=0.03801, over 2383452.58 frames. ], batch size: 32, lr: 3.62e-03, grad_scale: 8.0 2023-05-17 00:04:01,735 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=260928.0, num_to_drop=1, layers_to_drop={1} 2023-05-17 00:04:11,577 INFO [finetune.py:992] (0/2) Epoch 14, batch 1900, loss[loss=0.1703, simple_loss=0.2731, pruned_loss=0.03374, over 12267.00 frames. ], tot_loss[loss=0.1659, simple_loss=0.2555, pruned_loss=0.03811, over 2381510.45 frames. ], batch size: 37, lr: 3.62e-03, grad_scale: 8.0 2023-05-17 00:04:11,785 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=260942.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 00:04:38,632 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.025e+02 2.640e+02 3.082e+02 3.439e+02 6.112e+02, threshold=6.164e+02, percent-clipped=0.0 2023-05-17 00:04:47,261 INFO [finetune.py:992] (0/2) Epoch 14, batch 1950, loss[loss=0.1673, simple_loss=0.2468, pruned_loss=0.04391, over 12351.00 frames. ], tot_loss[loss=0.1655, simple_loss=0.2549, pruned_loss=0.03802, over 2378824.70 frames. ], batch size: 31, lr: 3.62e-03, grad_scale: 8.0 2023-05-17 00:05:14,969 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.0701, 4.7083, 4.8681, 4.8415, 4.7421, 4.9462, 4.7401, 2.8151], device='cuda:0'), covar=tensor([0.0098, 0.0070, 0.0081, 0.0064, 0.0046, 0.0092, 0.0086, 0.0774], device='cuda:0'), in_proj_covar=tensor([0.0068, 0.0078, 0.0081, 0.0073, 0.0060, 0.0092, 0.0081, 0.0096], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-17 00:05:19,584 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 2023-05-17 00:05:24,895 INFO [finetune.py:992] (0/2) Epoch 14, batch 2000, loss[loss=0.123, simple_loss=0.2049, pruned_loss=0.02055, over 12175.00 frames. ], tot_loss[loss=0.1651, simple_loss=0.2545, pruned_loss=0.03784, over 2379481.57 frames. ], batch size: 29, lr: 3.62e-03, grad_scale: 8.0 2023-05-17 00:05:35,443 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.41 vs. limit=2.0 2023-05-17 00:05:43,323 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.49 vs. limit=2.0 2023-05-17 00:05:52,007 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.884e+02 2.570e+02 2.979e+02 3.561e+02 5.774e+02, threshold=5.957e+02, percent-clipped=0.0 2023-05-17 00:05:56,500 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=261086.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 00:06:00,456 INFO [finetune.py:992] (0/2) Epoch 14, batch 2050, loss[loss=0.161, simple_loss=0.2505, pruned_loss=0.03577, over 12005.00 frames. ], tot_loss[loss=0.1649, simple_loss=0.2542, pruned_loss=0.03777, over 2378922.34 frames. ], batch size: 28, lr: 3.62e-03, grad_scale: 8.0 2023-05-17 00:06:11,436 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=261107.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 00:06:31,832 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=261136.0, num_to_drop=1, layers_to_drop={1} 2023-05-17 00:06:35,852 INFO [finetune.py:992] (0/2) Epoch 14, batch 2100, loss[loss=0.1549, simple_loss=0.2434, pruned_loss=0.03319, over 12098.00 frames. ], tot_loss[loss=0.1659, simple_loss=0.2554, pruned_loss=0.03818, over 2370619.80 frames. ], batch size: 32, lr: 3.62e-03, grad_scale: 8.0 2023-05-17 00:07:04,083 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.170e+02 2.873e+02 3.381e+02 4.160e+02 8.406e+02, threshold=6.762e+02, percent-clipped=4.0 2023-05-17 00:07:12,698 INFO [finetune.py:992] (0/2) Epoch 14, batch 2150, loss[loss=0.1442, simple_loss=0.2249, pruned_loss=0.03174, over 11802.00 frames. ], tot_loss[loss=0.1663, simple_loss=0.2558, pruned_loss=0.03839, over 2355961.55 frames. ], batch size: 26, lr: 3.62e-03, grad_scale: 8.0 2023-05-17 00:07:16,511 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=261197.0, num_to_drop=1, layers_to_drop={1} 2023-05-17 00:07:44,953 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=261237.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 00:07:48,473 INFO [finetune.py:992] (0/2) Epoch 14, batch 2200, loss[loss=0.154, simple_loss=0.2419, pruned_loss=0.03311, over 12108.00 frames. ], tot_loss[loss=0.1653, simple_loss=0.2549, pruned_loss=0.03787, over 2366276.76 frames. ], batch size: 33, lr: 3.61e-03, grad_scale: 8.0 2023-05-17 00:08:07,505 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=2.18 vs. limit=5.0 2023-05-17 00:08:15,595 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.012e+02 2.577e+02 3.065e+02 3.609e+02 7.867e+02, threshold=6.130e+02, percent-clipped=1.0 2023-05-17 00:08:24,235 INFO [finetune.py:992] (0/2) Epoch 14, batch 2250, loss[loss=0.1522, simple_loss=0.247, pruned_loss=0.02867, over 12144.00 frames. ], tot_loss[loss=0.1654, simple_loss=0.2548, pruned_loss=0.03803, over 2371922.18 frames. ], batch size: 34, lr: 3.61e-03, grad_scale: 8.0 2023-05-17 00:09:01,046 INFO [finetune.py:992] (0/2) Epoch 14, batch 2300, loss[loss=0.157, simple_loss=0.2445, pruned_loss=0.03479, over 12355.00 frames. ], tot_loss[loss=0.1646, simple_loss=0.254, pruned_loss=0.0376, over 2378235.86 frames. ], batch size: 31, lr: 3.61e-03, grad_scale: 8.0 2023-05-17 00:09:28,109 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.916e+02 2.585e+02 3.048e+02 3.546e+02 8.125e+02, threshold=6.097e+02, percent-clipped=1.0 2023-05-17 00:09:32,677 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=261386.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 00:09:36,717 INFO [finetune.py:992] (0/2) Epoch 14, batch 2350, loss[loss=0.1501, simple_loss=0.2271, pruned_loss=0.03655, over 11773.00 frames. ], tot_loss[loss=0.1644, simple_loss=0.2536, pruned_loss=0.03758, over 2380106.09 frames. ], batch size: 26, lr: 3.61e-03, grad_scale: 8.0 2023-05-17 00:09:47,832 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=261407.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 00:09:57,001 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.8854, 4.5007, 4.8327, 4.3084, 4.5548, 4.3782, 4.8667, 4.5727], device='cuda:0'), covar=tensor([0.0300, 0.0449, 0.0311, 0.0278, 0.0427, 0.0378, 0.0246, 0.0481], device='cuda:0'), in_proj_covar=tensor([0.0264, 0.0269, 0.0293, 0.0268, 0.0268, 0.0268, 0.0242, 0.0218], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-17 00:10:06,849 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=261434.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 00:10:07,808 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.14 vs. limit=2.0 2023-05-17 00:10:08,452 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 2023-05-17 00:10:12,306 INFO [finetune.py:992] (0/2) Epoch 14, batch 2400, loss[loss=0.1449, simple_loss=0.2424, pruned_loss=0.02367, over 12308.00 frames. ], tot_loss[loss=0.165, simple_loss=0.2545, pruned_loss=0.03778, over 2380553.32 frames. ], batch size: 34, lr: 3.61e-03, grad_scale: 8.0 2023-05-17 00:10:16,655 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.42 vs. limit=2.0 2023-05-17 00:10:21,905 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=261455.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 00:10:22,732 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=261456.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 00:10:33,724 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.8712, 4.5080, 4.6511, 4.7817, 4.6039, 4.8018, 4.6187, 2.6487], device='cuda:0'), covar=tensor([0.0103, 0.0067, 0.0082, 0.0058, 0.0051, 0.0093, 0.0075, 0.0782], device='cuda:0'), in_proj_covar=tensor([0.0069, 0.0079, 0.0082, 0.0073, 0.0061, 0.0093, 0.0082, 0.0096], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-17 00:10:40,726 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.718e+02 2.793e+02 3.197e+02 3.922e+02 8.062e+02, threshold=6.395e+02, percent-clipped=2.0 2023-05-17 00:10:49,440 INFO [finetune.py:992] (0/2) Epoch 14, batch 2450, loss[loss=0.1527, simple_loss=0.2378, pruned_loss=0.03382, over 12066.00 frames. ], tot_loss[loss=0.1655, simple_loss=0.2551, pruned_loss=0.0379, over 2376098.22 frames. ], batch size: 32, lr: 3.61e-03, grad_scale: 8.0 2023-05-17 00:10:49,519 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=261492.0, num_to_drop=1, layers_to_drop={3} 2023-05-17 00:11:07,552 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=261517.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 00:11:12,028 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.11 vs. limit=2.0 2023-05-17 00:11:20,829 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 2023-05-17 00:11:21,861 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=261537.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 00:11:25,236 INFO [finetune.py:992] (0/2) Epoch 14, batch 2500, loss[loss=0.1657, simple_loss=0.2616, pruned_loss=0.03488, over 12021.00 frames. ], tot_loss[loss=0.1655, simple_loss=0.2549, pruned_loss=0.03806, over 2381470.31 frames. ], batch size: 42, lr: 3.61e-03, grad_scale: 16.0 2023-05-17 00:11:37,371 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.9157, 5.9385, 5.6409, 5.2116, 5.0701, 5.7924, 5.4404, 5.1946], device='cuda:0'), covar=tensor([0.0722, 0.0883, 0.0775, 0.1563, 0.0871, 0.0782, 0.1576, 0.1070], device='cuda:0'), in_proj_covar=tensor([0.0625, 0.0570, 0.0523, 0.0648, 0.0427, 0.0739, 0.0796, 0.0577], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0004, 0.0004, 0.0003], device='cuda:0') 2023-05-17 00:11:51,217 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.39 vs. limit=2.0 2023-05-17 00:11:52,002 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.946e+02 2.621e+02 3.073e+02 3.822e+02 6.755e+02, threshold=6.147e+02, percent-clipped=2.0 2023-05-17 00:11:55,600 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=261585.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 00:11:56,563 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.5571, 4.4451, 4.2624, 4.6261, 3.2893, 4.0715, 2.9346, 4.3870], device='cuda:0'), covar=tensor([0.1456, 0.0558, 0.0908, 0.0640, 0.1046, 0.0584, 0.1613, 0.0901], device='cuda:0'), in_proj_covar=tensor([0.0231, 0.0265, 0.0298, 0.0363, 0.0242, 0.0243, 0.0263, 0.0375], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-17 00:12:00,725 INFO [finetune.py:992] (0/2) Epoch 14, batch 2550, loss[loss=0.1633, simple_loss=0.2596, pruned_loss=0.03353, over 12032.00 frames. ], tot_loss[loss=0.166, simple_loss=0.2553, pruned_loss=0.03836, over 2380602.73 frames. ], batch size: 37, lr: 3.61e-03, grad_scale: 16.0 2023-05-17 00:12:02,389 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.6784, 2.6415, 3.3982, 4.6099, 2.6140, 4.6494, 4.7081, 4.6697], device='cuda:0'), covar=tensor([0.0122, 0.1316, 0.0455, 0.0130, 0.1227, 0.0226, 0.0139, 0.0092], device='cuda:0'), in_proj_covar=tensor([0.0122, 0.0203, 0.0181, 0.0119, 0.0189, 0.0178, 0.0174, 0.0122], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-17 00:12:37,881 INFO [finetune.py:992] (0/2) Epoch 14, batch 2600, loss[loss=0.1718, simple_loss=0.2623, pruned_loss=0.04067, over 11658.00 frames. ], tot_loss[loss=0.1655, simple_loss=0.2547, pruned_loss=0.03809, over 2387369.61 frames. ], batch size: 48, lr: 3.61e-03, grad_scale: 16.0 2023-05-17 00:13:02,508 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.0100, 4.6154, 4.8658, 4.8830, 4.7532, 4.8764, 4.7400, 2.7251], device='cuda:0'), covar=tensor([0.0108, 0.0072, 0.0076, 0.0061, 0.0047, 0.0106, 0.0085, 0.0728], device='cuda:0'), in_proj_covar=tensor([0.0069, 0.0079, 0.0082, 0.0074, 0.0061, 0.0093, 0.0082, 0.0097], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-17 00:13:05,191 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.807e+02 2.688e+02 3.200e+02 3.722e+02 1.106e+03, threshold=6.400e+02, percent-clipped=4.0 2023-05-17 00:13:13,657 INFO [finetune.py:992] (0/2) Epoch 14, batch 2650, loss[loss=0.159, simple_loss=0.2623, pruned_loss=0.02787, over 12156.00 frames. ], tot_loss[loss=0.1649, simple_loss=0.2544, pruned_loss=0.03776, over 2383415.66 frames. ], batch size: 36, lr: 3.61e-03, grad_scale: 16.0 2023-05-17 00:13:49,369 INFO [finetune.py:992] (0/2) Epoch 14, batch 2700, loss[loss=0.152, simple_loss=0.2283, pruned_loss=0.03786, over 12308.00 frames. ], tot_loss[loss=0.1648, simple_loss=0.2538, pruned_loss=0.03788, over 2375881.29 frames. ], batch size: 28, lr: 3.61e-03, grad_scale: 16.0 2023-05-17 00:14:17,598 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.851e+02 2.853e+02 3.281e+02 3.725e+02 8.071e+02, threshold=6.562e+02, percent-clipped=2.0 2023-05-17 00:14:26,026 INFO [finetune.py:992] (0/2) Epoch 14, batch 2750, loss[loss=0.1677, simple_loss=0.2644, pruned_loss=0.03552, over 12207.00 frames. ], tot_loss[loss=0.1653, simple_loss=0.2545, pruned_loss=0.03806, over 2380990.95 frames. ], batch size: 35, lr: 3.61e-03, grad_scale: 16.0 2023-05-17 00:14:26,135 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=261792.0, num_to_drop=1, layers_to_drop={1} 2023-05-17 00:14:40,940 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=261812.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 00:14:43,226 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.4242, 2.3754, 3.2275, 4.3277, 2.2437, 4.4585, 4.4627, 4.4460], device='cuda:0'), covar=tensor([0.0129, 0.1415, 0.0491, 0.0166, 0.1333, 0.0215, 0.0137, 0.0092], device='cuda:0'), in_proj_covar=tensor([0.0123, 0.0203, 0.0182, 0.0119, 0.0188, 0.0178, 0.0174, 0.0122], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-17 00:14:47,575 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.2959, 2.7641, 3.8197, 3.1562, 3.5431, 3.3318, 2.7319, 3.6207], device='cuda:0'), covar=tensor([0.0148, 0.0320, 0.0159, 0.0261, 0.0187, 0.0191, 0.0358, 0.0146], device='cuda:0'), in_proj_covar=tensor([0.0183, 0.0208, 0.0195, 0.0193, 0.0223, 0.0171, 0.0205, 0.0191], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-17 00:14:49,351 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.62 vs. limit=2.0 2023-05-17 00:14:49,749 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=261824.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 00:14:56,187 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=261833.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 00:15:01,234 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=261840.0, num_to_drop=1, layers_to_drop={1} 2023-05-17 00:15:02,516 INFO [finetune.py:992] (0/2) Epoch 14, batch 2800, loss[loss=0.172, simple_loss=0.261, pruned_loss=0.04152, over 12195.00 frames. ], tot_loss[loss=0.165, simple_loss=0.2542, pruned_loss=0.03785, over 2380553.19 frames. ], batch size: 35, lr: 3.61e-03, grad_scale: 16.0 2023-05-17 00:15:29,555 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.082e+02 2.566e+02 2.904e+02 3.440e+02 8.546e+02, threshold=5.808e+02, percent-clipped=1.0 2023-05-17 00:15:33,442 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=261885.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 00:15:38,256 INFO [finetune.py:992] (0/2) Epoch 14, batch 2850, loss[loss=0.1692, simple_loss=0.2664, pruned_loss=0.03606, over 12354.00 frames. ], tot_loss[loss=0.1646, simple_loss=0.2538, pruned_loss=0.03772, over 2384609.38 frames. ], batch size: 36, lr: 3.61e-03, grad_scale: 16.0 2023-05-17 00:15:39,865 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=261894.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 00:15:58,451 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.3642, 5.2199, 5.3685, 5.4098, 4.9538, 5.0853, 4.8085, 5.3409], device='cuda:0'), covar=tensor([0.0795, 0.0604, 0.0830, 0.0566, 0.2057, 0.1280, 0.0587, 0.0937], device='cuda:0'), in_proj_covar=tensor([0.0548, 0.0703, 0.0626, 0.0632, 0.0854, 0.0751, 0.0563, 0.0480], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0005, 0.0005, 0.0003, 0.0003], device='cuda:0') 2023-05-17 00:16:15,242 INFO [finetune.py:992] (0/2) Epoch 14, batch 2900, loss[loss=0.1519, simple_loss=0.2382, pruned_loss=0.03283, over 12125.00 frames. ], tot_loss[loss=0.1649, simple_loss=0.2537, pruned_loss=0.03799, over 2379904.28 frames. ], batch size: 30, lr: 3.61e-03, grad_scale: 16.0 2023-05-17 00:16:18,345 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.4893, 2.6314, 3.5903, 4.5457, 3.9415, 4.4867, 3.9598, 3.4243], device='cuda:0'), covar=tensor([0.0054, 0.0431, 0.0162, 0.0049, 0.0128, 0.0087, 0.0119, 0.0334], device='cuda:0'), in_proj_covar=tensor([0.0090, 0.0125, 0.0108, 0.0079, 0.0106, 0.0118, 0.0099, 0.0141], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-17 00:16:34,674 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.77 vs. limit=2.0 2023-05-17 00:16:38,003 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-05-17 00:16:40,594 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=261977.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 00:16:42,506 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.858e+02 2.710e+02 3.195e+02 4.108e+02 8.461e+02, threshold=6.390e+02, percent-clipped=1.0 2023-05-17 00:16:50,978 INFO [finetune.py:992] (0/2) Epoch 14, batch 2950, loss[loss=0.1945, simple_loss=0.2853, pruned_loss=0.05179, over 12044.00 frames. ], tot_loss[loss=0.1654, simple_loss=0.2541, pruned_loss=0.03835, over 2366661.57 frames. ], batch size: 37, lr: 3.61e-03, grad_scale: 16.0 2023-05-17 00:16:55,406 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.3233, 5.2881, 5.2617, 4.6267, 4.7774, 5.4132, 4.6555, 4.8490], device='cuda:0'), covar=tensor([0.1277, 0.1492, 0.1137, 0.2908, 0.1417, 0.1283, 0.3496, 0.1765], device='cuda:0'), in_proj_covar=tensor([0.0621, 0.0560, 0.0517, 0.0643, 0.0419, 0.0729, 0.0790, 0.0570], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-05-17 00:16:57,003 INFO [checkpoint.py:75] (0/2) Saving checkpoint to pruned_transducer_stateless7/exp_giga_finetune/checkpoint-162000.pt 2023-05-17 00:17:27,944 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=262038.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 00:17:30,620 INFO [finetune.py:992] (0/2) Epoch 14, batch 3000, loss[loss=0.1912, simple_loss=0.2823, pruned_loss=0.05002, over 12133.00 frames. ], tot_loss[loss=0.1652, simple_loss=0.254, pruned_loss=0.03821, over 2374481.46 frames. ], batch size: 38, lr: 3.61e-03, grad_scale: 16.0 2023-05-17 00:17:30,621 INFO [finetune.py:1017] (0/2) Computing validation loss 2023-05-17 00:17:47,078 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.7557, 2.0079, 2.9128, 3.7554, 2.0093, 3.8567, 3.6713, 3.8558], device='cuda:0'), covar=tensor([0.0146, 0.1500, 0.0517, 0.0154, 0.1428, 0.0221, 0.0235, 0.0114], device='cuda:0'), in_proj_covar=tensor([0.0122, 0.0201, 0.0180, 0.0119, 0.0188, 0.0178, 0.0173, 0.0121], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-17 00:17:50,068 INFO [finetune.py:1026] (0/2) Epoch 14, validation: loss=0.315, simple_loss=0.3917, pruned_loss=0.1191, over 1020973.00 frames. 2023-05-17 00:17:50,069 INFO [finetune.py:1027] (0/2) Maximum memory allocated so far is 12827MB 2023-05-17 00:18:04,116 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.59 vs. limit=2.0 2023-05-17 00:18:17,816 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.703e+02 2.674e+02 3.142e+02 3.828e+02 1.600e+03, threshold=6.284e+02, percent-clipped=3.0 2023-05-17 00:18:25,565 INFO [finetune.py:992] (0/2) Epoch 14, batch 3050, loss[loss=0.1671, simple_loss=0.2554, pruned_loss=0.03942, over 12301.00 frames. ], tot_loss[loss=0.1657, simple_loss=0.2546, pruned_loss=0.03844, over 2364998.26 frames. ], batch size: 34, lr: 3.61e-03, grad_scale: 8.0 2023-05-17 00:18:34,837 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.0562, 4.7147, 4.8557, 4.9131, 4.7036, 4.8852, 4.8241, 2.8582], device='cuda:0'), covar=tensor([0.0128, 0.0076, 0.0080, 0.0064, 0.0050, 0.0112, 0.0098, 0.0747], device='cuda:0'), in_proj_covar=tensor([0.0070, 0.0080, 0.0083, 0.0074, 0.0062, 0.0094, 0.0083, 0.0097], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-17 00:18:39,634 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=262112.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 00:19:01,600 INFO [finetune.py:992] (0/2) Epoch 14, batch 3100, loss[loss=0.1831, simple_loss=0.2674, pruned_loss=0.04935, over 12025.00 frames. ], tot_loss[loss=0.1646, simple_loss=0.2533, pruned_loss=0.03799, over 2372317.84 frames. ], batch size: 40, lr: 3.61e-03, grad_scale: 8.0 2023-05-17 00:19:15,127 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=262160.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 00:19:29,385 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=262180.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 00:19:30,007 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.858e+02 2.766e+02 3.239e+02 4.149e+02 8.563e+02, threshold=6.478e+02, percent-clipped=3.0 2023-05-17 00:19:35,792 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=262189.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 00:19:37,766 INFO [finetune.py:992] (0/2) Epoch 14, batch 3150, loss[loss=0.1343, simple_loss=0.2293, pruned_loss=0.01966, over 12182.00 frames. ], tot_loss[loss=0.1652, simple_loss=0.2542, pruned_loss=0.0381, over 2373847.29 frames. ], batch size: 31, lr: 3.61e-03, grad_scale: 8.0 2023-05-17 00:20:13,803 INFO [finetune.py:992] (0/2) Epoch 14, batch 3200, loss[loss=0.1486, simple_loss=0.2318, pruned_loss=0.0327, over 12018.00 frames. ], tot_loss[loss=0.165, simple_loss=0.2542, pruned_loss=0.03794, over 2376920.65 frames. ], batch size: 28, lr: 3.61e-03, grad_scale: 8.0 2023-05-17 00:20:42,293 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.815e+02 2.540e+02 3.178e+02 3.757e+02 7.638e+02, threshold=6.357e+02, percent-clipped=1.0 2023-05-17 00:20:50,197 INFO [finetune.py:992] (0/2) Epoch 14, batch 3250, loss[loss=0.1723, simple_loss=0.2615, pruned_loss=0.04152, over 12131.00 frames. ], tot_loss[loss=0.1654, simple_loss=0.2545, pruned_loss=0.03815, over 2384189.24 frames. ], batch size: 33, lr: 3.61e-03, grad_scale: 8.0 2023-05-17 00:21:20,111 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=262333.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 00:21:26,377 INFO [finetune.py:992] (0/2) Epoch 14, batch 3300, loss[loss=0.1666, simple_loss=0.2572, pruned_loss=0.03803, over 12299.00 frames. ], tot_loss[loss=0.1659, simple_loss=0.2548, pruned_loss=0.03846, over 2388641.85 frames. ], batch size: 34, lr: 3.61e-03, grad_scale: 8.0 2023-05-17 00:21:53,886 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.934e+02 2.682e+02 3.147e+02 3.843e+02 7.432e+02, threshold=6.293e+02, percent-clipped=1.0 2023-05-17 00:22:00,256 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.40 vs. limit=2.0 2023-05-17 00:22:01,870 INFO [finetune.py:992] (0/2) Epoch 14, batch 3350, loss[loss=0.1374, simple_loss=0.2217, pruned_loss=0.02656, over 12186.00 frames. ], tot_loss[loss=0.1656, simple_loss=0.2545, pruned_loss=0.03837, over 2383081.66 frames. ], batch size: 29, lr: 3.61e-03, grad_scale: 8.0 2023-05-17 00:22:23,566 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.43 vs. limit=2.0 2023-05-17 00:22:25,583 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.91 vs. limit=2.0 2023-05-17 00:22:39,358 INFO [finetune.py:992] (0/2) Epoch 14, batch 3400, loss[loss=0.1497, simple_loss=0.234, pruned_loss=0.03269, over 12154.00 frames. ], tot_loss[loss=0.1657, simple_loss=0.2546, pruned_loss=0.03839, over 2380062.24 frames. ], batch size: 29, lr: 3.61e-03, grad_scale: 8.0 2023-05-17 00:23:05,132 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.91 vs. limit=2.0 2023-05-17 00:23:06,229 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=262480.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 00:23:06,787 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.107e+02 2.615e+02 3.016e+02 3.667e+02 7.524e+02, threshold=6.033e+02, percent-clipped=3.0 2023-05-17 00:23:12,499 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=262489.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 00:23:14,526 INFO [finetune.py:992] (0/2) Epoch 14, batch 3450, loss[loss=0.1878, simple_loss=0.2651, pruned_loss=0.05523, over 8420.00 frames. ], tot_loss[loss=0.1662, simple_loss=0.2553, pruned_loss=0.03859, over 2364509.86 frames. ], batch size: 99, lr: 3.60e-03, grad_scale: 8.0 2023-05-17 00:23:27,134 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2023-05-17 00:23:29,152 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=2.94 vs. limit=5.0 2023-05-17 00:23:30,407 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=3.28 vs. limit=5.0 2023-05-17 00:23:40,178 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=262528.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 00:23:46,685 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=262537.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 00:23:50,313 INFO [finetune.py:992] (0/2) Epoch 14, batch 3500, loss[loss=0.1532, simple_loss=0.2437, pruned_loss=0.03138, over 12338.00 frames. ], tot_loss[loss=0.1651, simple_loss=0.254, pruned_loss=0.03812, over 2362374.36 frames. ], batch size: 31, lr: 3.60e-03, grad_scale: 8.0 2023-05-17 00:24:19,094 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.767e+02 2.813e+02 3.162e+02 3.676e+02 6.941e+02, threshold=6.323e+02, percent-clipped=3.0 2023-05-17 00:24:25,683 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.8477, 5.8042, 5.6278, 5.1411, 5.0369, 5.7980, 5.3916, 5.1795], device='cuda:0'), covar=tensor([0.0730, 0.1032, 0.0758, 0.1620, 0.0692, 0.0681, 0.1601, 0.1009], device='cuda:0'), in_proj_covar=tensor([0.0615, 0.0561, 0.0513, 0.0638, 0.0418, 0.0728, 0.0787, 0.0569], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-05-17 00:24:27,028 INFO [finetune.py:992] (0/2) Epoch 14, batch 3550, loss[loss=0.2169, simple_loss=0.2826, pruned_loss=0.0756, over 8183.00 frames. ], tot_loss[loss=0.1657, simple_loss=0.2546, pruned_loss=0.03844, over 2361273.63 frames. ], batch size: 98, lr: 3.60e-03, grad_scale: 8.0 2023-05-17 00:24:55,952 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-05-17 00:24:56,407 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=262633.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 00:25:02,390 INFO [finetune.py:992] (0/2) Epoch 14, batch 3600, loss[loss=0.1793, simple_loss=0.2708, pruned_loss=0.04396, over 11606.00 frames. ], tot_loss[loss=0.1662, simple_loss=0.2549, pruned_loss=0.03874, over 2366285.95 frames. ], batch size: 48, lr: 3.60e-03, grad_scale: 8.0 2023-05-17 00:25:02,658 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.9774, 3.5236, 5.3216, 2.8221, 3.0554, 4.0404, 3.4163, 3.9488], device='cuda:0'), covar=tensor([0.0457, 0.1169, 0.0356, 0.1304, 0.2004, 0.1489, 0.1404, 0.1190], device='cuda:0'), in_proj_covar=tensor([0.0241, 0.0240, 0.0259, 0.0188, 0.0243, 0.0299, 0.0231, 0.0270], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-17 00:25:30,091 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.899e+02 2.801e+02 3.265e+02 4.111e+02 6.652e+02, threshold=6.530e+02, percent-clipped=1.0 2023-05-17 00:25:30,175 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=262681.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 00:25:38,086 INFO [finetune.py:992] (0/2) Epoch 14, batch 3650, loss[loss=0.1785, simple_loss=0.2682, pruned_loss=0.04442, over 11148.00 frames. ], tot_loss[loss=0.1654, simple_loss=0.2542, pruned_loss=0.03833, over 2378410.76 frames. ], batch size: 55, lr: 3.60e-03, grad_scale: 8.0 2023-05-17 00:26:14,946 INFO [finetune.py:992] (0/2) Epoch 14, batch 3700, loss[loss=0.169, simple_loss=0.2513, pruned_loss=0.04332, over 12036.00 frames. ], tot_loss[loss=0.1654, simple_loss=0.2543, pruned_loss=0.03819, over 2378786.30 frames. ], batch size: 31, lr: 3.60e-03, grad_scale: 8.0 2023-05-17 00:26:20,101 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=262749.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 00:26:42,898 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.944e+02 2.711e+02 3.026e+02 3.699e+02 9.052e+02, threshold=6.051e+02, percent-clipped=1.0 2023-05-17 00:26:50,650 INFO [finetune.py:992] (0/2) Epoch 14, batch 3750, loss[loss=0.2085, simple_loss=0.298, pruned_loss=0.05944, over 11362.00 frames. ], tot_loss[loss=0.166, simple_loss=0.2552, pruned_loss=0.03841, over 2376352.06 frames. ], batch size: 55, lr: 3.60e-03, grad_scale: 8.0 2023-05-17 00:27:03,612 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=262810.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 00:27:07,058 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.8599, 5.5833, 5.2250, 5.0839, 5.7046, 4.9961, 5.1240, 5.0670], device='cuda:0'), covar=tensor([0.1347, 0.0893, 0.0978, 0.2084, 0.0924, 0.2008, 0.1780, 0.1255], device='cuda:0'), in_proj_covar=tensor([0.0353, 0.0496, 0.0396, 0.0449, 0.0465, 0.0435, 0.0395, 0.0389], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-17 00:27:26,129 INFO [finetune.py:992] (0/2) Epoch 14, batch 3800, loss[loss=0.1582, simple_loss=0.245, pruned_loss=0.03569, over 12191.00 frames. ], tot_loss[loss=0.1664, simple_loss=0.2557, pruned_loss=0.03852, over 2373937.00 frames. ], batch size: 29, lr: 3.60e-03, grad_scale: 8.0 2023-05-17 00:27:55,340 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.699e+02 2.445e+02 2.946e+02 3.863e+02 7.594e+02, threshold=5.892e+02, percent-clipped=4.0 2023-05-17 00:27:55,826 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.38 vs. limit=2.0 2023-05-17 00:27:58,399 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.7982, 5.8420, 5.5678, 5.1057, 5.0614, 5.7622, 5.3945, 5.1306], device='cuda:0'), covar=tensor([0.0718, 0.0811, 0.0599, 0.1654, 0.0731, 0.0645, 0.1395, 0.1037], device='cuda:0'), in_proj_covar=tensor([0.0622, 0.0563, 0.0513, 0.0638, 0.0420, 0.0731, 0.0789, 0.0571], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0004, 0.0004, 0.0003], device='cuda:0') 2023-05-17 00:28:03,231 INFO [finetune.py:992] (0/2) Epoch 14, batch 3850, loss[loss=0.142, simple_loss=0.2237, pruned_loss=0.03016, over 12342.00 frames. ], tot_loss[loss=0.1663, simple_loss=0.2557, pruned_loss=0.03842, over 2376540.70 frames. ], batch size: 31, lr: 3.60e-03, grad_scale: 8.0 2023-05-17 00:28:12,689 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([6.0047, 5.9965, 5.7916, 5.3147, 5.2293, 5.9618, 5.5676, 5.3793], device='cuda:0'), covar=tensor([0.0755, 0.0931, 0.0682, 0.1715, 0.0683, 0.0712, 0.1532, 0.0970], device='cuda:0'), in_proj_covar=tensor([0.0622, 0.0563, 0.0514, 0.0639, 0.0419, 0.0732, 0.0789, 0.0571], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0004, 0.0004, 0.0003], device='cuda:0') 2023-05-17 00:28:38,949 INFO [finetune.py:992] (0/2) Epoch 14, batch 3900, loss[loss=0.229, simple_loss=0.2941, pruned_loss=0.08198, over 8108.00 frames. ], tot_loss[loss=0.1661, simple_loss=0.2553, pruned_loss=0.03843, over 2364879.63 frames. ], batch size: 97, lr: 3.60e-03, grad_scale: 8.0 2023-05-17 00:29:06,455 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.819e+02 2.531e+02 3.065e+02 3.595e+02 1.862e+03, threshold=6.130e+02, percent-clipped=2.0 2023-05-17 00:29:09,355 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.6608, 5.6430, 5.4271, 4.9585, 4.9547, 5.5731, 5.2420, 5.0553], device='cuda:0'), covar=tensor([0.0703, 0.0876, 0.0648, 0.1647, 0.0976, 0.0733, 0.1433, 0.0982], device='cuda:0'), in_proj_covar=tensor([0.0620, 0.0562, 0.0514, 0.0640, 0.0420, 0.0731, 0.0790, 0.0571], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0004, 0.0004, 0.0003], device='cuda:0') 2023-05-17 00:29:14,283 INFO [finetune.py:992] (0/2) Epoch 14, batch 3950, loss[loss=0.1478, simple_loss=0.2274, pruned_loss=0.03406, over 12183.00 frames. ], tot_loss[loss=0.1671, simple_loss=0.2566, pruned_loss=0.03879, over 2364350.58 frames. ], batch size: 29, lr: 3.60e-03, grad_scale: 8.0 2023-05-17 00:29:51,627 INFO [finetune.py:992] (0/2) Epoch 14, batch 4000, loss[loss=0.1706, simple_loss=0.2597, pruned_loss=0.04078, over 12144.00 frames. ], tot_loss[loss=0.1667, simple_loss=0.2564, pruned_loss=0.03852, over 2374577.00 frames. ], batch size: 34, lr: 3.60e-03, grad_scale: 8.0 2023-05-17 00:30:19,050 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.899e+02 2.610e+02 2.958e+02 3.557e+02 5.837e+02, threshold=5.916e+02, percent-clipped=0.0 2023-05-17 00:30:27,017 INFO [finetune.py:992] (0/2) Epoch 14, batch 4050, loss[loss=0.149, simple_loss=0.2444, pruned_loss=0.02679, over 12027.00 frames. ], tot_loss[loss=0.167, simple_loss=0.2562, pruned_loss=0.03891, over 2355994.22 frames. ], batch size: 31, lr: 3.60e-03, grad_scale: 8.0 2023-05-17 00:30:29,331 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=263095.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 00:30:36,211 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=263105.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 00:30:40,703 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=263111.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 00:31:01,186 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=263140.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 00:31:02,453 INFO [finetune.py:992] (0/2) Epoch 14, batch 4100, loss[loss=0.1654, simple_loss=0.2602, pruned_loss=0.03534, over 12058.00 frames. ], tot_loss[loss=0.1671, simple_loss=0.2561, pruned_loss=0.03904, over 2352433.70 frames. ], batch size: 37, lr: 3.60e-03, grad_scale: 8.0 2023-05-17 00:31:13,921 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=263156.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 00:31:14,537 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.3105, 4.6592, 2.9205, 2.7184, 3.9427, 2.4551, 3.9729, 3.3068], device='cuda:0'), covar=tensor([0.0774, 0.0581, 0.1157, 0.1483, 0.0361, 0.1470, 0.0537, 0.0765], device='cuda:0'), in_proj_covar=tensor([0.0191, 0.0259, 0.0177, 0.0202, 0.0143, 0.0183, 0.0200, 0.0176], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-17 00:31:25,134 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=263172.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 00:31:31,296 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.720e+02 2.679e+02 3.227e+02 3.947e+02 7.741e+02, threshold=6.455e+02, percent-clipped=5.0 2023-05-17 00:31:39,179 INFO [finetune.py:992] (0/2) Epoch 14, batch 4150, loss[loss=0.183, simple_loss=0.271, pruned_loss=0.04756, over 12364.00 frames. ], tot_loss[loss=0.1667, simple_loss=0.2556, pruned_loss=0.03887, over 2363962.89 frames. ], batch size: 36, lr: 3.60e-03, grad_scale: 8.0 2023-05-17 00:31:46,247 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=263201.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 00:31:55,281 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=263214.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 00:31:59,417 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.3904, 3.4744, 3.1982, 3.0597, 2.7752, 2.6388, 3.5261, 2.2979], device='cuda:0'), covar=tensor([0.0411, 0.0194, 0.0227, 0.0230, 0.0420, 0.0422, 0.0156, 0.0504], device='cuda:0'), in_proj_covar=tensor([0.0193, 0.0163, 0.0167, 0.0188, 0.0202, 0.0199, 0.0172, 0.0203], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-17 00:32:15,093 INFO [finetune.py:992] (0/2) Epoch 14, batch 4200, loss[loss=0.1523, simple_loss=0.238, pruned_loss=0.03332, over 12402.00 frames. ], tot_loss[loss=0.1661, simple_loss=0.2553, pruned_loss=0.03845, over 2370752.05 frames. ], batch size: 32, lr: 3.60e-03, grad_scale: 8.0 2023-05-17 00:32:18,125 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=263246.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 00:32:38,969 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=263275.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 00:32:42,993 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.563e+02 2.543e+02 3.024e+02 3.575e+02 6.631e+02, threshold=6.048e+02, percent-clipped=1.0 2023-05-17 00:32:51,482 INFO [finetune.py:992] (0/2) Epoch 14, batch 4250, loss[loss=0.1645, simple_loss=0.2585, pruned_loss=0.03527, over 11208.00 frames. ], tot_loss[loss=0.1653, simple_loss=0.2544, pruned_loss=0.03806, over 2373003.00 frames. ], batch size: 55, lr: 3.60e-03, grad_scale: 8.0 2023-05-17 00:32:59,494 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=4.51 vs. limit=5.0 2023-05-17 00:33:02,775 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=263307.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 00:33:27,725 INFO [finetune.py:992] (0/2) Epoch 14, batch 4300, loss[loss=0.2255, simple_loss=0.3003, pruned_loss=0.07534, over 8293.00 frames. ], tot_loss[loss=0.1653, simple_loss=0.2545, pruned_loss=0.03807, over 2367951.12 frames. ], batch size: 98, lr: 3.60e-03, grad_scale: 8.0 2023-05-17 00:33:55,194 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.758e+02 2.623e+02 2.984e+02 3.575e+02 6.098e+02, threshold=5.968e+02, percent-clipped=1.0 2023-05-17 00:33:56,144 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.4810, 3.4658, 3.1642, 3.0322, 2.7688, 2.6422, 3.4938, 2.3268], device='cuda:0'), covar=tensor([0.0358, 0.0140, 0.0196, 0.0198, 0.0409, 0.0336, 0.0144, 0.0449], device='cuda:0'), in_proj_covar=tensor([0.0193, 0.0163, 0.0166, 0.0188, 0.0202, 0.0200, 0.0172, 0.0204], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-17 00:34:02,857 INFO [finetune.py:992] (0/2) Epoch 14, batch 4350, loss[loss=0.1738, simple_loss=0.2608, pruned_loss=0.04341, over 12294.00 frames. ], tot_loss[loss=0.1658, simple_loss=0.2549, pruned_loss=0.0383, over 2367062.90 frames. ], batch size: 28, lr: 3.60e-03, grad_scale: 8.0 2023-05-17 00:34:09,814 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.4255, 2.4186, 3.8181, 3.2779, 3.6157, 3.3300, 2.6614, 3.6368], device='cuda:0'), covar=tensor([0.0123, 0.0435, 0.0177, 0.0245, 0.0164, 0.0181, 0.0390, 0.0158], device='cuda:0'), in_proj_covar=tensor([0.0180, 0.0205, 0.0193, 0.0189, 0.0219, 0.0167, 0.0201, 0.0188], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-17 00:34:12,444 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=263405.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 00:34:39,791 INFO [finetune.py:992] (0/2) Epoch 14, batch 4400, loss[loss=0.1355, simple_loss=0.2216, pruned_loss=0.02472, over 12015.00 frames. ], tot_loss[loss=0.1654, simple_loss=0.2546, pruned_loss=0.03814, over 2366491.07 frames. ], batch size: 28, lr: 3.60e-03, grad_scale: 8.0 2023-05-17 00:34:45,830 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=263451.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 00:34:47,346 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=263453.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 00:34:57,325 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=263467.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 00:35:07,230 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.773e+02 2.616e+02 3.252e+02 4.018e+02 9.612e+02, threshold=6.504e+02, percent-clipped=5.0 2023-05-17 00:35:15,016 INFO [finetune.py:992] (0/2) Epoch 14, batch 4450, loss[loss=0.1471, simple_loss=0.2324, pruned_loss=0.03088, over 12180.00 frames. ], tot_loss[loss=0.1652, simple_loss=0.2547, pruned_loss=0.03792, over 2373406.53 frames. ], batch size: 31, lr: 3.60e-03, grad_scale: 8.0 2023-05-17 00:35:17,829 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=263496.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 00:35:50,401 INFO [finetune.py:992] (0/2) Epoch 14, batch 4500, loss[loss=0.1529, simple_loss=0.2399, pruned_loss=0.03294, over 12185.00 frames. ], tot_loss[loss=0.1654, simple_loss=0.255, pruned_loss=0.03789, over 2374714.69 frames. ], batch size: 31, lr: 3.60e-03, grad_scale: 8.0 2023-05-17 00:35:58,064 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.34 vs. limit=2.0 2023-05-17 00:36:06,414 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([6.0024, 5.9653, 5.7186, 5.3178, 5.1966, 5.8753, 5.5286, 5.2286], device='cuda:0'), covar=tensor([0.0697, 0.0902, 0.0654, 0.1678, 0.0696, 0.0839, 0.1570, 0.1034], device='cuda:0'), in_proj_covar=tensor([0.0621, 0.0567, 0.0517, 0.0646, 0.0425, 0.0736, 0.0796, 0.0577], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0004, 0.0004, 0.0003], device='cuda:0') 2023-05-17 00:36:10,618 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=263570.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 00:36:18,907 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2023-05-17 00:36:19,218 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.823e+02 2.705e+02 3.261e+02 3.794e+02 6.330e+02, threshold=6.522e+02, percent-clipped=0.0 2023-05-17 00:36:27,803 INFO [finetune.py:992] (0/2) Epoch 14, batch 4550, loss[loss=0.1814, simple_loss=0.2804, pruned_loss=0.0412, over 12099.00 frames. ], tot_loss[loss=0.1656, simple_loss=0.2551, pruned_loss=0.03799, over 2378187.97 frames. ], batch size: 33, lr: 3.60e-03, grad_scale: 8.0 2023-05-17 00:36:35,128 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=263602.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 00:36:53,466 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.3573, 4.6678, 4.1748, 4.9532, 4.5432, 2.8823, 4.3222, 2.9524], device='cuda:0'), covar=tensor([0.0747, 0.0765, 0.1317, 0.0525, 0.1014, 0.1685, 0.0981, 0.3372], device='cuda:0'), in_proj_covar=tensor([0.0307, 0.0382, 0.0361, 0.0321, 0.0370, 0.0274, 0.0348, 0.0370], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-17 00:37:03,756 INFO [finetune.py:992] (0/2) Epoch 14, batch 4600, loss[loss=0.166, simple_loss=0.2555, pruned_loss=0.03831, over 12293.00 frames. ], tot_loss[loss=0.1656, simple_loss=0.2552, pruned_loss=0.03802, over 2382551.11 frames. ], batch size: 33, lr: 3.60e-03, grad_scale: 8.0 2023-05-17 00:37:31,550 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.900e+02 2.517e+02 3.041e+02 3.636e+02 5.915e+02, threshold=6.082e+02, percent-clipped=0.0 2023-05-17 00:37:39,350 INFO [finetune.py:992] (0/2) Epoch 14, batch 4650, loss[loss=0.1431, simple_loss=0.2361, pruned_loss=0.02504, over 12319.00 frames. ], tot_loss[loss=0.166, simple_loss=0.2559, pruned_loss=0.03806, over 2373915.98 frames. ], batch size: 30, lr: 3.60e-03, grad_scale: 8.0 2023-05-17 00:38:04,154 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.8836, 5.6794, 5.1749, 5.2554, 5.8161, 5.1761, 5.2911, 5.2359], device='cuda:0'), covar=tensor([0.1802, 0.1047, 0.1239, 0.2284, 0.0986, 0.2157, 0.2127, 0.1445], device='cuda:0'), in_proj_covar=tensor([0.0366, 0.0509, 0.0408, 0.0460, 0.0478, 0.0444, 0.0404, 0.0398], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-17 00:38:13,435 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.8127, 2.9621, 4.7589, 4.9263, 3.0650, 2.7532, 3.0786, 2.3474], device='cuda:0'), covar=tensor([0.1607, 0.2973, 0.0465, 0.0382, 0.1312, 0.2325, 0.2858, 0.3982], device='cuda:0'), in_proj_covar=tensor([0.0297, 0.0386, 0.0275, 0.0300, 0.0272, 0.0308, 0.0387, 0.0373], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-17 00:38:16,072 INFO [finetune.py:992] (0/2) Epoch 14, batch 4700, loss[loss=0.1238, simple_loss=0.2114, pruned_loss=0.01809, over 12269.00 frames. ], tot_loss[loss=0.1661, simple_loss=0.2559, pruned_loss=0.03815, over 2373368.41 frames. ], batch size: 28, lr: 3.59e-03, grad_scale: 8.0 2023-05-17 00:38:22,827 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=263751.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 00:38:34,145 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=263767.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 00:38:43,920 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.610e+02 2.488e+02 3.045e+02 3.623e+02 8.469e+02, threshold=6.090e+02, percent-clipped=2.0 2023-05-17 00:38:51,486 INFO [finetune.py:992] (0/2) Epoch 14, batch 4750, loss[loss=0.169, simple_loss=0.2574, pruned_loss=0.04031, over 12050.00 frames. ], tot_loss[loss=0.1661, simple_loss=0.2557, pruned_loss=0.03823, over 2379033.40 frames. ], batch size: 40, lr: 3.59e-03, grad_scale: 8.0 2023-05-17 00:38:54,275 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=263796.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 00:38:56,371 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=263799.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 00:39:07,960 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=263815.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 00:39:20,936 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=263833.0, num_to_drop=1, layers_to_drop={0} 2023-05-17 00:39:27,094 INFO [finetune.py:992] (0/2) Epoch 14, batch 4800, loss[loss=0.1608, simple_loss=0.2553, pruned_loss=0.03313, over 12272.00 frames. ], tot_loss[loss=0.1658, simple_loss=0.2556, pruned_loss=0.03803, over 2388771.32 frames. ], batch size: 37, lr: 3.59e-03, grad_scale: 8.0 2023-05-17 00:39:28,502 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=263844.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 00:39:47,085 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=263870.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 00:39:47,179 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.7850, 3.7492, 3.2186, 3.2311, 2.9394, 2.8829, 3.8162, 2.5575], device='cuda:0'), covar=tensor([0.0355, 0.0137, 0.0259, 0.0232, 0.0435, 0.0392, 0.0138, 0.0496], device='cuda:0'), in_proj_covar=tensor([0.0194, 0.0164, 0.0169, 0.0189, 0.0204, 0.0202, 0.0172, 0.0206], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-17 00:39:54,400 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=3.10 vs. limit=5.0 2023-05-17 00:39:55,298 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.824e+02 2.609e+02 3.113e+02 3.835e+02 5.503e+02, threshold=6.225e+02, percent-clipped=0.0 2023-05-17 00:40:03,767 INFO [finetune.py:992] (0/2) Epoch 14, batch 4850, loss[loss=0.1684, simple_loss=0.2578, pruned_loss=0.03954, over 12007.00 frames. ], tot_loss[loss=0.1652, simple_loss=0.2547, pruned_loss=0.03783, over 2388386.86 frames. ], batch size: 40, lr: 3.59e-03, grad_scale: 8.0 2023-05-17 00:40:05,287 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=263894.0, num_to_drop=1, layers_to_drop={0} 2023-05-17 00:40:10,827 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=263902.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 00:40:17,432 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.2440, 2.5690, 3.5494, 4.1939, 3.7513, 4.1837, 3.8017, 3.0080], device='cuda:0'), covar=tensor([0.0045, 0.0376, 0.0155, 0.0049, 0.0106, 0.0087, 0.0098, 0.0371], device='cuda:0'), in_proj_covar=tensor([0.0090, 0.0125, 0.0106, 0.0079, 0.0104, 0.0117, 0.0097, 0.0138], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-17 00:40:22,300 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=263918.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 00:40:39,480 INFO [finetune.py:992] (0/2) Epoch 14, batch 4900, loss[loss=0.184, simple_loss=0.2712, pruned_loss=0.04843, over 10660.00 frames. ], tot_loss[loss=0.1661, simple_loss=0.2554, pruned_loss=0.03835, over 2375247.19 frames. ], batch size: 70, lr: 3.59e-03, grad_scale: 8.0 2023-05-17 00:40:45,305 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=263950.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 00:40:52,653 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=263960.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 00:41:07,268 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.685e+02 2.685e+02 3.434e+02 4.193e+02 1.143e+03, threshold=6.868e+02, percent-clipped=7.0 2023-05-17 00:41:15,124 INFO [finetune.py:992] (0/2) Epoch 14, batch 4950, loss[loss=0.1293, simple_loss=0.2129, pruned_loss=0.02287, over 12335.00 frames. ], tot_loss[loss=0.1656, simple_loss=0.2544, pruned_loss=0.03834, over 2374578.51 frames. ], batch size: 30, lr: 3.59e-03, grad_scale: 8.0 2023-05-17 00:41:21,171 INFO [checkpoint.py:75] (0/2) Saving checkpoint to pruned_transducer_stateless7/exp_giga_finetune/checkpoint-164000.pt 2023-05-17 00:41:39,655 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.9358, 5.9350, 5.6352, 5.1870, 5.1160, 5.7947, 5.4641, 5.2369], device='cuda:0'), covar=tensor([0.0636, 0.0752, 0.0689, 0.1645, 0.0777, 0.0724, 0.1391, 0.0941], device='cuda:0'), in_proj_covar=tensor([0.0625, 0.0571, 0.0519, 0.0651, 0.0430, 0.0739, 0.0799, 0.0579], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0004, 0.0004, 0.0003], device='cuda:0') 2023-05-17 00:41:39,730 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=264021.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 00:41:55,095 INFO [finetune.py:992] (0/2) Epoch 14, batch 5000, loss[loss=0.1445, simple_loss=0.2294, pruned_loss=0.02984, over 12332.00 frames. ], tot_loss[loss=0.166, simple_loss=0.2547, pruned_loss=0.03869, over 2365994.34 frames. ], batch size: 31, lr: 3.59e-03, grad_scale: 8.0 2023-05-17 00:42:14,756 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.7209, 2.9744, 4.5435, 4.6666, 2.8073, 2.6536, 3.0140, 2.2370], device='cuda:0'), covar=tensor([0.1588, 0.3021, 0.0485, 0.0410, 0.1385, 0.2358, 0.2718, 0.3878], device='cuda:0'), in_proj_covar=tensor([0.0297, 0.0387, 0.0275, 0.0300, 0.0273, 0.0309, 0.0387, 0.0373], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-17 00:42:23,095 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.781e+02 2.696e+02 3.369e+02 3.930e+02 9.340e+02, threshold=6.738e+02, percent-clipped=2.0 2023-05-17 00:42:30,644 INFO [finetune.py:992] (0/2) Epoch 14, batch 5050, loss[loss=0.1718, simple_loss=0.261, pruned_loss=0.04125, over 11787.00 frames. ], tot_loss[loss=0.1662, simple_loss=0.2547, pruned_loss=0.03885, over 2373710.24 frames. ], batch size: 44, lr: 3.59e-03, grad_scale: 16.0 2023-05-17 00:42:41,126 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=4.71 vs. limit=5.0 2023-05-17 00:43:06,019 INFO [finetune.py:992] (0/2) Epoch 14, batch 5100, loss[loss=0.1793, simple_loss=0.2685, pruned_loss=0.04505, over 10330.00 frames. ], tot_loss[loss=0.1653, simple_loss=0.2541, pruned_loss=0.03828, over 2373685.80 frames. ], batch size: 68, lr: 3.59e-03, grad_scale: 16.0 2023-05-17 00:43:34,830 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.823e+02 2.640e+02 3.176e+02 3.824e+02 9.093e+02, threshold=6.353e+02, percent-clipped=2.0 2023-05-17 00:43:40,573 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=264189.0, num_to_drop=1, layers_to_drop={2} 2023-05-17 00:43:42,573 INFO [finetune.py:992] (0/2) Epoch 14, batch 5150, loss[loss=0.1665, simple_loss=0.2529, pruned_loss=0.04002, over 12082.00 frames. ], tot_loss[loss=0.1656, simple_loss=0.2544, pruned_loss=0.0384, over 2362183.77 frames. ], batch size: 32, lr: 3.59e-03, grad_scale: 16.0 2023-05-17 00:43:53,663 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.9293, 3.5061, 5.2400, 2.6427, 2.9551, 3.7668, 3.3996, 3.7640], device='cuda:0'), covar=tensor([0.0358, 0.1046, 0.0300, 0.1232, 0.1924, 0.1609, 0.1301, 0.1322], device='cuda:0'), in_proj_covar=tensor([0.0240, 0.0238, 0.0255, 0.0185, 0.0239, 0.0297, 0.0227, 0.0268], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-17 00:44:00,125 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.0460, 3.8313, 4.0142, 3.6737, 3.8385, 3.6794, 4.0227, 3.5848], device='cuda:0'), covar=tensor([0.0417, 0.0467, 0.0421, 0.0295, 0.0419, 0.0398, 0.0361, 0.1715], device='cuda:0'), in_proj_covar=tensor([0.0273, 0.0276, 0.0300, 0.0274, 0.0275, 0.0276, 0.0250, 0.0223], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-17 00:44:12,795 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=264234.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 00:44:18,389 INFO [finetune.py:992] (0/2) Epoch 14, batch 5200, loss[loss=0.1759, simple_loss=0.2657, pruned_loss=0.04311, over 10609.00 frames. ], tot_loss[loss=0.1656, simple_loss=0.2545, pruned_loss=0.03833, over 2358896.05 frames. ], batch size: 68, lr: 3.59e-03, grad_scale: 16.0 2023-05-17 00:44:42,123 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.6895, 2.9472, 4.6004, 4.8196, 2.8037, 2.6846, 2.9085, 2.2864], device='cuda:0'), covar=tensor([0.1736, 0.3063, 0.0497, 0.0434, 0.1402, 0.2393, 0.3025, 0.3974], device='cuda:0'), in_proj_covar=tensor([0.0298, 0.0388, 0.0276, 0.0301, 0.0273, 0.0309, 0.0390, 0.0375], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-17 00:44:44,909 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.5641, 3.5508, 3.2435, 3.0937, 2.8795, 2.6759, 3.6608, 2.3243], device='cuda:0'), covar=tensor([0.0385, 0.0195, 0.0211, 0.0224, 0.0442, 0.0394, 0.0121, 0.0498], device='cuda:0'), in_proj_covar=tensor([0.0194, 0.0164, 0.0167, 0.0190, 0.0203, 0.0201, 0.0172, 0.0205], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-17 00:44:46,095 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.808e+02 2.687e+02 3.109e+02 3.729e+02 7.458e+02, threshold=6.219e+02, percent-clipped=1.0 2023-05-17 00:44:53,720 INFO [finetune.py:992] (0/2) Epoch 14, batch 5250, loss[loss=0.184, simple_loss=0.2741, pruned_loss=0.04697, over 11588.00 frames. ], tot_loss[loss=0.1666, simple_loss=0.2555, pruned_loss=0.03888, over 2364579.53 frames. ], batch size: 48, lr: 3.59e-03, grad_scale: 16.0 2023-05-17 00:44:56,009 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=264295.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 00:45:12,357 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=264316.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 00:45:30,395 INFO [finetune.py:992] (0/2) Epoch 14, batch 5300, loss[loss=0.1433, simple_loss=0.2266, pruned_loss=0.03002, over 12275.00 frames. ], tot_loss[loss=0.1668, simple_loss=0.256, pruned_loss=0.03885, over 2367309.20 frames. ], batch size: 28, lr: 3.59e-03, grad_scale: 16.0 2023-05-17 00:45:32,651 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.6594, 2.8399, 3.7442, 4.5919, 3.9981, 4.7052, 3.9932, 3.3246], device='cuda:0'), covar=tensor([0.0034, 0.0351, 0.0136, 0.0040, 0.0109, 0.0066, 0.0101, 0.0333], device='cuda:0'), in_proj_covar=tensor([0.0090, 0.0126, 0.0107, 0.0079, 0.0105, 0.0118, 0.0098, 0.0139], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-17 00:45:34,289 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.62 vs. limit=2.0 2023-05-17 00:45:58,266 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.919e+02 2.985e+02 3.539e+02 4.100e+02 7.031e+02, threshold=7.077e+02, percent-clipped=2.0 2023-05-17 00:45:59,499 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.58 vs. limit=2.0 2023-05-17 00:46:06,212 INFO [finetune.py:992] (0/2) Epoch 14, batch 5350, loss[loss=0.1936, simple_loss=0.2907, pruned_loss=0.04827, over 12280.00 frames. ], tot_loss[loss=0.1663, simple_loss=0.2556, pruned_loss=0.03856, over 2370668.03 frames. ], batch size: 37, lr: 3.59e-03, grad_scale: 16.0 2023-05-17 00:46:12,555 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.68 vs. limit=2.0 2023-05-17 00:46:14,532 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.1754, 4.5834, 2.8763, 2.5954, 3.9050, 2.4822, 3.9644, 3.1924], device='cuda:0'), covar=tensor([0.0766, 0.0494, 0.1172, 0.1635, 0.0347, 0.1391, 0.0527, 0.0817], device='cuda:0'), in_proj_covar=tensor([0.0194, 0.0262, 0.0181, 0.0205, 0.0146, 0.0185, 0.0204, 0.0178], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-17 00:46:38,409 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.6457, 2.5699, 3.3112, 4.5526, 2.5450, 4.5156, 4.5990, 4.6375], device='cuda:0'), covar=tensor([0.0138, 0.1336, 0.0499, 0.0145, 0.1393, 0.0218, 0.0140, 0.0094], device='cuda:0'), in_proj_covar=tensor([0.0125, 0.0205, 0.0184, 0.0122, 0.0193, 0.0182, 0.0177, 0.0125], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-17 00:46:42,441 INFO [finetune.py:992] (0/2) Epoch 14, batch 5400, loss[loss=0.175, simple_loss=0.2666, pruned_loss=0.0417, over 12362.00 frames. ], tot_loss[loss=0.1666, simple_loss=0.2558, pruned_loss=0.03872, over 2368460.73 frames. ], batch size: 38, lr: 3.59e-03, grad_scale: 16.0 2023-05-17 00:46:42,838 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=3.01 vs. limit=5.0 2023-05-17 00:47:09,391 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=264479.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 00:47:10,569 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.765e+02 2.548e+02 2.961e+02 3.629e+02 7.539e+02, threshold=5.921e+02, percent-clipped=1.0 2023-05-17 00:47:15,705 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.4502, 5.2170, 5.3989, 5.4150, 5.0362, 5.0166, 4.8218, 5.3282], device='cuda:0'), covar=tensor([0.0701, 0.0637, 0.0824, 0.0611, 0.1939, 0.1566, 0.0581, 0.1048], device='cuda:0'), in_proj_covar=tensor([0.0549, 0.0709, 0.0629, 0.0647, 0.0872, 0.0764, 0.0573, 0.0491], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0005, 0.0004, 0.0004, 0.0005, 0.0005, 0.0004, 0.0003], device='cuda:0') 2023-05-17 00:47:16,447 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=264489.0, num_to_drop=1, layers_to_drop={2} 2023-05-17 00:47:17,773 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.1295, 4.9157, 5.1087, 5.1092, 4.7442, 4.7460, 4.5448, 5.0348], device='cuda:0'), covar=tensor([0.0783, 0.0675, 0.0892, 0.0644, 0.2072, 0.1608, 0.0618, 0.1013], device='cuda:0'), in_proj_covar=tensor([0.0550, 0.0709, 0.0629, 0.0647, 0.0873, 0.0765, 0.0573, 0.0491], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0005, 0.0004, 0.0004, 0.0005, 0.0005, 0.0004, 0.0003], device='cuda:0') 2023-05-17 00:47:18,389 INFO [finetune.py:992] (0/2) Epoch 14, batch 5450, loss[loss=0.167, simple_loss=0.2591, pruned_loss=0.03749, over 11878.00 frames. ], tot_loss[loss=0.1665, simple_loss=0.2558, pruned_loss=0.03856, over 2367682.51 frames. ], batch size: 44, lr: 3.59e-03, grad_scale: 16.0 2023-05-17 00:47:24,193 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=264500.0, num_to_drop=1, layers_to_drop={0} 2023-05-17 00:47:34,780 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.0201, 3.6112, 5.3218, 2.6827, 3.0219, 3.8856, 3.3814, 3.9283], device='cuda:0'), covar=tensor([0.0387, 0.1039, 0.0344, 0.1226, 0.1922, 0.1586, 0.1341, 0.1230], device='cuda:0'), in_proj_covar=tensor([0.0240, 0.0238, 0.0255, 0.0185, 0.0239, 0.0298, 0.0227, 0.0268], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-17 00:47:50,027 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=2.99 vs. limit=5.0 2023-05-17 00:47:50,390 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=264537.0, num_to_drop=1, layers_to_drop={1} 2023-05-17 00:47:52,634 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=264540.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 00:47:53,785 INFO [finetune.py:992] (0/2) Epoch 14, batch 5500, loss[loss=0.1753, simple_loss=0.2776, pruned_loss=0.03653, over 12349.00 frames. ], tot_loss[loss=0.1655, simple_loss=0.2547, pruned_loss=0.03814, over 2372343.78 frames. ], batch size: 36, lr: 3.59e-03, grad_scale: 16.0 2023-05-17 00:48:03,272 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.5868, 3.6493, 3.2707, 3.3022, 2.9565, 2.8242, 3.6582, 2.3935], device='cuda:0'), covar=tensor([0.0378, 0.0132, 0.0216, 0.0182, 0.0379, 0.0363, 0.0136, 0.0503], device='cuda:0'), in_proj_covar=tensor([0.0195, 0.0165, 0.0167, 0.0189, 0.0205, 0.0203, 0.0172, 0.0206], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-17 00:48:07,557 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=264561.0, num_to_drop=1, layers_to_drop={0} 2023-05-17 00:48:15,674 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.68 vs. limit=2.0 2023-05-17 00:48:21,551 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.865e+02 2.628e+02 3.099e+02 3.637e+02 7.816e+02, threshold=6.199e+02, percent-clipped=1.0 2023-05-17 00:48:28,124 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=264590.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 00:48:30,153 INFO [finetune.py:992] (0/2) Epoch 14, batch 5550, loss[loss=0.1638, simple_loss=0.2559, pruned_loss=0.03587, over 12070.00 frames. ], tot_loss[loss=0.1651, simple_loss=0.2544, pruned_loss=0.03788, over 2370683.83 frames. ], batch size: 32, lr: 3.59e-03, grad_scale: 16.0 2023-05-17 00:48:48,566 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=264616.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 00:49:00,733 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.4291, 4.8704, 4.1844, 5.0512, 4.6619, 3.1791, 4.3494, 3.1219], device='cuda:0'), covar=tensor([0.0914, 0.0819, 0.1529, 0.0506, 0.1125, 0.1645, 0.1076, 0.3491], device='cuda:0'), in_proj_covar=tensor([0.0309, 0.0384, 0.0364, 0.0323, 0.0373, 0.0275, 0.0353, 0.0370], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-17 00:49:06,840 INFO [finetune.py:992] (0/2) Epoch 14, batch 5600, loss[loss=0.1582, simple_loss=0.2497, pruned_loss=0.03333, over 12287.00 frames. ], tot_loss[loss=0.166, simple_loss=0.2551, pruned_loss=0.03843, over 2366349.21 frames. ], batch size: 33, lr: 3.59e-03, grad_scale: 16.0 2023-05-17 00:49:22,504 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=264664.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 00:49:22,617 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.9946, 4.8948, 4.8614, 4.8919, 4.5003, 5.0163, 4.9784, 5.2533], device='cuda:0'), covar=tensor([0.0355, 0.0156, 0.0221, 0.0338, 0.0804, 0.0373, 0.0169, 0.0168], device='cuda:0'), in_proj_covar=tensor([0.0196, 0.0201, 0.0193, 0.0250, 0.0243, 0.0221, 0.0178, 0.0227], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0005, 0.0004, 0.0004, 0.0003, 0.0005], device='cuda:0') 2023-05-17 00:49:29,169 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.4138, 2.8679, 3.8645, 3.3074, 3.6615, 3.4449, 2.9887, 3.7544], device='cuda:0'), covar=tensor([0.0109, 0.0302, 0.0131, 0.0202, 0.0144, 0.0152, 0.0262, 0.0116], device='cuda:0'), in_proj_covar=tensor([0.0183, 0.0208, 0.0197, 0.0193, 0.0223, 0.0171, 0.0204, 0.0193], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-17 00:49:34,635 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.642e+02 2.679e+02 3.341e+02 3.846e+02 7.061e+02, threshold=6.682e+02, percent-clipped=3.0 2023-05-17 00:49:42,210 INFO [finetune.py:992] (0/2) Epoch 14, batch 5650, loss[loss=0.1716, simple_loss=0.2625, pruned_loss=0.0404, over 12009.00 frames. ], tot_loss[loss=0.1654, simple_loss=0.2547, pruned_loss=0.03806, over 2380098.09 frames. ], batch size: 40, lr: 3.59e-03, grad_scale: 16.0 2023-05-17 00:50:04,489 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.3327, 2.5323, 3.8239, 3.1685, 3.6462, 3.3655, 2.6269, 3.6562], device='cuda:0'), covar=tensor([0.0142, 0.0445, 0.0152, 0.0281, 0.0147, 0.0168, 0.0427, 0.0141], device='cuda:0'), in_proj_covar=tensor([0.0183, 0.0209, 0.0197, 0.0193, 0.0223, 0.0171, 0.0204, 0.0193], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-17 00:50:18,492 INFO [finetune.py:992] (0/2) Epoch 14, batch 5700, loss[loss=0.1292, simple_loss=0.2158, pruned_loss=0.02131, over 12003.00 frames. ], tot_loss[loss=0.1661, simple_loss=0.2556, pruned_loss=0.03827, over 2379967.57 frames. ], batch size: 28, lr: 3.59e-03, grad_scale: 16.0 2023-05-17 00:50:32,021 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.6278, 2.5779, 3.9657, 4.0591, 2.7424, 2.5084, 2.7020, 2.1793], device='cuda:0'), covar=tensor([0.1643, 0.2896, 0.0590, 0.0499, 0.1385, 0.2492, 0.2864, 0.3996], device='cuda:0'), in_proj_covar=tensor([0.0301, 0.0393, 0.0279, 0.0303, 0.0276, 0.0313, 0.0394, 0.0380], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-17 00:50:34,083 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.5218, 2.4405, 3.7108, 4.3917, 3.9352, 4.6011, 3.8505, 3.1510], device='cuda:0'), covar=tensor([0.0040, 0.0458, 0.0150, 0.0051, 0.0116, 0.0066, 0.0138, 0.0372], device='cuda:0'), in_proj_covar=tensor([0.0090, 0.0125, 0.0106, 0.0079, 0.0104, 0.0116, 0.0098, 0.0138], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-17 00:50:46,731 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.066e+02 2.696e+02 3.141e+02 3.814e+02 8.222e+02, threshold=6.283e+02, percent-clipped=3.0 2023-05-17 00:50:54,495 INFO [finetune.py:992] (0/2) Epoch 14, batch 5750, loss[loss=0.1743, simple_loss=0.2641, pruned_loss=0.04231, over 12296.00 frames. ], tot_loss[loss=0.1665, simple_loss=0.2559, pruned_loss=0.03852, over 2380518.25 frames. ], batch size: 34, lr: 3.59e-03, grad_scale: 16.0 2023-05-17 00:51:03,395 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=264804.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 00:51:25,690 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=264835.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 00:51:27,350 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=2.66 vs. limit=5.0 2023-05-17 00:51:30,559 INFO [finetune.py:992] (0/2) Epoch 14, batch 5800, loss[loss=0.1682, simple_loss=0.2579, pruned_loss=0.03921, over 12055.00 frames. ], tot_loss[loss=0.1666, simple_loss=0.256, pruned_loss=0.03862, over 2377863.01 frames. ], batch size: 37, lr: 3.59e-03, grad_scale: 16.0 2023-05-17 00:51:36,511 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.88 vs. limit=2.0 2023-05-17 00:51:40,284 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=264856.0, num_to_drop=1, layers_to_drop={0} 2023-05-17 00:51:46,812 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=264865.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 00:51:57,903 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.831e+02 2.752e+02 3.123e+02 3.647e+02 6.224e+02, threshold=6.246e+02, percent-clipped=0.0 2023-05-17 00:52:04,949 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=264890.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 00:52:06,247 INFO [finetune.py:992] (0/2) Epoch 14, batch 5850, loss[loss=0.1571, simple_loss=0.2492, pruned_loss=0.03249, over 12134.00 frames. ], tot_loss[loss=0.1667, simple_loss=0.2558, pruned_loss=0.0388, over 2377237.16 frames. ], batch size: 30, lr: 3.59e-03, grad_scale: 16.0 2023-05-17 00:52:24,055 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.3685, 4.9714, 5.3675, 4.7104, 4.9737, 4.8373, 5.3948, 5.0185], device='cuda:0'), covar=tensor([0.0293, 0.0384, 0.0279, 0.0284, 0.0459, 0.0362, 0.0224, 0.0351], device='cuda:0'), in_proj_covar=tensor([0.0272, 0.0274, 0.0299, 0.0274, 0.0274, 0.0275, 0.0249, 0.0223], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-17 00:52:39,351 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=264938.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 00:52:42,206 INFO [finetune.py:992] (0/2) Epoch 14, batch 5900, loss[loss=0.1602, simple_loss=0.2536, pruned_loss=0.03343, over 12340.00 frames. ], tot_loss[loss=0.1662, simple_loss=0.2554, pruned_loss=0.03845, over 2379720.57 frames. ], batch size: 36, lr: 3.59e-03, grad_scale: 16.0 2023-05-17 00:53:10,020 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.805e+02 2.718e+02 3.176e+02 3.638e+02 7.396e+02, threshold=6.352e+02, percent-clipped=1.0 2023-05-17 00:53:17,774 INFO [finetune.py:992] (0/2) Epoch 14, batch 5950, loss[loss=0.161, simple_loss=0.2487, pruned_loss=0.03666, over 12181.00 frames. ], tot_loss[loss=0.1663, simple_loss=0.2557, pruned_loss=0.03847, over 2376692.34 frames. ], batch size: 31, lr: 3.58e-03, grad_scale: 16.0 2023-05-17 00:53:24,760 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.7215, 2.2377, 3.3544, 3.6554, 3.4628, 3.7610, 3.3808, 2.6529], device='cuda:0'), covar=tensor([0.0062, 0.0440, 0.0159, 0.0065, 0.0120, 0.0093, 0.0122, 0.0387], device='cuda:0'), in_proj_covar=tensor([0.0090, 0.0124, 0.0105, 0.0079, 0.0104, 0.0116, 0.0098, 0.0138], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-17 00:53:54,407 INFO [finetune.py:992] (0/2) Epoch 14, batch 6000, loss[loss=0.1673, simple_loss=0.2532, pruned_loss=0.04072, over 12289.00 frames. ], tot_loss[loss=0.1663, simple_loss=0.2555, pruned_loss=0.03858, over 2373069.14 frames. ], batch size: 37, lr: 3.58e-03, grad_scale: 16.0 2023-05-17 00:53:54,408 INFO [finetune.py:1017] (0/2) Computing validation loss 2023-05-17 00:54:05,130 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.4432, 3.0578, 5.1880, 2.5926, 2.7213, 4.0421, 3.0455, 4.0163], device='cuda:0'), covar=tensor([0.0520, 0.1412, 0.0137, 0.1242, 0.2044, 0.1233, 0.1609, 0.0852], device='cuda:0'), in_proj_covar=tensor([0.0238, 0.0237, 0.0252, 0.0183, 0.0237, 0.0295, 0.0225, 0.0265], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-17 00:54:06,425 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.6400, 1.9662, 2.7563, 3.6000, 2.0906, 3.7313, 3.2157, 3.6514], device='cuda:0'), covar=tensor([0.0166, 0.1408, 0.0543, 0.0169, 0.1471, 0.0185, 0.0392, 0.0147], device='cuda:0'), in_proj_covar=tensor([0.0125, 0.0206, 0.0185, 0.0123, 0.0195, 0.0182, 0.0179, 0.0126], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-17 00:54:12,932 INFO [finetune.py:1026] (0/2) Epoch 14, validation: loss=0.3136, simple_loss=0.3909, pruned_loss=0.1181, over 1020973.00 frames. 2023-05-17 00:54:12,933 INFO [finetune.py:1027] (0/2) Maximum memory allocated so far is 12827MB 2023-05-17 00:54:40,684 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.995e+02 2.610e+02 3.085e+02 3.720e+02 7.946e+02, threshold=6.170e+02, percent-clipped=2.0 2023-05-17 00:54:48,373 INFO [finetune.py:992] (0/2) Epoch 14, batch 6050, loss[loss=0.165, simple_loss=0.2394, pruned_loss=0.04533, over 11765.00 frames. ], tot_loss[loss=0.1671, simple_loss=0.2558, pruned_loss=0.03919, over 2359565.84 frames. ], batch size: 26, lr: 3.58e-03, grad_scale: 16.0 2023-05-17 00:54:58,443 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.3620, 5.1901, 5.2676, 5.3325, 4.9886, 5.0006, 4.7950, 5.2579], device='cuda:0'), covar=tensor([0.0660, 0.0543, 0.0881, 0.0609, 0.1811, 0.1322, 0.0560, 0.1036], device='cuda:0'), in_proj_covar=tensor([0.0547, 0.0705, 0.0629, 0.0650, 0.0867, 0.0761, 0.0574, 0.0492], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0005, 0.0005, 0.0004, 0.0003], device='cuda:0') 2023-05-17 00:55:19,488 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=265135.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 00:55:24,139 INFO [finetune.py:992] (0/2) Epoch 14, batch 6100, loss[loss=0.1428, simple_loss=0.2318, pruned_loss=0.0269, over 12145.00 frames. ], tot_loss[loss=0.1676, simple_loss=0.2567, pruned_loss=0.03922, over 2360606.53 frames. ], batch size: 29, lr: 3.58e-03, grad_scale: 16.0 2023-05-17 00:55:27,376 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.12 vs. limit=2.0 2023-05-17 00:55:34,379 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=265156.0, num_to_drop=1, layers_to_drop={2} 2023-05-17 00:55:37,244 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=265160.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 00:55:52,810 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.824e+02 2.534e+02 2.916e+02 4.072e+02 6.898e+02, threshold=5.833e+02, percent-clipped=2.0 2023-05-17 00:55:54,312 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=265183.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 00:56:00,675 INFO [finetune.py:992] (0/2) Epoch 14, batch 6150, loss[loss=0.1681, simple_loss=0.262, pruned_loss=0.03715, over 11550.00 frames. ], tot_loss[loss=0.1673, simple_loss=0.2563, pruned_loss=0.03917, over 2361207.54 frames. ], batch size: 48, lr: 3.58e-03, grad_scale: 16.0 2023-05-17 00:56:09,307 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=265204.0, num_to_drop=1, layers_to_drop={1} 2023-05-17 00:56:10,107 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.4556, 2.6149, 3.6687, 4.3828, 3.9398, 4.5243, 3.7902, 3.0610], device='cuda:0'), covar=tensor([0.0037, 0.0385, 0.0156, 0.0053, 0.0109, 0.0066, 0.0147, 0.0357], device='cuda:0'), in_proj_covar=tensor([0.0089, 0.0124, 0.0105, 0.0079, 0.0104, 0.0116, 0.0098, 0.0137], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-17 00:56:18,589 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([6.0825, 6.0324, 5.8318, 5.3563, 5.1441, 5.9857, 5.5426, 5.3656], device='cuda:0'), covar=tensor([0.0663, 0.0991, 0.0652, 0.1793, 0.0717, 0.0677, 0.1685, 0.0944], device='cuda:0'), in_proj_covar=tensor([0.0628, 0.0577, 0.0525, 0.0654, 0.0430, 0.0745, 0.0808, 0.0584], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0004, 0.0004, 0.0003], device='cuda:0') 2023-05-17 00:56:29,050 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.36 vs. limit=2.0 2023-05-17 00:56:36,203 INFO [finetune.py:992] (0/2) Epoch 14, batch 6200, loss[loss=0.1585, simple_loss=0.2557, pruned_loss=0.03061, over 12189.00 frames. ], tot_loss[loss=0.167, simple_loss=0.256, pruned_loss=0.03903, over 2369114.73 frames. ], batch size: 35, lr: 3.58e-03, grad_scale: 16.0 2023-05-17 00:57:01,815 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=265278.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 00:57:04,370 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.721e+02 2.684e+02 3.070e+02 3.874e+02 6.926e+02, threshold=6.140e+02, percent-clipped=2.0 2023-05-17 00:57:12,073 INFO [finetune.py:992] (0/2) Epoch 14, batch 6250, loss[loss=0.1391, simple_loss=0.2217, pruned_loss=0.02824, over 12149.00 frames. ], tot_loss[loss=0.1665, simple_loss=0.2553, pruned_loss=0.03885, over 2371819.20 frames. ], batch size: 30, lr: 3.58e-03, grad_scale: 16.0 2023-05-17 00:57:36,912 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.4030, 2.5874, 3.5278, 4.2258, 3.7308, 4.3964, 3.6322, 2.9151], device='cuda:0'), covar=tensor([0.0037, 0.0389, 0.0175, 0.0063, 0.0132, 0.0075, 0.0152, 0.0398], device='cuda:0'), in_proj_covar=tensor([0.0089, 0.0124, 0.0105, 0.0079, 0.0104, 0.0116, 0.0098, 0.0137], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-17 00:57:46,060 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=265339.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 00:57:47,480 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.5417, 2.6882, 3.7051, 4.4290, 3.8597, 4.5391, 3.8128, 3.0351], device='cuda:0'), covar=tensor([0.0041, 0.0408, 0.0139, 0.0055, 0.0132, 0.0088, 0.0139, 0.0403], device='cuda:0'), in_proj_covar=tensor([0.0089, 0.0124, 0.0105, 0.0079, 0.0104, 0.0116, 0.0098, 0.0137], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-17 00:57:48,005 INFO [finetune.py:992] (0/2) Epoch 14, batch 6300, loss[loss=0.172, simple_loss=0.2617, pruned_loss=0.04114, over 12308.00 frames. ], tot_loss[loss=0.1665, simple_loss=0.2551, pruned_loss=0.03896, over 2371774.08 frames. ], batch size: 34, lr: 3.58e-03, grad_scale: 16.0 2023-05-17 00:58:13,915 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.6201, 3.2565, 5.0182, 2.4278, 2.7082, 3.6685, 3.1061, 3.7016], device='cuda:0'), covar=tensor([0.0442, 0.1220, 0.0297, 0.1340, 0.2020, 0.1652, 0.1427, 0.1165], device='cuda:0'), in_proj_covar=tensor([0.0239, 0.0237, 0.0253, 0.0184, 0.0237, 0.0296, 0.0225, 0.0265], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-17 00:58:15,805 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.961e+02 2.676e+02 3.101e+02 3.558e+02 6.054e+02, threshold=6.203e+02, percent-clipped=0.0 2023-05-17 00:58:23,344 INFO [finetune.py:992] (0/2) Epoch 14, batch 6350, loss[loss=0.1568, simple_loss=0.2483, pruned_loss=0.03268, over 12352.00 frames. ], tot_loss[loss=0.1667, simple_loss=0.2556, pruned_loss=0.03892, over 2377986.58 frames. ], batch size: 30, lr: 3.58e-03, grad_scale: 16.0 2023-05-17 00:58:59,880 INFO [finetune.py:992] (0/2) Epoch 14, batch 6400, loss[loss=0.1629, simple_loss=0.2588, pruned_loss=0.03348, over 12357.00 frames. ], tot_loss[loss=0.1666, simple_loss=0.2557, pruned_loss=0.03875, over 2371465.20 frames. ], batch size: 36, lr: 3.58e-03, grad_scale: 16.0 2023-05-17 00:59:05,143 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.6691, 2.9420, 4.2518, 4.4846, 2.7421, 2.5950, 2.8760, 2.1385], device='cuda:0'), covar=tensor([0.1692, 0.2997, 0.0576, 0.0489, 0.1433, 0.2548, 0.2961, 0.4220], device='cuda:0'), in_proj_covar=tensor([0.0300, 0.0393, 0.0278, 0.0303, 0.0275, 0.0312, 0.0392, 0.0377], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-17 00:59:12,761 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=265460.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 00:59:28,014 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.064e+02 2.672e+02 3.215e+02 3.915e+02 6.084e+02, threshold=6.430e+02, percent-clipped=0.0 2023-05-17 00:59:34,585 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.4421, 2.2806, 3.0971, 4.2722, 2.0911, 4.4055, 4.3601, 4.4531], device='cuda:0'), covar=tensor([0.0150, 0.1271, 0.0524, 0.0186, 0.1423, 0.0267, 0.0190, 0.0110], device='cuda:0'), in_proj_covar=tensor([0.0123, 0.0203, 0.0183, 0.0120, 0.0191, 0.0179, 0.0177, 0.0124], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-17 00:59:35,805 INFO [finetune.py:992] (0/2) Epoch 14, batch 6450, loss[loss=0.1297, simple_loss=0.2159, pruned_loss=0.02178, over 11993.00 frames. ], tot_loss[loss=0.1668, simple_loss=0.2562, pruned_loss=0.0387, over 2371292.39 frames. ], batch size: 28, lr: 3.58e-03, grad_scale: 16.0 2023-05-17 00:59:47,373 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=265508.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 01:00:01,667 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.6460, 2.3851, 3.2063, 4.5092, 2.4408, 4.5883, 4.5339, 4.6537], device='cuda:0'), covar=tensor([0.0117, 0.1360, 0.0545, 0.0133, 0.1362, 0.0188, 0.0165, 0.0091], device='cuda:0'), in_proj_covar=tensor([0.0123, 0.0203, 0.0183, 0.0120, 0.0192, 0.0179, 0.0177, 0.0124], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-17 01:00:02,983 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=265530.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 01:00:10,924 INFO [finetune.py:992] (0/2) Epoch 14, batch 6500, loss[loss=0.156, simple_loss=0.2406, pruned_loss=0.03572, over 12180.00 frames. ], tot_loss[loss=0.1675, simple_loss=0.2567, pruned_loss=0.03911, over 2373045.29 frames. ], batch size: 31, lr: 3.58e-03, grad_scale: 16.0 2023-05-17 01:00:36,493 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.0487, 5.9648, 5.5484, 5.4811, 6.0226, 5.3236, 5.3081, 5.4997], device='cuda:0'), covar=tensor([0.1670, 0.0909, 0.1065, 0.2093, 0.1028, 0.2307, 0.2270, 0.1265], device='cuda:0'), in_proj_covar=tensor([0.0365, 0.0511, 0.0408, 0.0459, 0.0478, 0.0445, 0.0408, 0.0400], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-17 01:00:39,182 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.899e+02 2.785e+02 3.332e+02 3.898e+02 5.510e+02, threshold=6.664e+02, percent-clipped=0.0 2023-05-17 01:00:46,683 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=265591.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 01:00:47,228 INFO [finetune.py:992] (0/2) Epoch 14, batch 6550, loss[loss=0.1423, simple_loss=0.2294, pruned_loss=0.02759, over 12163.00 frames. ], tot_loss[loss=0.1668, simple_loss=0.2559, pruned_loss=0.03881, over 2376627.34 frames. ], batch size: 29, lr: 3.58e-03, grad_scale: 16.0 2023-05-17 01:01:04,101 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=265614.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 01:01:12,182 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.5273, 4.8755, 3.0099, 2.5206, 4.2686, 2.3335, 4.0926, 3.1562], device='cuda:0'), covar=tensor([0.0632, 0.0532, 0.1219, 0.1950, 0.0310, 0.1823, 0.0518, 0.1029], device='cuda:0'), in_proj_covar=tensor([0.0193, 0.0264, 0.0181, 0.0206, 0.0145, 0.0185, 0.0203, 0.0179], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-17 01:01:18,586 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=265634.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 01:01:24,298 INFO [finetune.py:992] (0/2) Epoch 14, batch 6600, loss[loss=0.167, simple_loss=0.2527, pruned_loss=0.0407, over 12351.00 frames. ], tot_loss[loss=0.1664, simple_loss=0.256, pruned_loss=0.03845, over 2379214.44 frames. ], batch size: 31, lr: 3.58e-03, grad_scale: 16.0 2023-05-17 01:01:47,357 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.8809, 5.8437, 5.6500, 5.1147, 5.0714, 5.7820, 5.4068, 5.2275], device='cuda:0'), covar=tensor([0.0705, 0.0884, 0.0647, 0.1632, 0.0840, 0.0724, 0.1470, 0.0917], device='cuda:0'), in_proj_covar=tensor([0.0626, 0.0571, 0.0524, 0.0649, 0.0424, 0.0743, 0.0803, 0.0583], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0004, 0.0004, 0.0003], device='cuda:0') 2023-05-17 01:01:48,150 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=265675.0, num_to_drop=1, layers_to_drop={3} 2023-05-17 01:01:52,165 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.730e+02 2.631e+02 2.986e+02 3.533e+02 9.316e+02, threshold=5.973e+02, percent-clipped=1.0 2023-05-17 01:02:00,136 INFO [finetune.py:992] (0/2) Epoch 14, batch 6650, loss[loss=0.1646, simple_loss=0.26, pruned_loss=0.03458, over 11640.00 frames. ], tot_loss[loss=0.1671, simple_loss=0.2569, pruned_loss=0.03865, over 2373349.92 frames. ], batch size: 48, lr: 3.58e-03, grad_scale: 16.0 2023-05-17 01:02:16,991 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=265715.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 01:02:36,492 INFO [finetune.py:992] (0/2) Epoch 14, batch 6700, loss[loss=0.1649, simple_loss=0.2501, pruned_loss=0.03982, over 12339.00 frames. ], tot_loss[loss=0.1668, simple_loss=0.2568, pruned_loss=0.03845, over 2369467.98 frames. ], batch size: 31, lr: 3.58e-03, grad_scale: 16.0 2023-05-17 01:03:01,713 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=265776.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 01:03:04,967 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.051e+02 2.711e+02 3.255e+02 3.641e+02 5.853e+02, threshold=6.511e+02, percent-clipped=0.0 2023-05-17 01:03:12,931 INFO [finetune.py:992] (0/2) Epoch 14, batch 6750, loss[loss=0.1459, simple_loss=0.2367, pruned_loss=0.02754, over 12043.00 frames. ], tot_loss[loss=0.1658, simple_loss=0.2557, pruned_loss=0.0379, over 2372095.50 frames. ], batch size: 31, lr: 3.58e-03, grad_scale: 16.0 2023-05-17 01:03:48,590 INFO [finetune.py:992] (0/2) Epoch 14, batch 6800, loss[loss=0.1522, simple_loss=0.2391, pruned_loss=0.03264, over 11986.00 frames. ], tot_loss[loss=0.1667, simple_loss=0.2563, pruned_loss=0.03854, over 2363940.24 frames. ], batch size: 28, lr: 3.58e-03, grad_scale: 16.0 2023-05-17 01:03:54,529 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.3405, 3.9055, 3.9896, 4.3252, 2.8668, 3.7522, 2.4656, 3.9024], device='cuda:0'), covar=tensor([0.1532, 0.0726, 0.0848, 0.0579, 0.1186, 0.0618, 0.1880, 0.1304], device='cuda:0'), in_proj_covar=tensor([0.0230, 0.0266, 0.0297, 0.0360, 0.0242, 0.0242, 0.0260, 0.0376], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-17 01:04:06,615 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.3032, 4.8475, 5.2619, 4.5700, 4.9026, 4.7179, 5.3021, 4.9201], device='cuda:0'), covar=tensor([0.0276, 0.0430, 0.0291, 0.0306, 0.0434, 0.0352, 0.0220, 0.0365], device='cuda:0'), in_proj_covar=tensor([0.0274, 0.0277, 0.0302, 0.0276, 0.0275, 0.0276, 0.0250, 0.0225], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-17 01:04:10,706 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=265872.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 01:04:17,068 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.788e+02 2.693e+02 3.222e+02 3.796e+02 7.708e+02, threshold=6.444e+02, percent-clipped=1.0 2023-05-17 01:04:17,231 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.6193, 4.4501, 4.5695, 4.6259, 4.2927, 4.3659, 4.1765, 4.5448], device='cuda:0'), covar=tensor([0.0793, 0.0733, 0.0979, 0.0615, 0.2099, 0.1280, 0.0669, 0.0991], device='cuda:0'), in_proj_covar=tensor([0.0552, 0.0713, 0.0629, 0.0654, 0.0880, 0.0764, 0.0583, 0.0493], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0005, 0.0004, 0.0004, 0.0005, 0.0005, 0.0004, 0.0003], device='cuda:0') 2023-05-17 01:04:20,606 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=265886.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 01:04:21,004 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.42 vs. limit=2.0 2023-05-17 01:04:25,561 INFO [finetune.py:992] (0/2) Epoch 14, batch 6850, loss[loss=0.2176, simple_loss=0.3084, pruned_loss=0.06338, over 7904.00 frames. ], tot_loss[loss=0.1677, simple_loss=0.2571, pruned_loss=0.03913, over 2358078.31 frames. ], batch size: 97, lr: 3.58e-03, grad_scale: 16.0 2023-05-17 01:04:43,580 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=265917.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 01:04:55,047 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=265933.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 01:04:55,718 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=265934.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 01:05:01,299 INFO [finetune.py:992] (0/2) Epoch 14, batch 6900, loss[loss=0.1597, simple_loss=0.2513, pruned_loss=0.03403, over 12189.00 frames. ], tot_loss[loss=0.1674, simple_loss=0.2568, pruned_loss=0.03904, over 2360284.62 frames. ], batch size: 35, lr: 3.58e-03, grad_scale: 16.0 2023-05-17 01:05:17,873 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 2023-05-17 01:05:21,151 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=265970.0, num_to_drop=1, layers_to_drop={1} 2023-05-17 01:05:26,978 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=265978.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 01:05:28,996 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.838e+02 2.809e+02 3.282e+02 3.911e+02 1.414e+03, threshold=6.565e+02, percent-clipped=6.0 2023-05-17 01:05:29,807 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=265982.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 01:05:36,842 INFO [finetune.py:992] (0/2) Epoch 14, batch 6950, loss[loss=0.146, simple_loss=0.2258, pruned_loss=0.03308, over 12008.00 frames. ], tot_loss[loss=0.1672, simple_loss=0.2563, pruned_loss=0.039, over 2355470.18 frames. ], batch size: 28, lr: 3.58e-03, grad_scale: 16.0 2023-05-17 01:05:42,785 INFO [checkpoint.py:75] (0/2) Saving checkpoint to pruned_transducer_stateless7/exp_giga_finetune/checkpoint-166000.pt 2023-05-17 01:05:49,148 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.8342, 2.5801, 3.6060, 3.6296, 2.8855, 2.6542, 2.7126, 2.3978], device='cuda:0'), covar=tensor([0.1308, 0.2624, 0.0659, 0.0537, 0.1116, 0.2280, 0.2509, 0.3508], device='cuda:0'), in_proj_covar=tensor([0.0297, 0.0388, 0.0276, 0.0299, 0.0272, 0.0309, 0.0388, 0.0373], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-17 01:06:16,368 INFO [finetune.py:992] (0/2) Epoch 14, batch 7000, loss[loss=0.1668, simple_loss=0.2553, pruned_loss=0.03912, over 12350.00 frames. ], tot_loss[loss=0.1663, simple_loss=0.2557, pruned_loss=0.03849, over 2364117.13 frames. ], batch size: 36, lr: 3.58e-03, grad_scale: 16.0 2023-05-17 01:06:37,172 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=266071.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 01:06:37,986 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.5666, 5.1433, 5.5005, 4.8537, 5.1213, 4.9350, 5.5691, 5.1959], device='cuda:0'), covar=tensor([0.0235, 0.0350, 0.0264, 0.0264, 0.0358, 0.0405, 0.0218, 0.0239], device='cuda:0'), in_proj_covar=tensor([0.0276, 0.0278, 0.0304, 0.0278, 0.0276, 0.0278, 0.0251, 0.0226], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-17 01:06:44,282 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.849e+02 2.623e+02 3.097e+02 3.755e+02 1.095e+03, threshold=6.194e+02, percent-clipped=4.0 2023-05-17 01:06:52,094 INFO [finetune.py:992] (0/2) Epoch 14, batch 7050, loss[loss=0.1487, simple_loss=0.2331, pruned_loss=0.03211, over 12399.00 frames. ], tot_loss[loss=0.1669, simple_loss=0.2563, pruned_loss=0.03873, over 2362039.10 frames. ], batch size: 32, lr: 3.58e-03, grad_scale: 32.0 2023-05-17 01:06:58,673 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.3277, 4.9197, 5.2561, 4.6432, 4.8983, 4.7524, 5.3076, 5.0122], device='cuda:0'), covar=tensor([0.0284, 0.0407, 0.0330, 0.0269, 0.0425, 0.0358, 0.0248, 0.0378], device='cuda:0'), in_proj_covar=tensor([0.0274, 0.0276, 0.0303, 0.0276, 0.0275, 0.0276, 0.0250, 0.0225], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-17 01:07:23,890 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.6490, 3.3651, 5.2149, 2.4602, 2.8051, 3.8524, 3.1753, 3.9487], device='cuda:0'), covar=tensor([0.0541, 0.1191, 0.0261, 0.1252, 0.2037, 0.1428, 0.1403, 0.1009], device='cuda:0'), in_proj_covar=tensor([0.0239, 0.0237, 0.0253, 0.0184, 0.0238, 0.0297, 0.0226, 0.0265], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-17 01:07:27,339 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.2796, 4.4372, 2.7898, 2.4442, 3.8352, 2.6467, 3.8337, 3.1560], device='cuda:0'), covar=tensor([0.0655, 0.0614, 0.1191, 0.1641, 0.0318, 0.1236, 0.0544, 0.0752], device='cuda:0'), in_proj_covar=tensor([0.0192, 0.0260, 0.0178, 0.0204, 0.0143, 0.0184, 0.0201, 0.0177], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-17 01:07:27,803 INFO [finetune.py:992] (0/2) Epoch 14, batch 7100, loss[loss=0.1675, simple_loss=0.261, pruned_loss=0.03698, over 11774.00 frames. ], tot_loss[loss=0.1671, simple_loss=0.2568, pruned_loss=0.03872, over 2360413.00 frames. ], batch size: 44, lr: 3.58e-03, grad_scale: 32.0 2023-05-17 01:07:56,839 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.947e+02 2.647e+02 3.091e+02 3.573e+02 6.537e+02, threshold=6.182e+02, percent-clipped=2.0 2023-05-17 01:08:00,410 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=266186.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 01:08:04,718 INFO [finetune.py:992] (0/2) Epoch 14, batch 7150, loss[loss=0.1498, simple_loss=0.2322, pruned_loss=0.03374, over 12352.00 frames. ], tot_loss[loss=0.1667, simple_loss=0.2561, pruned_loss=0.03863, over 2369020.13 frames. ], batch size: 30, lr: 3.58e-03, grad_scale: 32.0 2023-05-17 01:08:14,907 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.5101, 2.6410, 3.7952, 4.3924, 3.8168, 4.4414, 3.7996, 2.8210], device='cuda:0'), covar=tensor([0.0035, 0.0363, 0.0137, 0.0047, 0.0118, 0.0080, 0.0130, 0.0395], device='cuda:0'), in_proj_covar=tensor([0.0090, 0.0124, 0.0105, 0.0080, 0.0104, 0.0117, 0.0098, 0.0138], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-17 01:08:21,323 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=266214.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 01:08:29,732 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=266226.0, num_to_drop=1, layers_to_drop={0} 2023-05-17 01:08:31,093 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=266228.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 01:08:35,281 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=266234.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 01:08:40,992 INFO [finetune.py:992] (0/2) Epoch 14, batch 7200, loss[loss=0.1729, simple_loss=0.2704, pruned_loss=0.03769, over 12346.00 frames. ], tot_loss[loss=0.166, simple_loss=0.2553, pruned_loss=0.03839, over 2375084.76 frames. ], batch size: 36, lr: 3.57e-03, grad_scale: 32.0 2023-05-17 01:08:49,617 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([6.2416, 6.2154, 6.0547, 5.5233, 5.3745, 6.1819, 5.7814, 5.5518], device='cuda:0'), covar=tensor([0.0613, 0.0935, 0.0595, 0.1539, 0.0592, 0.0633, 0.1492, 0.0916], device='cuda:0'), in_proj_covar=tensor([0.0622, 0.0564, 0.0518, 0.0644, 0.0422, 0.0734, 0.0793, 0.0576], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0004, 0.0004, 0.0003], device='cuda:0') 2023-05-17 01:09:01,348 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=266270.0, num_to_drop=1, layers_to_drop={0} 2023-05-17 01:09:03,381 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=266273.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 01:09:04,963 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=266275.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 01:09:09,044 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.785e+02 2.507e+02 2.959e+02 3.672e+02 9.427e+02, threshold=5.917e+02, percent-clipped=2.0 2023-05-17 01:09:13,455 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=266287.0, num_to_drop=1, layers_to_drop={0} 2023-05-17 01:09:16,674 INFO [finetune.py:992] (0/2) Epoch 14, batch 7250, loss[loss=0.1385, simple_loss=0.2251, pruned_loss=0.02601, over 12170.00 frames. ], tot_loss[loss=0.1664, simple_loss=0.2555, pruned_loss=0.03867, over 2371384.96 frames. ], batch size: 29, lr: 3.57e-03, grad_scale: 32.0 2023-05-17 01:09:20,979 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.3014, 2.2873, 3.8275, 4.1985, 3.7709, 4.2345, 3.8248, 2.8440], device='cuda:0'), covar=tensor([0.0056, 0.0542, 0.0129, 0.0077, 0.0151, 0.0101, 0.0118, 0.0433], device='cuda:0'), in_proj_covar=tensor([0.0090, 0.0124, 0.0105, 0.0080, 0.0104, 0.0117, 0.0098, 0.0138], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-17 01:09:35,902 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=266318.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 01:09:53,355 INFO [finetune.py:992] (0/2) Epoch 14, batch 7300, loss[loss=0.237, simple_loss=0.3124, pruned_loss=0.0808, over 7722.00 frames. ], tot_loss[loss=0.1666, simple_loss=0.2558, pruned_loss=0.03867, over 2370959.70 frames. ], batch size: 98, lr: 3.57e-03, grad_scale: 32.0 2023-05-17 01:09:56,349 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=266346.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 01:10:13,839 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=266371.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 01:10:20,024 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.39 vs. limit=2.0 2023-05-17 01:10:20,976 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.913e+02 2.881e+02 3.416e+02 4.291e+02 1.034e+03, threshold=6.831e+02, percent-clipped=8.0 2023-05-17 01:10:28,856 INFO [finetune.py:992] (0/2) Epoch 14, batch 7350, loss[loss=0.1484, simple_loss=0.2406, pruned_loss=0.02814, over 12352.00 frames. ], tot_loss[loss=0.1672, simple_loss=0.2565, pruned_loss=0.03896, over 2370908.85 frames. ], batch size: 35, lr: 3.57e-03, grad_scale: 32.0 2023-05-17 01:10:40,052 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=266407.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 01:10:48,453 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=266419.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 01:11:05,619 INFO [finetune.py:992] (0/2) Epoch 14, batch 7400, loss[loss=0.1591, simple_loss=0.2592, pruned_loss=0.02945, over 12350.00 frames. ], tot_loss[loss=0.1667, simple_loss=0.256, pruned_loss=0.03868, over 2368064.05 frames. ], batch size: 36, lr: 3.57e-03, grad_scale: 32.0 2023-05-17 01:11:12,964 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.5248, 4.8512, 2.9920, 2.7736, 4.2687, 2.7041, 4.0687, 3.3415], device='cuda:0'), covar=tensor([0.0639, 0.0532, 0.1137, 0.1501, 0.0238, 0.1296, 0.0469, 0.0764], device='cuda:0'), in_proj_covar=tensor([0.0190, 0.0259, 0.0178, 0.0202, 0.0143, 0.0183, 0.0199, 0.0176], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-17 01:11:21,610 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=266464.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 01:11:30,798 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.8497, 2.5310, 3.3368, 2.8600, 3.2315, 3.0209, 2.3800, 3.3161], device='cuda:0'), covar=tensor([0.0163, 0.0355, 0.0190, 0.0262, 0.0170, 0.0195, 0.0394, 0.0149], device='cuda:0'), in_proj_covar=tensor([0.0186, 0.0210, 0.0200, 0.0194, 0.0225, 0.0173, 0.0205, 0.0196], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-17 01:11:34,133 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.831e+02 2.530e+02 2.995e+02 3.621e+02 5.505e+02, threshold=5.991e+02, percent-clipped=0.0 2023-05-17 01:11:34,553 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=4.41 vs. limit=5.0 2023-05-17 01:11:41,973 INFO [finetune.py:992] (0/2) Epoch 14, batch 7450, loss[loss=0.1355, simple_loss=0.2269, pruned_loss=0.02199, over 12201.00 frames. ], tot_loss[loss=0.167, simple_loss=0.2563, pruned_loss=0.03884, over 2369113.38 frames. ], batch size: 29, lr: 3.57e-03, grad_scale: 32.0 2023-05-17 01:12:05,545 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=266525.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 01:12:07,708 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=266528.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 01:12:17,566 INFO [finetune.py:992] (0/2) Epoch 14, batch 7500, loss[loss=0.1776, simple_loss=0.2775, pruned_loss=0.03882, over 12296.00 frames. ], tot_loss[loss=0.166, simple_loss=0.255, pruned_loss=0.03845, over 2372074.06 frames. ], batch size: 33, lr: 3.57e-03, grad_scale: 32.0 2023-05-17 01:12:37,905 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=266570.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 01:12:40,177 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=266573.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 01:12:42,897 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=266576.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 01:12:46,353 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.995e+02 2.645e+02 3.187e+02 3.673e+02 5.507e+02, threshold=6.374e+02, percent-clipped=0.0 2023-05-17 01:12:47,103 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=266582.0, num_to_drop=1, layers_to_drop={1} 2023-05-17 01:12:54,150 INFO [finetune.py:992] (0/2) Epoch 14, batch 7550, loss[loss=0.1942, simple_loss=0.2791, pruned_loss=0.0546, over 8456.00 frames. ], tot_loss[loss=0.1658, simple_loss=0.2547, pruned_loss=0.03844, over 2372043.29 frames. ], batch size: 100, lr: 3.57e-03, grad_scale: 32.0 2023-05-17 01:13:15,491 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=266621.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 01:13:30,113 INFO [finetune.py:992] (0/2) Epoch 14, batch 7600, loss[loss=0.1717, simple_loss=0.2607, pruned_loss=0.0413, over 12352.00 frames. ], tot_loss[loss=0.1666, simple_loss=0.2557, pruned_loss=0.03879, over 2372653.60 frames. ], batch size: 36, lr: 3.57e-03, grad_scale: 32.0 2023-05-17 01:13:56,648 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=266679.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 01:13:57,938 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.664e+02 2.549e+02 3.060e+02 3.704e+02 6.506e+02, threshold=6.120e+02, percent-clipped=2.0 2023-05-17 01:14:05,912 INFO [finetune.py:992] (0/2) Epoch 14, batch 7650, loss[loss=0.1586, simple_loss=0.2542, pruned_loss=0.03146, over 12256.00 frames. ], tot_loss[loss=0.1665, simple_loss=0.2556, pruned_loss=0.03869, over 2365842.72 frames. ], batch size: 37, lr: 3.57e-03, grad_scale: 32.0 2023-05-17 01:14:07,006 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=2.98 vs. limit=5.0 2023-05-17 01:14:13,116 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=266702.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 01:14:14,599 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=266704.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 01:14:22,109 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.36 vs. limit=2.0 2023-05-17 01:14:41,122 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=266740.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 01:14:41,355 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.63 vs. limit=2.0 2023-05-17 01:14:42,303 INFO [finetune.py:992] (0/2) Epoch 14, batch 7700, loss[loss=0.1679, simple_loss=0.2619, pruned_loss=0.037, over 12183.00 frames. ], tot_loss[loss=0.1668, simple_loss=0.2558, pruned_loss=0.03887, over 2368876.52 frames. ], batch size: 35, lr: 3.57e-03, grad_scale: 32.0 2023-05-17 01:14:59,001 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=266765.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 01:15:10,713 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.702e+02 2.682e+02 3.206e+02 3.905e+02 6.001e+02, threshold=6.412e+02, percent-clipped=0.0 2023-05-17 01:15:14,501 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.0428, 4.8385, 4.8254, 4.8511, 4.5578, 4.9704, 5.0486, 5.1518], device='cuda:0'), covar=tensor([0.0223, 0.0176, 0.0187, 0.0363, 0.0733, 0.0313, 0.0152, 0.0189], device='cuda:0'), in_proj_covar=tensor([0.0201, 0.0204, 0.0197, 0.0254, 0.0248, 0.0224, 0.0182, 0.0232], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0005, 0.0004, 0.0004, 0.0003, 0.0005], device='cuda:0') 2023-05-17 01:15:18,607 INFO [finetune.py:992] (0/2) Epoch 14, batch 7750, loss[loss=0.1441, simple_loss=0.2423, pruned_loss=0.02292, over 12306.00 frames. ], tot_loss[loss=0.1669, simple_loss=0.2559, pruned_loss=0.03894, over 2370509.71 frames. ], batch size: 34, lr: 3.57e-03, grad_scale: 32.0 2023-05-17 01:15:23,718 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.9918, 4.7961, 4.7555, 4.7959, 4.4840, 4.9084, 4.9396, 5.1020], device='cuda:0'), covar=tensor([0.0251, 0.0179, 0.0173, 0.0378, 0.0747, 0.0303, 0.0149, 0.0188], device='cuda:0'), in_proj_covar=tensor([0.0201, 0.0204, 0.0197, 0.0254, 0.0248, 0.0224, 0.0182, 0.0232], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0005, 0.0004, 0.0004, 0.0003, 0.0005], device='cuda:0') 2023-05-17 01:15:36,327 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.4601, 3.6509, 3.2122, 3.2326, 2.9232, 2.7839, 3.6452, 2.4092], device='cuda:0'), covar=tensor([0.0430, 0.0133, 0.0216, 0.0226, 0.0424, 0.0374, 0.0143, 0.0489], device='cuda:0'), in_proj_covar=tensor([0.0194, 0.0165, 0.0168, 0.0189, 0.0205, 0.0201, 0.0173, 0.0206], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-17 01:15:38,962 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=266820.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 01:15:54,837 INFO [finetune.py:992] (0/2) Epoch 14, batch 7800, loss[loss=0.1817, simple_loss=0.2798, pruned_loss=0.04182, over 11227.00 frames. ], tot_loss[loss=0.1665, simple_loss=0.2559, pruned_loss=0.0386, over 2376639.84 frames. ], batch size: 55, lr: 3.57e-03, grad_scale: 32.0 2023-05-17 01:16:15,841 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=266870.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 01:16:23,496 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.701e+02 2.578e+02 2.944e+02 3.565e+02 1.039e+03, threshold=5.888e+02, percent-clipped=3.0 2023-05-17 01:16:24,349 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=266882.0, num_to_drop=1, layers_to_drop={2} 2023-05-17 01:16:31,375 INFO [finetune.py:992] (0/2) Epoch 14, batch 7850, loss[loss=0.1701, simple_loss=0.2677, pruned_loss=0.0363, over 12018.00 frames. ], tot_loss[loss=0.1661, simple_loss=0.2555, pruned_loss=0.03831, over 2379273.10 frames. ], batch size: 42, lr: 3.57e-03, grad_scale: 32.0 2023-05-17 01:16:47,534 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.43 vs. limit=2.0 2023-05-17 01:16:49,897 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=266918.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 01:16:52,227 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=266920.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 01:16:59,259 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=266930.0, num_to_drop=1, layers_to_drop={0} 2023-05-17 01:17:07,565 INFO [finetune.py:992] (0/2) Epoch 14, batch 7900, loss[loss=0.1878, simple_loss=0.2765, pruned_loss=0.04956, over 10733.00 frames. ], tot_loss[loss=0.1665, simple_loss=0.256, pruned_loss=0.03853, over 2376575.92 frames. ], batch size: 68, lr: 3.57e-03, grad_scale: 16.0 2023-05-17 01:17:35,582 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=266981.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 01:17:36,068 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.887e+02 2.705e+02 3.150e+02 3.986e+02 7.942e+02, threshold=6.299e+02, percent-clipped=5.0 2023-05-17 01:17:43,131 INFO [finetune.py:992] (0/2) Epoch 14, batch 7950, loss[loss=0.1784, simple_loss=0.263, pruned_loss=0.04691, over 12008.00 frames. ], tot_loss[loss=0.1674, simple_loss=0.2567, pruned_loss=0.03905, over 2368056.01 frames. ], batch size: 40, lr: 3.57e-03, grad_scale: 16.0 2023-05-17 01:17:50,575 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=267002.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 01:18:06,980 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2023-05-17 01:18:14,540 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=267035.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 01:18:19,547 INFO [finetune.py:992] (0/2) Epoch 14, batch 8000, loss[loss=0.1688, simple_loss=0.2584, pruned_loss=0.03962, over 12119.00 frames. ], tot_loss[loss=0.1674, simple_loss=0.2564, pruned_loss=0.03922, over 2369121.58 frames. ], batch size: 38, lr: 3.57e-03, grad_scale: 16.0 2023-05-17 01:18:25,334 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=267050.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 01:18:32,316 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=267060.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 01:18:48,578 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.781e+02 2.680e+02 3.132e+02 3.791e+02 6.983e+02, threshold=6.264e+02, percent-clipped=3.0 2023-05-17 01:18:48,772 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=267082.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 01:18:55,801 INFO [finetune.py:992] (0/2) Epoch 14, batch 8050, loss[loss=0.1559, simple_loss=0.24, pruned_loss=0.03585, over 12187.00 frames. ], tot_loss[loss=0.1676, simple_loss=0.2567, pruned_loss=0.0392, over 2366643.72 frames. ], batch size: 29, lr: 3.57e-03, grad_scale: 16.0 2023-05-17 01:19:00,308 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=267098.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 01:19:15,943 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=267120.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 01:19:31,345 INFO [finetune.py:992] (0/2) Epoch 14, batch 8100, loss[loss=0.1746, simple_loss=0.2688, pruned_loss=0.04018, over 12103.00 frames. ], tot_loss[loss=0.1675, simple_loss=0.2568, pruned_loss=0.03907, over 2368077.21 frames. ], batch size: 32, lr: 3.57e-03, grad_scale: 16.0 2023-05-17 01:19:32,327 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=267143.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 01:19:36,938 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.84 vs. limit=2.0 2023-05-17 01:19:44,510 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=267159.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 01:19:50,790 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=267168.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 01:20:00,808 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.910e+02 2.604e+02 3.178e+02 3.850e+02 8.583e+02, threshold=6.356e+02, percent-clipped=3.0 2023-05-17 01:20:07,920 INFO [finetune.py:992] (0/2) Epoch 14, batch 8150, loss[loss=0.1843, simple_loss=0.2748, pruned_loss=0.04694, over 11601.00 frames. ], tot_loss[loss=0.1676, simple_loss=0.2569, pruned_loss=0.03919, over 2369049.35 frames. ], batch size: 48, lr: 3.57e-03, grad_scale: 16.0 2023-05-17 01:20:21,202 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.1753, 5.0472, 4.9698, 5.0625, 4.7161, 5.1702, 5.1289, 5.3143], device='cuda:0'), covar=tensor([0.0187, 0.0155, 0.0173, 0.0332, 0.0752, 0.0322, 0.0147, 0.0176], device='cuda:0'), in_proj_covar=tensor([0.0201, 0.0204, 0.0198, 0.0255, 0.0248, 0.0224, 0.0182, 0.0232], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0005, 0.0004, 0.0004, 0.0003, 0.0005], device='cuda:0') 2023-05-17 01:20:44,812 INFO [finetune.py:992] (0/2) Epoch 14, batch 8200, loss[loss=0.149, simple_loss=0.2368, pruned_loss=0.03059, over 12174.00 frames. ], tot_loss[loss=0.1665, simple_loss=0.2557, pruned_loss=0.03867, over 2379577.88 frames. ], batch size: 31, lr: 3.57e-03, grad_scale: 16.0 2023-05-17 01:21:09,020 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=267276.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 01:21:13,150 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.956e+02 2.546e+02 3.025e+02 3.497e+02 5.221e+02, threshold=6.051e+02, percent-clipped=0.0 2023-05-17 01:21:20,192 INFO [finetune.py:992] (0/2) Epoch 14, batch 8250, loss[loss=0.1736, simple_loss=0.2636, pruned_loss=0.04183, over 12195.00 frames. ], tot_loss[loss=0.1667, simple_loss=0.2558, pruned_loss=0.03883, over 2375096.93 frames. ], batch size: 35, lr: 3.57e-03, grad_scale: 16.0 2023-05-17 01:21:39,591 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.7497, 2.3836, 3.2147, 2.7695, 3.0287, 2.9577, 2.3577, 3.1190], device='cuda:0'), covar=tensor([0.0166, 0.0372, 0.0174, 0.0258, 0.0176, 0.0193, 0.0338, 0.0164], device='cuda:0'), in_proj_covar=tensor([0.0184, 0.0210, 0.0198, 0.0194, 0.0223, 0.0172, 0.0204, 0.0194], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-17 01:21:41,629 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.8162, 5.5234, 5.0977, 5.1573, 5.6810, 5.0696, 5.1090, 5.0881], device='cuda:0'), covar=tensor([0.1495, 0.1015, 0.1158, 0.1790, 0.0963, 0.2228, 0.2062, 0.1205], device='cuda:0'), in_proj_covar=tensor([0.0357, 0.0501, 0.0404, 0.0451, 0.0471, 0.0436, 0.0403, 0.0392], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-17 01:21:51,563 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=267335.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 01:21:56,313 INFO [finetune.py:992] (0/2) Epoch 14, batch 8300, loss[loss=0.1941, simple_loss=0.29, pruned_loss=0.04914, over 12094.00 frames. ], tot_loss[loss=0.1672, simple_loss=0.2563, pruned_loss=0.03908, over 2365106.82 frames. ], batch size: 42, lr: 3.57e-03, grad_scale: 16.0 2023-05-17 01:22:10,046 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=267360.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 01:22:12,914 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.3639, 4.7185, 2.9901, 2.8141, 4.0379, 2.5501, 3.9475, 3.1978], device='cuda:0'), covar=tensor([0.0830, 0.0620, 0.1128, 0.1560, 0.0295, 0.1458, 0.0619, 0.0873], device='cuda:0'), in_proj_covar=tensor([0.0193, 0.0262, 0.0179, 0.0205, 0.0145, 0.0185, 0.0202, 0.0178], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-17 01:22:25,570 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.047e+02 2.816e+02 3.145e+02 3.654e+02 7.270e+02, threshold=6.290e+02, percent-clipped=2.0 2023-05-17 01:22:26,379 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=267383.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 01:22:30,287 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.83 vs. limit=2.0 2023-05-17 01:22:32,725 INFO [finetune.py:992] (0/2) Epoch 14, batch 8350, loss[loss=0.1562, simple_loss=0.2455, pruned_loss=0.03346, over 12024.00 frames. ], tot_loss[loss=0.1669, simple_loss=0.2562, pruned_loss=0.03885, over 2372561.79 frames. ], batch size: 31, lr: 3.57e-03, grad_scale: 16.0 2023-05-17 01:22:44,698 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=267408.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 01:23:04,297 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.34 vs. limit=2.0 2023-05-17 01:23:05,955 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=267438.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 01:23:08,749 INFO [finetune.py:992] (0/2) Epoch 14, batch 8400, loss[loss=0.1628, simple_loss=0.2609, pruned_loss=0.03233, over 12304.00 frames. ], tot_loss[loss=0.1667, simple_loss=0.256, pruned_loss=0.03873, over 2373220.91 frames. ], batch size: 34, lr: 3.57e-03, grad_scale: 16.0 2023-05-17 01:23:18,316 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=267454.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 01:23:38,235 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.715e+02 2.794e+02 3.288e+02 4.070e+02 9.921e+02, threshold=6.576e+02, percent-clipped=2.0 2023-05-17 01:23:44,896 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2023-05-17 01:23:45,276 INFO [finetune.py:992] (0/2) Epoch 14, batch 8450, loss[loss=0.1434, simple_loss=0.2303, pruned_loss=0.02828, over 12125.00 frames. ], tot_loss[loss=0.1661, simple_loss=0.2552, pruned_loss=0.03853, over 2379340.83 frames. ], batch size: 30, lr: 3.56e-03, grad_scale: 16.0 2023-05-17 01:23:46,888 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=267494.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 01:24:08,151 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.7443, 5.6789, 5.4619, 5.0137, 5.0004, 5.6333, 5.2127, 5.0217], device='cuda:0'), covar=tensor([0.0845, 0.1023, 0.0748, 0.1707, 0.0880, 0.0770, 0.1631, 0.1173], device='cuda:0'), in_proj_covar=tensor([0.0631, 0.0569, 0.0525, 0.0651, 0.0427, 0.0742, 0.0798, 0.0584], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0004, 0.0004, 0.0003], device='cuda:0') 2023-05-17 01:24:21,401 INFO [finetune.py:992] (0/2) Epoch 14, batch 8500, loss[loss=0.1801, simple_loss=0.2691, pruned_loss=0.04555, over 12310.00 frames. ], tot_loss[loss=0.1662, simple_loss=0.2555, pruned_loss=0.03849, over 2375188.87 frames. ], batch size: 34, lr: 3.56e-03, grad_scale: 16.0 2023-05-17 01:24:22,251 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([6.0955, 6.0339, 5.7831, 5.4199, 5.2440, 5.9655, 5.5466, 5.2937], device='cuda:0'), covar=tensor([0.0816, 0.1136, 0.0777, 0.1727, 0.0627, 0.0730, 0.1646, 0.1126], device='cuda:0'), in_proj_covar=tensor([0.0632, 0.0570, 0.0526, 0.0652, 0.0428, 0.0745, 0.0800, 0.0586], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0004, 0.0004, 0.0003], device='cuda:0') 2023-05-17 01:24:30,919 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=267555.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 01:24:37,448 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.4047, 4.9712, 5.3660, 4.7492, 5.0117, 4.7931, 5.3952, 5.0254], device='cuda:0'), covar=tensor([0.0277, 0.0400, 0.0264, 0.0261, 0.0382, 0.0313, 0.0220, 0.0323], device='cuda:0'), in_proj_covar=tensor([0.0269, 0.0273, 0.0299, 0.0274, 0.0271, 0.0272, 0.0247, 0.0222], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-17 01:24:45,953 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=267576.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 01:24:48,083 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.2521, 5.2035, 5.0205, 4.6140, 4.7594, 5.1979, 4.8490, 4.6733], device='cuda:0'), covar=tensor([0.0773, 0.1002, 0.0693, 0.1571, 0.1083, 0.0756, 0.1531, 0.1118], device='cuda:0'), in_proj_covar=tensor([0.0627, 0.0565, 0.0521, 0.0648, 0.0426, 0.0739, 0.0795, 0.0581], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0004, 0.0004, 0.0003], device='cuda:0') 2023-05-17 01:24:50,066 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.866e+02 2.604e+02 3.237e+02 3.803e+02 9.594e+02, threshold=6.474e+02, percent-clipped=1.0 2023-05-17 01:24:58,012 INFO [finetune.py:992] (0/2) Epoch 14, batch 8550, loss[loss=0.1815, simple_loss=0.276, pruned_loss=0.04356, over 11316.00 frames. ], tot_loss[loss=0.1659, simple_loss=0.2554, pruned_loss=0.03824, over 2381414.03 frames. ], batch size: 55, lr: 3.56e-03, grad_scale: 16.0 2023-05-17 01:25:20,932 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=267624.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 01:25:33,770 INFO [finetune.py:992] (0/2) Epoch 14, batch 8600, loss[loss=0.1589, simple_loss=0.2509, pruned_loss=0.03346, over 12113.00 frames. ], tot_loss[loss=0.1662, simple_loss=0.2554, pruned_loss=0.03855, over 2381569.03 frames. ], batch size: 33, lr: 3.56e-03, grad_scale: 16.0 2023-05-17 01:25:40,301 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.2751, 4.5914, 2.7758, 2.4072, 3.8433, 2.4256, 3.9508, 2.9553], device='cuda:0'), covar=tensor([0.0820, 0.0466, 0.1262, 0.1828, 0.0321, 0.1444, 0.0495, 0.0938], device='cuda:0'), in_proj_covar=tensor([0.0195, 0.0264, 0.0182, 0.0208, 0.0146, 0.0187, 0.0203, 0.0180], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-17 01:25:51,898 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.39 vs. limit=2.0 2023-05-17 01:26:02,846 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.852e+02 2.742e+02 3.259e+02 3.597e+02 7.182e+02, threshold=6.518e+02, percent-clipped=1.0 2023-05-17 01:26:09,985 INFO [finetune.py:992] (0/2) Epoch 14, batch 8650, loss[loss=0.1935, simple_loss=0.2863, pruned_loss=0.05028, over 12258.00 frames. ], tot_loss[loss=0.1664, simple_loss=0.2553, pruned_loss=0.03871, over 2385629.41 frames. ], batch size: 37, lr: 3.56e-03, grad_scale: 16.0 2023-05-17 01:26:25,489 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.5445, 4.9106, 3.0763, 2.8414, 4.2579, 2.6439, 4.2134, 3.3589], device='cuda:0'), covar=tensor([0.0712, 0.0480, 0.1138, 0.1490, 0.0263, 0.1351, 0.0447, 0.0850], device='cuda:0'), in_proj_covar=tensor([0.0193, 0.0262, 0.0181, 0.0206, 0.0146, 0.0186, 0.0202, 0.0179], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-17 01:26:43,783 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=267738.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 01:26:46,473 INFO [finetune.py:992] (0/2) Epoch 14, batch 8700, loss[loss=0.1736, simple_loss=0.2581, pruned_loss=0.04459, over 12338.00 frames. ], tot_loss[loss=0.1664, simple_loss=0.2555, pruned_loss=0.03866, over 2384173.66 frames. ], batch size: 31, lr: 3.56e-03, grad_scale: 16.0 2023-05-17 01:26:55,012 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=267754.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 01:27:14,865 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.871e+02 2.664e+02 3.049e+02 3.883e+02 6.983e+02, threshold=6.098e+02, percent-clipped=1.0 2023-05-17 01:27:17,698 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=267786.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 01:27:21,747 INFO [finetune.py:992] (0/2) Epoch 14, batch 8750, loss[loss=0.1523, simple_loss=0.2416, pruned_loss=0.03152, over 12252.00 frames. ], tot_loss[loss=0.1664, simple_loss=0.2554, pruned_loss=0.03868, over 2383410.50 frames. ], batch size: 32, lr: 3.56e-03, grad_scale: 16.0 2023-05-17 01:27:29,434 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=267802.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 01:27:30,470 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-05-17 01:27:35,132 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.5169, 2.4874, 3.6572, 4.4144, 3.9366, 4.5883, 3.9140, 3.2234], device='cuda:0'), covar=tensor([0.0050, 0.0429, 0.0162, 0.0055, 0.0123, 0.0066, 0.0123, 0.0360], device='cuda:0'), in_proj_covar=tensor([0.0091, 0.0124, 0.0106, 0.0080, 0.0104, 0.0118, 0.0097, 0.0139], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-17 01:27:58,320 INFO [finetune.py:992] (0/2) Epoch 14, batch 8800, loss[loss=0.1977, simple_loss=0.2886, pruned_loss=0.05339, over 12285.00 frames. ], tot_loss[loss=0.1662, simple_loss=0.2552, pruned_loss=0.03858, over 2384724.65 frames. ], batch size: 37, lr: 3.56e-03, grad_scale: 16.0 2023-05-17 01:28:04,143 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=267850.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 01:28:20,764 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 2023-05-17 01:28:27,484 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.021e+02 2.878e+02 3.369e+02 4.005e+02 1.931e+03, threshold=6.738e+02, percent-clipped=8.0 2023-05-17 01:28:34,583 INFO [finetune.py:992] (0/2) Epoch 14, batch 8850, loss[loss=0.1676, simple_loss=0.2607, pruned_loss=0.03727, over 12250.00 frames. ], tot_loss[loss=0.1666, simple_loss=0.256, pruned_loss=0.0386, over 2378461.19 frames. ], batch size: 32, lr: 3.56e-03, grad_scale: 16.0 2023-05-17 01:28:49,937 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.6938, 3.7883, 3.2939, 3.2999, 3.0630, 2.8623, 3.7822, 2.5008], device='cuda:0'), covar=tensor([0.0409, 0.0146, 0.0225, 0.0199, 0.0411, 0.0395, 0.0152, 0.0510], device='cuda:0'), in_proj_covar=tensor([0.0197, 0.0167, 0.0169, 0.0191, 0.0206, 0.0203, 0.0175, 0.0207], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-17 01:29:10,539 INFO [finetune.py:992] (0/2) Epoch 14, batch 8900, loss[loss=0.1519, simple_loss=0.2488, pruned_loss=0.02745, over 12280.00 frames. ], tot_loss[loss=0.1666, simple_loss=0.2557, pruned_loss=0.03877, over 2374710.01 frames. ], batch size: 33, lr: 3.56e-03, grad_scale: 16.0 2023-05-17 01:29:21,266 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.1706, 6.1407, 5.7700, 5.7107, 6.2069, 5.5390, 5.6245, 5.6734], device='cuda:0'), covar=tensor([0.1590, 0.0870, 0.1029, 0.1989, 0.0938, 0.2227, 0.2230, 0.1197], device='cuda:0'), in_proj_covar=tensor([0.0363, 0.0506, 0.0404, 0.0454, 0.0472, 0.0436, 0.0405, 0.0395], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-17 01:29:38,327 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.5591, 4.4144, 4.5429, 4.5433, 4.1190, 4.0869, 4.0904, 4.4299], device='cuda:0'), covar=tensor([0.1136, 0.0914, 0.1378, 0.0975, 0.2667, 0.2138, 0.0845, 0.1433], device='cuda:0'), in_proj_covar=tensor([0.0557, 0.0720, 0.0635, 0.0662, 0.0875, 0.0766, 0.0580, 0.0496], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0005, 0.0004, 0.0004, 0.0005, 0.0005, 0.0004, 0.0003], device='cuda:0') 2023-05-17 01:29:39,547 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.954e+02 2.749e+02 3.348e+02 3.854e+02 9.928e+02, threshold=6.695e+02, percent-clipped=3.0 2023-05-17 01:29:45,994 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.4733, 5.3001, 5.3914, 5.3997, 5.0907, 5.0171, 4.8560, 5.4105], device='cuda:0'), covar=tensor([0.0684, 0.0554, 0.0835, 0.0670, 0.1695, 0.1384, 0.0558, 0.0888], device='cuda:0'), in_proj_covar=tensor([0.0556, 0.0718, 0.0634, 0.0660, 0.0873, 0.0764, 0.0579, 0.0495], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0005, 0.0004, 0.0004, 0.0005, 0.0005, 0.0004, 0.0003], device='cuda:0') 2023-05-17 01:29:46,569 INFO [finetune.py:992] (0/2) Epoch 14, batch 8950, loss[loss=0.1718, simple_loss=0.2652, pruned_loss=0.03917, over 12262.00 frames. ], tot_loss[loss=0.1666, simple_loss=0.2556, pruned_loss=0.03876, over 2369764.15 frames. ], batch size: 32, lr: 3.56e-03, grad_scale: 16.0 2023-05-17 01:29:52,406 INFO [checkpoint.py:75] (0/2) Saving checkpoint to pruned_transducer_stateless7/exp_giga_finetune/checkpoint-168000.pt 2023-05-17 01:30:21,451 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.5038, 2.6703, 3.1520, 4.2705, 2.3983, 4.3417, 4.4054, 4.5074], device='cuda:0'), covar=tensor([0.0139, 0.1263, 0.0507, 0.0166, 0.1322, 0.0330, 0.0173, 0.0117], device='cuda:0'), in_proj_covar=tensor([0.0126, 0.0205, 0.0186, 0.0123, 0.0193, 0.0183, 0.0178, 0.0126], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-17 01:30:24,383 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.3912, 2.9299, 3.7739, 2.3663, 2.6766, 3.0411, 2.9068, 3.1727], device='cuda:0'), covar=tensor([0.0588, 0.1132, 0.0438, 0.1147, 0.1705, 0.1471, 0.1249, 0.1112], device='cuda:0'), in_proj_covar=tensor([0.0240, 0.0238, 0.0256, 0.0183, 0.0238, 0.0298, 0.0225, 0.0267], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-17 01:30:26,259 INFO [finetune.py:992] (0/2) Epoch 14, batch 9000, loss[loss=0.1582, simple_loss=0.25, pruned_loss=0.03321, over 12332.00 frames. ], tot_loss[loss=0.1655, simple_loss=0.2546, pruned_loss=0.03817, over 2373449.45 frames. ], batch size: 31, lr: 3.56e-03, grad_scale: 16.0 2023-05-17 01:30:26,260 INFO [finetune.py:1017] (0/2) Computing validation loss 2023-05-17 01:30:44,387 INFO [finetune.py:1026] (0/2) Epoch 14, validation: loss=0.3235, simple_loss=0.3964, pruned_loss=0.1253, over 1020973.00 frames. 2023-05-17 01:30:44,387 INFO [finetune.py:1027] (0/2) Maximum memory allocated so far is 12827MB 2023-05-17 01:31:13,641 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.779e+02 2.652e+02 3.054e+02 3.579e+02 6.577e+02, threshold=6.108e+02, percent-clipped=0.0 2023-05-17 01:31:20,735 INFO [finetune.py:992] (0/2) Epoch 14, batch 9050, loss[loss=0.1647, simple_loss=0.2497, pruned_loss=0.03992, over 12261.00 frames. ], tot_loss[loss=0.1657, simple_loss=0.2547, pruned_loss=0.03835, over 2374362.42 frames. ], batch size: 32, lr: 3.56e-03, grad_scale: 16.0 2023-05-17 01:31:33,075 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.4332, 2.9134, 3.9787, 3.3106, 3.8206, 3.3589, 2.9083, 3.8513], device='cuda:0'), covar=tensor([0.0138, 0.0319, 0.0138, 0.0261, 0.0144, 0.0193, 0.0331, 0.0135], device='cuda:0'), in_proj_covar=tensor([0.0185, 0.0209, 0.0197, 0.0193, 0.0223, 0.0172, 0.0203, 0.0195], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-17 01:31:52,041 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.3984, 5.2014, 5.3348, 5.3261, 4.9420, 5.0202, 4.8042, 5.2703], device='cuda:0'), covar=tensor([0.0643, 0.0591, 0.0875, 0.0650, 0.2108, 0.1345, 0.0567, 0.0983], device='cuda:0'), in_proj_covar=tensor([0.0552, 0.0714, 0.0631, 0.0659, 0.0868, 0.0759, 0.0578, 0.0493], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0005, 0.0004, 0.0004, 0.0005, 0.0005, 0.0004, 0.0003], device='cuda:0') 2023-05-17 01:31:56,831 INFO [finetune.py:992] (0/2) Epoch 14, batch 9100, loss[loss=0.1625, simple_loss=0.2596, pruned_loss=0.0327, over 12038.00 frames. ], tot_loss[loss=0.1664, simple_loss=0.2556, pruned_loss=0.03859, over 2373024.92 frames. ], batch size: 42, lr: 3.56e-03, grad_scale: 16.0 2023-05-17 01:32:02,621 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=268150.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 01:32:18,584 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.40 vs. limit=2.0 2023-05-17 01:32:25,037 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.134e+02 2.594e+02 3.160e+02 3.768e+02 7.595e+02, threshold=6.320e+02, percent-clipped=1.0 2023-05-17 01:32:32,968 INFO [finetune.py:992] (0/2) Epoch 14, batch 9150, loss[loss=0.152, simple_loss=0.2424, pruned_loss=0.03083, over 12306.00 frames. ], tot_loss[loss=0.1666, simple_loss=0.2558, pruned_loss=0.03865, over 2376947.18 frames. ], batch size: 34, lr: 3.56e-03, grad_scale: 16.0 2023-05-17 01:32:37,232 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=268198.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 01:32:47,542 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=268212.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 01:33:08,908 INFO [finetune.py:992] (0/2) Epoch 14, batch 9200, loss[loss=0.1445, simple_loss=0.2329, pruned_loss=0.02801, over 12182.00 frames. ], tot_loss[loss=0.1655, simple_loss=0.2551, pruned_loss=0.03799, over 2375910.15 frames. ], batch size: 31, lr: 3.56e-03, grad_scale: 16.0 2023-05-17 01:33:31,665 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=268273.0, num_to_drop=1, layers_to_drop={3} 2023-05-17 01:33:33,085 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=268275.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 01:33:36,147 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2023-05-17 01:33:37,945 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.774e+02 2.664e+02 3.128e+02 3.850e+02 8.698e+02, threshold=6.257e+02, percent-clipped=2.0 2023-05-17 01:33:45,157 INFO [finetune.py:992] (0/2) Epoch 14, batch 9250, loss[loss=0.1725, simple_loss=0.2677, pruned_loss=0.03862, over 10433.00 frames. ], tot_loss[loss=0.1656, simple_loss=0.2553, pruned_loss=0.03801, over 2376895.63 frames. ], batch size: 68, lr: 3.56e-03, grad_scale: 16.0 2023-05-17 01:33:54,151 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.12 vs. limit=2.0 2023-05-17 01:34:17,449 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=268336.0, num_to_drop=1, layers_to_drop={0} 2023-05-17 01:34:21,584 INFO [finetune.py:992] (0/2) Epoch 14, batch 9300, loss[loss=0.1621, simple_loss=0.2418, pruned_loss=0.04119, over 12355.00 frames. ], tot_loss[loss=0.1664, simple_loss=0.2559, pruned_loss=0.03844, over 2371125.86 frames. ], batch size: 31, lr: 3.56e-03, grad_scale: 16.0 2023-05-17 01:34:49,812 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.762e+02 2.778e+02 3.151e+02 3.808e+02 5.509e+02, threshold=6.302e+02, percent-clipped=0.0 2023-05-17 01:34:56,957 INFO [finetune.py:992] (0/2) Epoch 14, batch 9350, loss[loss=0.166, simple_loss=0.2574, pruned_loss=0.03728, over 12126.00 frames. ], tot_loss[loss=0.1653, simple_loss=0.2548, pruned_loss=0.03786, over 2377134.28 frames. ], batch size: 33, lr: 3.56e-03, grad_scale: 16.0 2023-05-17 01:35:33,566 INFO [finetune.py:992] (0/2) Epoch 14, batch 9400, loss[loss=0.1727, simple_loss=0.2634, pruned_loss=0.04097, over 12382.00 frames. ], tot_loss[loss=0.1649, simple_loss=0.2545, pruned_loss=0.03766, over 2383986.35 frames. ], batch size: 38, lr: 3.56e-03, grad_scale: 16.0 2023-05-17 01:35:48,874 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.9866, 3.5795, 5.3270, 2.7289, 3.0008, 3.8279, 3.3944, 3.8175], device='cuda:0'), covar=tensor([0.0362, 0.1018, 0.0298, 0.1136, 0.1765, 0.1646, 0.1214, 0.1150], device='cuda:0'), in_proj_covar=tensor([0.0236, 0.0235, 0.0253, 0.0181, 0.0235, 0.0295, 0.0224, 0.0264], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-17 01:36:02,121 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.649e+02 2.584e+02 2.948e+02 3.482e+02 7.106e+02, threshold=5.896e+02, percent-clipped=2.0 2023-05-17 01:36:09,261 INFO [finetune.py:992] (0/2) Epoch 14, batch 9450, loss[loss=0.1313, simple_loss=0.2114, pruned_loss=0.02564, over 12294.00 frames. ], tot_loss[loss=0.1644, simple_loss=0.2541, pruned_loss=0.03732, over 2385395.64 frames. ], batch size: 28, lr: 3.56e-03, grad_scale: 16.0 2023-05-17 01:36:10,053 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([6.0693, 6.0414, 5.7733, 5.2277, 5.1137, 5.9247, 5.5381, 5.2550], device='cuda:0'), covar=tensor([0.0781, 0.0987, 0.0727, 0.1632, 0.0793, 0.0707, 0.1684, 0.1095], device='cuda:0'), in_proj_covar=tensor([0.0639, 0.0577, 0.0530, 0.0661, 0.0435, 0.0751, 0.0814, 0.0590], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0004, 0.0004, 0.0003], device='cuda:0') 2023-05-17 01:36:20,798 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.0861, 4.6951, 4.8417, 4.9478, 4.7998, 4.9276, 4.8553, 2.6447], device='cuda:0'), covar=tensor([0.0114, 0.0080, 0.0102, 0.0060, 0.0050, 0.0101, 0.0107, 0.0836], device='cuda:0'), in_proj_covar=tensor([0.0071, 0.0082, 0.0085, 0.0076, 0.0063, 0.0096, 0.0084, 0.0100], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-17 01:36:30,461 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=4.45 vs. limit=5.0 2023-05-17 01:36:33,118 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.7093, 3.9972, 3.5649, 4.2235, 3.8730, 2.6883, 3.6666, 2.8658], device='cuda:0'), covar=tensor([0.0998, 0.1006, 0.1662, 0.0707, 0.1372, 0.1830, 0.1259, 0.3362], device='cuda:0'), in_proj_covar=tensor([0.0309, 0.0385, 0.0364, 0.0325, 0.0374, 0.0276, 0.0355, 0.0372], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-17 01:36:45,614 INFO [finetune.py:992] (0/2) Epoch 14, batch 9500, loss[loss=0.1881, simple_loss=0.2834, pruned_loss=0.04639, over 12345.00 frames. ], tot_loss[loss=0.1645, simple_loss=0.2543, pruned_loss=0.03737, over 2391107.91 frames. ], batch size: 36, lr: 3.56e-03, grad_scale: 16.0 2023-05-17 01:37:04,954 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=268568.0, num_to_drop=1, layers_to_drop={0} 2023-05-17 01:37:14,969 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.916e+02 2.701e+02 3.279e+02 3.811e+02 1.640e+03, threshold=6.558e+02, percent-clipped=2.0 2023-05-17 01:37:22,005 INFO [finetune.py:992] (0/2) Epoch 14, batch 9550, loss[loss=0.1714, simple_loss=0.2649, pruned_loss=0.03891, over 12202.00 frames. ], tot_loss[loss=0.1642, simple_loss=0.254, pruned_loss=0.03718, over 2392038.96 frames. ], batch size: 35, lr: 3.56e-03, grad_scale: 16.0 2023-05-17 01:37:25,713 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.5073, 4.9938, 5.4626, 4.7833, 5.0907, 4.8106, 5.5510, 5.1202], device='cuda:0'), covar=tensor([0.0273, 0.0433, 0.0291, 0.0286, 0.0412, 0.0394, 0.0191, 0.0281], device='cuda:0'), in_proj_covar=tensor([0.0269, 0.0273, 0.0299, 0.0273, 0.0272, 0.0273, 0.0247, 0.0221], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-17 01:37:28,810 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=268601.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 01:37:45,513 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.7022, 2.4999, 3.3493, 4.5019, 2.4340, 4.6265, 4.6986, 4.7063], device='cuda:0'), covar=tensor([0.0132, 0.1434, 0.0514, 0.0140, 0.1411, 0.0216, 0.0151, 0.0117], device='cuda:0'), in_proj_covar=tensor([0.0124, 0.0205, 0.0185, 0.0123, 0.0192, 0.0181, 0.0177, 0.0125], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-17 01:37:49,779 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=268631.0, num_to_drop=1, layers_to_drop={0} 2023-05-17 01:37:58,109 INFO [finetune.py:992] (0/2) Epoch 14, batch 9600, loss[loss=0.1454, simple_loss=0.2397, pruned_loss=0.0255, over 12069.00 frames. ], tot_loss[loss=0.1641, simple_loss=0.2537, pruned_loss=0.03729, over 2386798.68 frames. ], batch size: 32, lr: 3.56e-03, grad_scale: 16.0 2023-05-17 01:38:12,628 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=268662.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 01:38:17,712 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.0882, 4.7042, 4.9093, 4.9824, 4.8314, 4.9078, 4.8334, 2.8421], device='cuda:0'), covar=tensor([0.0096, 0.0078, 0.0085, 0.0052, 0.0041, 0.0105, 0.0091, 0.0745], device='cuda:0'), in_proj_covar=tensor([0.0070, 0.0081, 0.0085, 0.0075, 0.0063, 0.0095, 0.0084, 0.0099], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-17 01:38:26,700 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.698e+02 2.534e+02 3.172e+02 3.807e+02 1.348e+03, threshold=6.344e+02, percent-clipped=3.0 2023-05-17 01:38:33,749 INFO [finetune.py:992] (0/2) Epoch 14, batch 9650, loss[loss=0.1849, simple_loss=0.2771, pruned_loss=0.04634, over 12365.00 frames. ], tot_loss[loss=0.1648, simple_loss=0.2543, pruned_loss=0.03766, over 2383376.92 frames. ], batch size: 38, lr: 3.56e-03, grad_scale: 16.0 2023-05-17 01:39:10,284 INFO [finetune.py:992] (0/2) Epoch 14, batch 9700, loss[loss=0.1704, simple_loss=0.2599, pruned_loss=0.04043, over 12034.00 frames. ], tot_loss[loss=0.165, simple_loss=0.2542, pruned_loss=0.03786, over 2382509.77 frames. ], batch size: 37, lr: 3.56e-03, grad_scale: 16.0 2023-05-17 01:39:32,766 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=2.58 vs. limit=5.0 2023-05-17 01:39:38,637 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.791e+02 2.581e+02 3.284e+02 3.800e+02 7.476e+02, threshold=6.568e+02, percent-clipped=3.0 2023-05-17 01:39:46,390 INFO [finetune.py:992] (0/2) Epoch 14, batch 9750, loss[loss=0.2091, simple_loss=0.2859, pruned_loss=0.06621, over 8269.00 frames. ], tot_loss[loss=0.1653, simple_loss=0.2547, pruned_loss=0.03799, over 2370960.12 frames. ], batch size: 99, lr: 3.55e-03, grad_scale: 16.0 2023-05-17 01:39:56,235 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.7818, 2.9448, 4.7227, 4.8011, 2.9185, 2.7381, 3.1160, 2.1963], device='cuda:0'), covar=tensor([0.1626, 0.2930, 0.0438, 0.0427, 0.1371, 0.2450, 0.2552, 0.4105], device='cuda:0'), in_proj_covar=tensor([0.0300, 0.0390, 0.0278, 0.0305, 0.0274, 0.0311, 0.0391, 0.0375], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-17 01:40:14,051 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-05-17 01:40:22,314 INFO [finetune.py:992] (0/2) Epoch 14, batch 9800, loss[loss=0.1926, simple_loss=0.2862, pruned_loss=0.04956, over 12197.00 frames. ], tot_loss[loss=0.166, simple_loss=0.2552, pruned_loss=0.03835, over 2371452.22 frames. ], batch size: 35, lr: 3.55e-03, grad_scale: 16.0 2023-05-17 01:40:27,499 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.6062, 2.7395, 3.7651, 4.5328, 4.0031, 4.6453, 4.0767, 3.3813], device='cuda:0'), covar=tensor([0.0037, 0.0369, 0.0150, 0.0045, 0.0122, 0.0051, 0.0119, 0.0341], device='cuda:0'), in_proj_covar=tensor([0.0093, 0.0127, 0.0108, 0.0081, 0.0106, 0.0119, 0.0100, 0.0141], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-17 01:40:41,527 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=268868.0, num_to_drop=1, layers_to_drop={2} 2023-05-17 01:40:50,293 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2023-05-17 01:40:51,365 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.847e+02 2.816e+02 3.389e+02 4.188e+02 5.492e+02, threshold=6.779e+02, percent-clipped=0.0 2023-05-17 01:40:58,417 INFO [finetune.py:992] (0/2) Epoch 14, batch 9850, loss[loss=0.153, simple_loss=0.2357, pruned_loss=0.03512, over 12037.00 frames. ], tot_loss[loss=0.166, simple_loss=0.2552, pruned_loss=0.03846, over 2369273.93 frames. ], batch size: 31, lr: 3.55e-03, grad_scale: 16.0 2023-05-17 01:41:00,096 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=268894.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 01:41:03,571 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.2346, 5.0427, 5.1792, 5.1855, 4.8356, 4.8740, 4.6635, 5.0929], device='cuda:0'), covar=tensor([0.0613, 0.0599, 0.0833, 0.0584, 0.1847, 0.1332, 0.0560, 0.1000], device='cuda:0'), in_proj_covar=tensor([0.0550, 0.0710, 0.0622, 0.0654, 0.0864, 0.0753, 0.0572, 0.0485], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0005, 0.0005, 0.0003, 0.0003], device='cuda:0') 2023-05-17 01:41:15,624 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=268916.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 01:41:15,831 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.6828, 4.4115, 4.2514, 4.6262, 3.6359, 4.2065, 3.2502, 4.3925], device='cuda:0'), covar=tensor([0.1400, 0.0622, 0.0906, 0.0713, 0.0963, 0.0549, 0.1473, 0.1130], device='cuda:0'), in_proj_covar=tensor([0.0232, 0.0269, 0.0303, 0.0365, 0.0244, 0.0246, 0.0263, 0.0381], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-17 01:41:22,234 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.5193, 4.8838, 3.1444, 3.0183, 4.1862, 3.0036, 4.0868, 3.4659], device='cuda:0'), covar=tensor([0.0713, 0.0586, 0.1047, 0.1367, 0.0347, 0.1190, 0.0537, 0.0775], device='cuda:0'), in_proj_covar=tensor([0.0193, 0.0263, 0.0181, 0.0205, 0.0145, 0.0186, 0.0203, 0.0179], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-17 01:41:27,159 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=268931.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 01:41:34,904 INFO [finetune.py:992] (0/2) Epoch 14, batch 9900, loss[loss=0.1789, simple_loss=0.2701, pruned_loss=0.04384, over 11230.00 frames. ], tot_loss[loss=0.1663, simple_loss=0.2553, pruned_loss=0.03864, over 2365044.20 frames. ], batch size: 55, lr: 3.55e-03, grad_scale: 32.0 2023-05-17 01:41:44,322 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=268955.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 01:41:45,595 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=268957.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 01:42:00,983 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=268979.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 01:42:03,057 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.779e+02 2.743e+02 3.231e+02 3.791e+02 6.341e+02, threshold=6.462e+02, percent-clipped=0.0 2023-05-17 01:42:10,411 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=268991.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 01:42:10,916 INFO [finetune.py:992] (0/2) Epoch 14, batch 9950, loss[loss=0.1465, simple_loss=0.2366, pruned_loss=0.02824, over 12024.00 frames. ], tot_loss[loss=0.1661, simple_loss=0.2555, pruned_loss=0.03838, over 2370302.22 frames. ], batch size: 31, lr: 3.55e-03, grad_scale: 32.0 2023-05-17 01:42:16,051 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.9841, 3.9100, 3.9412, 4.0077, 3.7633, 3.7987, 3.6484, 3.9149], device='cuda:0'), covar=tensor([0.1011, 0.0735, 0.1311, 0.0790, 0.1891, 0.1339, 0.0692, 0.1006], device='cuda:0'), in_proj_covar=tensor([0.0550, 0.0712, 0.0625, 0.0655, 0.0867, 0.0756, 0.0575, 0.0487], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0005, 0.0005, 0.0004, 0.0003], device='cuda:0') 2023-05-17 01:42:47,297 INFO [finetune.py:992] (0/2) Epoch 14, batch 10000, loss[loss=0.1369, simple_loss=0.2239, pruned_loss=0.02499, over 12409.00 frames. ], tot_loss[loss=0.1652, simple_loss=0.2546, pruned_loss=0.03792, over 2375979.45 frames. ], batch size: 32, lr: 3.55e-03, grad_scale: 32.0 2023-05-17 01:42:54,647 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=269052.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 01:43:07,490 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.1287, 2.0582, 2.5602, 3.1270, 2.1853, 3.2333, 3.0665, 3.2567], device='cuda:0'), covar=tensor([0.0200, 0.1270, 0.0560, 0.0208, 0.1164, 0.0368, 0.0395, 0.0172], device='cuda:0'), in_proj_covar=tensor([0.0125, 0.0206, 0.0186, 0.0124, 0.0193, 0.0182, 0.0179, 0.0126], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-17 01:43:16,320 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.894e+02 2.649e+02 3.101e+02 3.484e+02 6.410e+02, threshold=6.202e+02, percent-clipped=0.0 2023-05-17 01:43:23,338 INFO [finetune.py:992] (0/2) Epoch 14, batch 10050, loss[loss=0.1374, simple_loss=0.2214, pruned_loss=0.02669, over 12190.00 frames. ], tot_loss[loss=0.1651, simple_loss=0.2544, pruned_loss=0.0379, over 2372593.94 frames. ], batch size: 29, lr: 3.55e-03, grad_scale: 32.0 2023-05-17 01:43:26,397 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.5736, 4.5520, 4.4101, 4.0940, 4.1652, 4.5679, 4.2679, 4.0418], device='cuda:0'), covar=tensor([0.1041, 0.1209, 0.0868, 0.1569, 0.2056, 0.0869, 0.1758, 0.1501], device='cuda:0'), in_proj_covar=tensor([0.0638, 0.0572, 0.0525, 0.0657, 0.0431, 0.0749, 0.0805, 0.0586], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0004, 0.0004, 0.0003], device='cuda:0') 2023-05-17 01:43:38,933 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=5.37 vs. limit=5.0 2023-05-17 01:43:44,301 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.58 vs. limit=2.0 2023-05-17 01:43:45,963 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.8406, 5.6627, 5.3194, 5.1591, 5.8087, 5.1104, 5.1167, 5.1508], device='cuda:0'), covar=tensor([0.1572, 0.1015, 0.1060, 0.2069, 0.0888, 0.2217, 0.2069, 0.1141], device='cuda:0'), in_proj_covar=tensor([0.0359, 0.0501, 0.0402, 0.0454, 0.0470, 0.0435, 0.0405, 0.0391], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-17 01:43:59,163 INFO [finetune.py:992] (0/2) Epoch 14, batch 10100, loss[loss=0.1522, simple_loss=0.2396, pruned_loss=0.03245, over 12075.00 frames. ], tot_loss[loss=0.1657, simple_loss=0.2549, pruned_loss=0.03821, over 2370638.70 frames. ], batch size: 32, lr: 3.55e-03, grad_scale: 16.0 2023-05-17 01:44:12,558 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([6.0251, 6.0117, 5.7822, 5.3242, 5.1452, 5.9400, 5.5688, 5.2811], device='cuda:0'), covar=tensor([0.0681, 0.0838, 0.0668, 0.1633, 0.0645, 0.0689, 0.1480, 0.1083], device='cuda:0'), in_proj_covar=tensor([0.0635, 0.0568, 0.0525, 0.0656, 0.0430, 0.0745, 0.0803, 0.0584], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0004, 0.0004, 0.0003], device='cuda:0') 2023-05-17 01:44:28,227 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.305e+02 2.721e+02 3.314e+02 4.067e+02 1.296e+03, threshold=6.628e+02, percent-clipped=5.0 2023-05-17 01:44:34,825 INFO [finetune.py:992] (0/2) Epoch 14, batch 10150, loss[loss=0.2011, simple_loss=0.2883, pruned_loss=0.05693, over 12379.00 frames. ], tot_loss[loss=0.1661, simple_loss=0.2551, pruned_loss=0.03856, over 2373465.43 frames. ], batch size: 38, lr: 3.55e-03, grad_scale: 16.0 2023-05-17 01:44:40,085 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.2534, 5.0716, 5.2285, 5.2549, 4.7975, 4.9126, 4.6602, 5.2384], device='cuda:0'), covar=tensor([0.0697, 0.0631, 0.0745, 0.0566, 0.1971, 0.1285, 0.0603, 0.0885], device='cuda:0'), in_proj_covar=tensor([0.0553, 0.0715, 0.0630, 0.0654, 0.0870, 0.0756, 0.0577, 0.0490], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0005, 0.0004, 0.0004, 0.0005, 0.0005, 0.0004, 0.0003], device='cuda:0') 2023-05-17 01:45:10,495 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=269241.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 01:45:11,058 INFO [finetune.py:992] (0/2) Epoch 14, batch 10200, loss[loss=0.1661, simple_loss=0.2646, pruned_loss=0.03377, over 12266.00 frames. ], tot_loss[loss=0.1667, simple_loss=0.2557, pruned_loss=0.0389, over 2375799.42 frames. ], batch size: 37, lr: 3.55e-03, grad_scale: 16.0 2023-05-17 01:45:16,821 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=269250.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 01:45:19,785 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.9894, 5.7077, 5.3522, 5.2442, 5.8507, 5.0683, 5.2943, 5.3380], device='cuda:0'), covar=tensor([0.1357, 0.0941, 0.1277, 0.1850, 0.0824, 0.2072, 0.1873, 0.1047], device='cuda:0'), in_proj_covar=tensor([0.0359, 0.0501, 0.0401, 0.0455, 0.0469, 0.0434, 0.0404, 0.0390], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-17 01:45:22,005 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=269257.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 01:45:40,049 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.763e+02 2.691e+02 3.133e+02 3.921e+02 6.696e+02, threshold=6.266e+02, percent-clipped=1.0 2023-05-17 01:45:47,112 INFO [finetune.py:992] (0/2) Epoch 14, batch 10250, loss[loss=0.1597, simple_loss=0.2396, pruned_loss=0.03992, over 12247.00 frames. ], tot_loss[loss=0.167, simple_loss=0.256, pruned_loss=0.03899, over 2375495.10 frames. ], batch size: 32, lr: 3.55e-03, grad_scale: 16.0 2023-05-17 01:45:54,471 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=269302.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 01:45:56,483 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=269305.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 01:46:19,266 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=3.36 vs. limit=5.0 2023-05-17 01:46:22,515 INFO [finetune.py:992] (0/2) Epoch 14, batch 10300, loss[loss=0.1535, simple_loss=0.2426, pruned_loss=0.0322, over 12179.00 frames. ], tot_loss[loss=0.1676, simple_loss=0.2567, pruned_loss=0.0393, over 2369304.08 frames. ], batch size: 31, lr: 3.55e-03, grad_scale: 16.0 2023-05-17 01:46:25,493 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.2209, 6.0819, 5.6307, 5.6498, 6.1297, 5.4010, 5.4723, 5.5861], device='cuda:0'), covar=tensor([0.1407, 0.0837, 0.1213, 0.1683, 0.0907, 0.2064, 0.2130, 0.1210], device='cuda:0'), in_proj_covar=tensor([0.0359, 0.0500, 0.0400, 0.0455, 0.0468, 0.0432, 0.0404, 0.0390], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-17 01:46:26,231 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=269347.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 01:46:29,853 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=269352.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 01:46:34,763 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.9989, 2.3691, 3.6259, 2.9635, 3.3649, 3.1645, 2.5962, 3.4671], device='cuda:0'), covar=tensor([0.0208, 0.0451, 0.0204, 0.0323, 0.0200, 0.0224, 0.0435, 0.0175], device='cuda:0'), in_proj_covar=tensor([0.0189, 0.0213, 0.0200, 0.0197, 0.0227, 0.0175, 0.0206, 0.0199], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-17 01:46:52,126 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.801e+02 2.825e+02 3.136e+02 3.593e+02 1.295e+03, threshold=6.272e+02, percent-clipped=3.0 2023-05-17 01:46:58,649 INFO [finetune.py:992] (0/2) Epoch 14, batch 10350, loss[loss=0.1418, simple_loss=0.2284, pruned_loss=0.02759, over 12185.00 frames. ], tot_loss[loss=0.1667, simple_loss=0.2559, pruned_loss=0.0387, over 2373452.72 frames. ], batch size: 29, lr: 3.55e-03, grad_scale: 16.0 2023-05-17 01:47:14,157 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=269413.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 01:47:20,376 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.0704, 5.9093, 5.5381, 5.4852, 6.0130, 5.2931, 5.4989, 5.4939], device='cuda:0'), covar=tensor([0.1550, 0.0879, 0.1002, 0.1757, 0.0931, 0.1982, 0.1878, 0.1217], device='cuda:0'), in_proj_covar=tensor([0.0360, 0.0503, 0.0402, 0.0458, 0.0471, 0.0434, 0.0405, 0.0391], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-17 01:47:29,772 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.3118, 4.1976, 4.1641, 4.6037, 3.1679, 4.1103, 2.8467, 4.3521], device='cuda:0'), covar=tensor([0.1534, 0.0642, 0.1011, 0.0569, 0.1066, 0.0553, 0.1674, 0.0958], device='cuda:0'), in_proj_covar=tensor([0.0232, 0.0269, 0.0304, 0.0365, 0.0244, 0.0245, 0.0263, 0.0380], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-17 01:47:34,726 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=269441.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 01:47:35,278 INFO [finetune.py:992] (0/2) Epoch 14, batch 10400, loss[loss=0.144, simple_loss=0.2333, pruned_loss=0.02734, over 12121.00 frames. ], tot_loss[loss=0.1664, simple_loss=0.2556, pruned_loss=0.03861, over 2368569.70 frames. ], batch size: 30, lr: 3.55e-03, grad_scale: 16.0 2023-05-17 01:48:04,618 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.686e+02 2.696e+02 3.077e+02 3.789e+02 7.478e+02, threshold=6.154e+02, percent-clipped=3.0 2023-05-17 01:48:11,105 INFO [finetune.py:992] (0/2) Epoch 14, batch 10450, loss[loss=0.1303, simple_loss=0.2109, pruned_loss=0.02486, over 12259.00 frames. ], tot_loss[loss=0.1649, simple_loss=0.2543, pruned_loss=0.0377, over 2375529.53 frames. ], batch size: 28, lr: 3.55e-03, grad_scale: 16.0 2023-05-17 01:48:18,184 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=269502.0, num_to_drop=1, layers_to_drop={1} 2023-05-17 01:48:34,838 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.95 vs. limit=2.0 2023-05-17 01:48:47,300 INFO [finetune.py:992] (0/2) Epoch 14, batch 10500, loss[loss=0.1354, simple_loss=0.2247, pruned_loss=0.02303, over 12310.00 frames. ], tot_loss[loss=0.165, simple_loss=0.2544, pruned_loss=0.03782, over 2368524.59 frames. ], batch size: 28, lr: 3.55e-03, grad_scale: 16.0 2023-05-17 01:48:53,289 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=269550.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 01:49:17,244 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.959e+02 2.582e+02 3.202e+02 3.705e+02 6.357e+02, threshold=6.404e+02, percent-clipped=1.0 2023-05-17 01:49:23,432 INFO [finetune.py:992] (0/2) Epoch 14, batch 10550, loss[loss=0.1822, simple_loss=0.2791, pruned_loss=0.04266, over 12205.00 frames. ], tot_loss[loss=0.1653, simple_loss=0.2546, pruned_loss=0.03795, over 2369402.36 frames. ], batch size: 35, lr: 3.55e-03, grad_scale: 16.0 2023-05-17 01:49:27,117 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=269597.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 01:49:27,809 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=269598.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 01:49:31,818 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.4364, 2.6386, 3.6311, 4.4489, 3.9415, 4.5409, 3.8934, 3.1757], device='cuda:0'), covar=tensor([0.0046, 0.0404, 0.0150, 0.0039, 0.0100, 0.0060, 0.0111, 0.0364], device='cuda:0'), in_proj_covar=tensor([0.0093, 0.0127, 0.0108, 0.0080, 0.0105, 0.0118, 0.0099, 0.0140], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-17 01:49:32,556 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.6249, 3.8064, 3.2833, 3.2513, 3.0375, 2.8749, 3.7876, 2.3713], device='cuda:0'), covar=tensor([0.0396, 0.0115, 0.0194, 0.0212, 0.0363, 0.0351, 0.0134, 0.0504], device='cuda:0'), in_proj_covar=tensor([0.0200, 0.0169, 0.0172, 0.0195, 0.0207, 0.0205, 0.0177, 0.0211], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-17 01:49:57,566 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.4110, 2.9603, 3.9004, 3.3442, 3.7998, 3.4592, 2.8990, 3.8486], device='cuda:0'), covar=tensor([0.0129, 0.0301, 0.0150, 0.0263, 0.0136, 0.0182, 0.0308, 0.0130], device='cuda:0'), in_proj_covar=tensor([0.0186, 0.0210, 0.0198, 0.0194, 0.0224, 0.0173, 0.0203, 0.0195], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-17 01:49:58,823 INFO [finetune.py:992] (0/2) Epoch 14, batch 10600, loss[loss=0.1543, simple_loss=0.2426, pruned_loss=0.033, over 12088.00 frames. ], tot_loss[loss=0.1652, simple_loss=0.2548, pruned_loss=0.0378, over 2374763.51 frames. ], batch size: 32, lr: 3.55e-03, grad_scale: 16.0 2023-05-17 01:50:02,543 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=269647.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 01:50:28,526 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.817e+02 2.539e+02 3.181e+02 3.899e+02 5.844e+02, threshold=6.361e+02, percent-clipped=0.0 2023-05-17 01:50:30,911 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.3781, 4.9447, 5.3377, 4.6548, 4.9794, 4.7436, 5.3846, 4.9797], device='cuda:0'), covar=tensor([0.0248, 0.0379, 0.0265, 0.0266, 0.0439, 0.0346, 0.0250, 0.0404], device='cuda:0'), in_proj_covar=tensor([0.0271, 0.0274, 0.0302, 0.0274, 0.0274, 0.0273, 0.0250, 0.0222], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-17 01:50:32,428 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.3247, 3.2033, 2.9736, 2.9040, 2.6713, 2.5872, 3.1648, 2.0021], device='cuda:0'), covar=tensor([0.0423, 0.0173, 0.0218, 0.0227, 0.0397, 0.0317, 0.0159, 0.0557], device='cuda:0'), in_proj_covar=tensor([0.0200, 0.0168, 0.0172, 0.0195, 0.0207, 0.0205, 0.0177, 0.0210], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-17 01:50:35,021 INFO [finetune.py:992] (0/2) Epoch 14, batch 10650, loss[loss=0.1383, simple_loss=0.2241, pruned_loss=0.0263, over 12140.00 frames. ], tot_loss[loss=0.1654, simple_loss=0.2547, pruned_loss=0.03806, over 2373827.17 frames. ], batch size: 30, lr: 3.55e-03, grad_scale: 16.0 2023-05-17 01:50:37,309 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=269695.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 01:50:46,460 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=269708.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 01:50:48,070 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.5703, 3.7199, 3.2350, 3.2207, 2.9280, 2.8581, 3.7415, 2.3822], device='cuda:0'), covar=tensor([0.0430, 0.0166, 0.0244, 0.0242, 0.0408, 0.0378, 0.0143, 0.0521], device='cuda:0'), in_proj_covar=tensor([0.0200, 0.0169, 0.0172, 0.0195, 0.0207, 0.0205, 0.0177, 0.0210], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-17 01:51:11,123 INFO [finetune.py:992] (0/2) Epoch 14, batch 10700, loss[loss=0.2454, simple_loss=0.3251, pruned_loss=0.08289, over 8157.00 frames. ], tot_loss[loss=0.1655, simple_loss=0.255, pruned_loss=0.03797, over 2374388.86 frames. ], batch size: 97, lr: 3.55e-03, grad_scale: 16.0 2023-05-17 01:51:33,344 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=269773.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 01:51:40,222 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.943e+02 2.824e+02 3.372e+02 3.895e+02 7.438e+02, threshold=6.744e+02, percent-clipped=2.0 2023-05-17 01:51:46,545 INFO [finetune.py:992] (0/2) Epoch 14, batch 10750, loss[loss=0.1643, simple_loss=0.2471, pruned_loss=0.04076, over 12177.00 frames. ], tot_loss[loss=0.1655, simple_loss=0.255, pruned_loss=0.03797, over 2367483.04 frames. ], batch size: 31, lr: 3.55e-03, grad_scale: 16.0 2023-05-17 01:51:50,132 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=269797.0, num_to_drop=1, layers_to_drop={0} 2023-05-17 01:52:17,131 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=269834.0, num_to_drop=1, layers_to_drop={2} 2023-05-17 01:52:18,469 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=269836.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 01:52:22,660 INFO [finetune.py:992] (0/2) Epoch 14, batch 10800, loss[loss=0.146, simple_loss=0.2349, pruned_loss=0.02854, over 12137.00 frames. ], tot_loss[loss=0.1657, simple_loss=0.2551, pruned_loss=0.03812, over 2373175.35 frames. ], batch size: 30, lr: 3.55e-03, grad_scale: 16.0 2023-05-17 01:52:25,109 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.7631, 2.9363, 4.7390, 4.8556, 2.7789, 2.6619, 2.9862, 2.2098], device='cuda:0'), covar=tensor([0.1678, 0.2956, 0.0447, 0.0409, 0.1459, 0.2518, 0.2869, 0.4022], device='cuda:0'), in_proj_covar=tensor([0.0303, 0.0394, 0.0280, 0.0307, 0.0277, 0.0315, 0.0395, 0.0379], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-17 01:52:52,310 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.039e+02 2.900e+02 3.363e+02 3.885e+02 6.041e+02, threshold=6.726e+02, percent-clipped=0.0 2023-05-17 01:52:58,613 INFO [finetune.py:992] (0/2) Epoch 14, batch 10850, loss[loss=0.1551, simple_loss=0.2389, pruned_loss=0.03565, over 12137.00 frames. ], tot_loss[loss=0.1653, simple_loss=0.2547, pruned_loss=0.03799, over 2372978.25 frames. ], batch size: 30, lr: 3.55e-03, grad_scale: 16.0 2023-05-17 01:53:02,307 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=269897.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 01:53:02,366 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.2186, 2.7066, 3.7349, 3.1482, 3.5089, 3.3208, 2.7335, 3.5910], device='cuda:0'), covar=tensor([0.0142, 0.0302, 0.0151, 0.0237, 0.0172, 0.0177, 0.0309, 0.0143], device='cuda:0'), in_proj_covar=tensor([0.0185, 0.0208, 0.0196, 0.0192, 0.0223, 0.0172, 0.0202, 0.0194], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-17 01:53:02,371 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=269897.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 01:53:34,546 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=3.10 vs. limit=5.0 2023-05-17 01:53:34,835 INFO [finetune.py:992] (0/2) Epoch 14, batch 10900, loss[loss=0.1755, simple_loss=0.2715, pruned_loss=0.03976, over 12153.00 frames. ], tot_loss[loss=0.1664, simple_loss=0.2556, pruned_loss=0.0386, over 2370775.08 frames. ], batch size: 34, lr: 3.55e-03, grad_scale: 16.0 2023-05-17 01:53:37,029 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=269945.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 01:54:02,067 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.0957, 3.5222, 3.6819, 4.1672, 2.9311, 3.6825, 2.5375, 3.5063], device='cuda:0'), covar=tensor([0.1844, 0.0928, 0.0898, 0.0582, 0.1158, 0.0704, 0.1954, 0.1139], device='cuda:0'), in_proj_covar=tensor([0.0228, 0.0265, 0.0296, 0.0357, 0.0239, 0.0242, 0.0259, 0.0373], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-17 01:54:04,643 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.781e+02 2.784e+02 3.347e+02 3.934e+02 6.582e+02, threshold=6.694e+02, percent-clipped=0.0 2023-05-17 01:54:11,050 INFO [finetune.py:992] (0/2) Epoch 14, batch 10950, loss[loss=0.1929, simple_loss=0.2783, pruned_loss=0.05375, over 12150.00 frames. ], tot_loss[loss=0.1675, simple_loss=0.2567, pruned_loss=0.03919, over 2370769.08 frames. ], batch size: 34, lr: 3.55e-03, grad_scale: 16.0 2023-05-17 01:54:16,921 INFO [checkpoint.py:75] (0/2) Saving checkpoint to pruned_transducer_stateless7/exp_giga_finetune/checkpoint-170000.pt 2023-05-17 01:54:25,628 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=270008.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 01:54:43,537 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.2404, 6.0083, 5.7296, 5.4982, 6.1171, 5.3998, 5.5830, 5.5458], device='cuda:0'), covar=tensor([0.1383, 0.0932, 0.0819, 0.2230, 0.0958, 0.2243, 0.1861, 0.1196], device='cuda:0'), in_proj_covar=tensor([0.0368, 0.0516, 0.0413, 0.0471, 0.0485, 0.0450, 0.0415, 0.0403], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-17 01:54:49,717 INFO [finetune.py:992] (0/2) Epoch 14, batch 11000, loss[loss=0.1519, simple_loss=0.2353, pruned_loss=0.03423, over 12342.00 frames. ], tot_loss[loss=0.169, simple_loss=0.2583, pruned_loss=0.0398, over 2357973.79 frames. ], batch size: 30, lr: 3.55e-03, grad_scale: 16.0 2023-05-17 01:54:59,733 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=270056.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 01:55:19,151 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.924e+02 2.921e+02 3.787e+02 4.330e+02 1.308e+03, threshold=7.573e+02, percent-clipped=3.0 2023-05-17 01:55:25,267 INFO [finetune.py:992] (0/2) Epoch 14, batch 11050, loss[loss=0.2295, simple_loss=0.3264, pruned_loss=0.06632, over 10227.00 frames. ], tot_loss[loss=0.174, simple_loss=0.2627, pruned_loss=0.04265, over 2294947.03 frames. ], batch size: 68, lr: 3.54e-03, grad_scale: 16.0 2023-05-17 01:55:28,911 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=270097.0, num_to_drop=1, layers_to_drop={1} 2023-05-17 01:55:37,312 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.8462, 3.2672, 2.5023, 2.2672, 2.8733, 2.3527, 3.0956, 2.6564], device='cuda:0'), covar=tensor([0.0674, 0.0714, 0.0963, 0.1440, 0.0301, 0.1176, 0.0571, 0.0796], device='cuda:0'), in_proj_covar=tensor([0.0194, 0.0263, 0.0181, 0.0204, 0.0147, 0.0186, 0.0203, 0.0178], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-17 01:55:51,965 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=270129.0, num_to_drop=1, layers_to_drop={1} 2023-05-17 01:56:00,757 INFO [finetune.py:992] (0/2) Epoch 14, batch 11100, loss[loss=0.1389, simple_loss=0.2241, pruned_loss=0.02684, over 12347.00 frames. ], tot_loss[loss=0.1778, simple_loss=0.2661, pruned_loss=0.04474, over 2261020.29 frames. ], batch size: 31, lr: 3.54e-03, grad_scale: 16.0 2023-05-17 01:56:02,975 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=270145.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 01:56:29,927 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.965e+02 3.413e+02 4.097e+02 4.880e+02 1.017e+03, threshold=8.194e+02, percent-clipped=2.0 2023-05-17 01:56:36,313 INFO [finetune.py:992] (0/2) Epoch 14, batch 11150, loss[loss=0.1651, simple_loss=0.2549, pruned_loss=0.03765, over 12047.00 frames. ], tot_loss[loss=0.1834, simple_loss=0.2712, pruned_loss=0.04783, over 2215551.57 frames. ], batch size: 40, lr: 3.54e-03, grad_scale: 16.0 2023-05-17 01:56:36,412 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=270192.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 01:57:12,069 INFO [finetune.py:992] (0/2) Epoch 14, batch 11200, loss[loss=0.2575, simple_loss=0.3401, pruned_loss=0.08748, over 10152.00 frames. ], tot_loss[loss=0.1898, simple_loss=0.2772, pruned_loss=0.05122, over 2165625.74 frames. ], batch size: 68, lr: 3.54e-03, grad_scale: 16.0 2023-05-17 01:57:32,682 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.6601, 2.8122, 3.9440, 4.0321, 2.9608, 2.7032, 2.8139, 2.1472], device='cuda:0'), covar=tensor([0.1561, 0.2439, 0.0548, 0.0532, 0.1182, 0.2326, 0.2637, 0.4118], device='cuda:0'), in_proj_covar=tensor([0.0299, 0.0388, 0.0276, 0.0302, 0.0272, 0.0311, 0.0390, 0.0375], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-17 01:57:40,385 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=270280.0, num_to_drop=1, layers_to_drop={0} 2023-05-17 01:57:42,323 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.206e+02 3.467e+02 4.075e+02 5.309e+02 8.859e+02, threshold=8.151e+02, percent-clipped=3.0 2023-05-17 01:57:48,253 INFO [finetune.py:992] (0/2) Epoch 14, batch 11250, loss[loss=0.2927, simple_loss=0.3603, pruned_loss=0.1126, over 6850.00 frames. ], tot_loss[loss=0.1963, simple_loss=0.2826, pruned_loss=0.05496, over 2107002.22 frames. ], batch size: 98, lr: 3.54e-03, grad_scale: 16.0 2023-05-17 01:58:23,027 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=270341.0, num_to_drop=1, layers_to_drop={1} 2023-05-17 01:58:23,469 INFO [finetune.py:992] (0/2) Epoch 14, batch 11300, loss[loss=0.1893, simple_loss=0.2827, pruned_loss=0.04799, over 12151.00 frames. ], tot_loss[loss=0.2026, simple_loss=0.2881, pruned_loss=0.05857, over 2041871.81 frames. ], batch size: 36, lr: 3.54e-03, grad_scale: 16.0 2023-05-17 01:58:41,031 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 2023-05-17 01:58:52,085 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.411e+02 3.330e+02 3.992e+02 4.745e+02 8.371e+02, threshold=7.984e+02, percent-clipped=1.0 2023-05-17 01:58:57,996 INFO [finetune.py:992] (0/2) Epoch 14, batch 11350, loss[loss=0.2572, simple_loss=0.3335, pruned_loss=0.09042, over 10364.00 frames. ], tot_loss[loss=0.2072, simple_loss=0.2923, pruned_loss=0.06103, over 2012214.64 frames. ], batch size: 68, lr: 3.54e-03, grad_scale: 16.0 2023-05-17 01:59:01,635 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=270396.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 01:59:24,448 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=270429.0, num_to_drop=1, layers_to_drop={1} 2023-05-17 01:59:31,342 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=270439.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 01:59:33,158 INFO [finetune.py:992] (0/2) Epoch 14, batch 11400, loss[loss=0.2492, simple_loss=0.3188, pruned_loss=0.08974, over 6655.00 frames. ], tot_loss[loss=0.2107, simple_loss=0.2954, pruned_loss=0.06301, over 1979860.41 frames. ], batch size: 98, lr: 3.54e-03, grad_scale: 16.0 2023-05-17 01:59:43,698 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=270457.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 01:59:58,598 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=270477.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 02:00:02,121 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.8487, 2.5990, 3.5077, 3.5620, 2.9200, 2.7113, 2.6820, 2.3760], device='cuda:0'), covar=tensor([0.1152, 0.2282, 0.0528, 0.0487, 0.0883, 0.1923, 0.2483, 0.3564], device='cuda:0'), in_proj_covar=tensor([0.0297, 0.0384, 0.0273, 0.0298, 0.0270, 0.0308, 0.0388, 0.0372], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-17 02:00:02,442 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.325e+02 3.507e+02 4.046e+02 4.865e+02 1.035e+03, threshold=8.092e+02, percent-clipped=3.0 2023-05-17 02:00:08,530 INFO [finetune.py:992] (0/2) Epoch 14, batch 11450, loss[loss=0.1929, simple_loss=0.2869, pruned_loss=0.04945, over 10291.00 frames. ], tot_loss[loss=0.2153, simple_loss=0.2991, pruned_loss=0.06575, over 1927248.20 frames. ], batch size: 68, lr: 3.54e-03, grad_scale: 16.0 2023-05-17 02:00:08,658 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=270492.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 02:00:14,167 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=270500.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 02:00:42,142 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=270540.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 02:00:43,345 INFO [finetune.py:992] (0/2) Epoch 14, batch 11500, loss[loss=0.206, simple_loss=0.2983, pruned_loss=0.05686, over 11765.00 frames. ], tot_loss[loss=0.2202, simple_loss=0.3026, pruned_loss=0.06894, over 1865765.63 frames. ], batch size: 44, lr: 3.54e-03, grad_scale: 16.0 2023-05-17 02:00:46,329 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.2088, 3.4435, 3.1732, 3.5494, 3.3246, 2.5460, 3.1802, 2.7674], device='cuda:0'), covar=tensor([0.0935, 0.1004, 0.1414, 0.0736, 0.1182, 0.1718, 0.1277, 0.2911], device='cuda:0'), in_proj_covar=tensor([0.0293, 0.0363, 0.0343, 0.0305, 0.0353, 0.0263, 0.0334, 0.0353], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-17 02:01:10,843 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.541e+02 3.383e+02 4.011e+02 4.670e+02 1.056e+03, threshold=8.021e+02, percent-clipped=3.0 2023-05-17 02:01:17,531 INFO [finetune.py:992] (0/2) Epoch 14, batch 11550, loss[loss=0.2762, simple_loss=0.3405, pruned_loss=0.106, over 6705.00 frames. ], tot_loss[loss=0.2225, simple_loss=0.3042, pruned_loss=0.07041, over 1845512.00 frames. ], batch size: 97, lr: 3.54e-03, grad_scale: 16.0 2023-05-17 02:01:32,794 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=270613.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 02:01:47,810 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=270636.0, num_to_drop=1, layers_to_drop={3} 2023-05-17 02:01:48,521 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.4739, 5.3228, 5.2714, 5.2772, 4.8895, 5.3474, 5.4424, 5.4120], device='cuda:0'), covar=tensor([0.0135, 0.0112, 0.0126, 0.0271, 0.0622, 0.0228, 0.0112, 0.0153], device='cuda:0'), in_proj_covar=tensor([0.0187, 0.0187, 0.0181, 0.0234, 0.0229, 0.0206, 0.0169, 0.0219], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:0') 2023-05-17 02:01:51,857 INFO [finetune.py:992] (0/2) Epoch 14, batch 11600, loss[loss=0.2247, simple_loss=0.2995, pruned_loss=0.07499, over 11362.00 frames. ], tot_loss[loss=0.2253, simple_loss=0.3062, pruned_loss=0.07222, over 1824037.98 frames. ], batch size: 55, lr: 3.54e-03, grad_scale: 16.0 2023-05-17 02:02:13,928 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.8363, 3.8081, 3.8035, 3.8807, 3.6915, 3.7226, 3.6431, 3.7797], device='cuda:0'), covar=tensor([0.1105, 0.0724, 0.1470, 0.0703, 0.1625, 0.1186, 0.0591, 0.1064], device='cuda:0'), in_proj_covar=tensor([0.0516, 0.0668, 0.0588, 0.0607, 0.0801, 0.0709, 0.0541, 0.0465], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0004, 0.0004, 0.0005, 0.0004, 0.0003, 0.0003], device='cuda:0') 2023-05-17 02:02:15,492 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=270674.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 02:02:21,159 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.6063, 4.5222, 4.5966, 4.6089, 4.2523, 4.2402, 4.2409, 4.4324], device='cuda:0'), covar=tensor([0.1013, 0.0830, 0.1250, 0.0754, 0.2539, 0.1765, 0.0678, 0.1442], device='cuda:0'), in_proj_covar=tensor([0.0517, 0.0669, 0.0589, 0.0608, 0.0803, 0.0710, 0.0542, 0.0467], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0005, 0.0004, 0.0003, 0.0003], device='cuda:0') 2023-05-17 02:02:21,595 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.195e+02 3.446e+02 3.907e+02 4.532e+02 7.259e+02, threshold=7.814e+02, percent-clipped=0.0 2023-05-17 02:02:28,184 INFO [finetune.py:992] (0/2) Epoch 14, batch 11650, loss[loss=0.2016, simple_loss=0.2925, pruned_loss=0.05535, over 10157.00 frames. ], tot_loss[loss=0.226, simple_loss=0.3061, pruned_loss=0.07302, over 1807701.88 frames. ], batch size: 68, lr: 3.54e-03, grad_scale: 16.0 2023-05-17 02:02:35,752 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.8936, 2.1809, 2.6226, 2.9147, 2.2716, 3.0414, 2.9740, 3.0573], device='cuda:0'), covar=tensor([0.0190, 0.1110, 0.0429, 0.0186, 0.1026, 0.0311, 0.0282, 0.0157], device='cuda:0'), in_proj_covar=tensor([0.0122, 0.0201, 0.0179, 0.0120, 0.0187, 0.0176, 0.0171, 0.0121], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-17 02:02:53,051 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=270726.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 02:03:03,567 INFO [finetune.py:992] (0/2) Epoch 14, batch 11700, loss[loss=0.2147, simple_loss=0.3001, pruned_loss=0.06464, over 10344.00 frames. ], tot_loss[loss=0.2267, simple_loss=0.3059, pruned_loss=0.07368, over 1786445.76 frames. ], batch size: 68, lr: 3.54e-03, grad_scale: 16.0 2023-05-17 02:03:10,419 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=270752.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 02:03:32,508 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.544e+02 3.420e+02 4.029e+02 4.880e+02 9.418e+02, threshold=8.057e+02, percent-clipped=2.0 2023-05-17 02:03:35,444 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=270787.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 02:03:38,562 INFO [finetune.py:992] (0/2) Epoch 14, batch 11750, loss[loss=0.2251, simple_loss=0.2952, pruned_loss=0.07753, over 6278.00 frames. ], tot_loss[loss=0.2297, simple_loss=0.3078, pruned_loss=0.07575, over 1733522.45 frames. ], batch size: 98, lr: 3.54e-03, grad_scale: 16.0 2023-05-17 02:03:40,739 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=270795.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 02:04:14,132 INFO [finetune.py:992] (0/2) Epoch 14, batch 11800, loss[loss=0.2005, simple_loss=0.2913, pruned_loss=0.05487, over 10164.00 frames. ], tot_loss[loss=0.2321, simple_loss=0.3098, pruned_loss=0.0772, over 1723416.50 frames. ], batch size: 68, lr: 3.54e-03, grad_scale: 16.0 2023-05-17 02:04:19,452 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.51 vs. limit=2.0 2023-05-17 02:04:42,101 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.651e+02 3.576e+02 4.134e+02 5.028e+02 8.984e+02, threshold=8.267e+02, percent-clipped=2.0 2023-05-17 02:04:48,198 INFO [finetune.py:992] (0/2) Epoch 14, batch 11850, loss[loss=0.248, simple_loss=0.3205, pruned_loss=0.08771, over 7067.00 frames. ], tot_loss[loss=0.2332, simple_loss=0.3114, pruned_loss=0.07751, over 1710881.14 frames. ], batch size: 98, lr: 3.54e-03, grad_scale: 16.0 2023-05-17 02:04:56,631 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([1.9434, 2.2579, 2.1533, 2.1156, 1.8989, 1.9606, 2.0529, 1.6124], device='cuda:0'), covar=tensor([0.0349, 0.0187, 0.0211, 0.0206, 0.0340, 0.0272, 0.0205, 0.0472], device='cuda:0'), in_proj_covar=tensor([0.0191, 0.0161, 0.0163, 0.0187, 0.0198, 0.0197, 0.0170, 0.0202], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-17 02:05:04,693 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.5138, 4.4644, 4.4045, 4.0958, 4.0619, 4.4859, 4.2773, 4.0715], device='cuda:0'), covar=tensor([0.0918, 0.0954, 0.0662, 0.1329, 0.2606, 0.0837, 0.1445, 0.1105], device='cuda:0'), in_proj_covar=tensor([0.0594, 0.0533, 0.0489, 0.0603, 0.0398, 0.0690, 0.0732, 0.0544], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002], device='cuda:0') 2023-05-17 02:05:06,766 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=270917.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 02:05:19,518 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=270936.0, num_to_drop=1, layers_to_drop={0} 2023-05-17 02:05:23,454 INFO [finetune.py:992] (0/2) Epoch 14, batch 11900, loss[loss=0.1981, simple_loss=0.2867, pruned_loss=0.05478, over 12192.00 frames. ], tot_loss[loss=0.2313, simple_loss=0.3102, pruned_loss=0.07613, over 1705328.53 frames. ], batch size: 35, lr: 3.54e-03, grad_scale: 16.0 2023-05-17 02:05:24,360 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.9094, 3.7839, 3.9119, 3.6541, 3.7962, 3.6556, 3.8536, 3.5814], device='cuda:0'), covar=tensor([0.0556, 0.0529, 0.0516, 0.0345, 0.0467, 0.0406, 0.0490, 0.1559], device='cuda:0'), in_proj_covar=tensor([0.0252, 0.0257, 0.0279, 0.0256, 0.0256, 0.0253, 0.0232, 0.0204], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-17 02:05:37,855 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.8848, 3.8626, 3.8770, 3.9291, 3.7523, 3.7744, 3.7126, 3.8202], device='cuda:0'), covar=tensor([0.1136, 0.0771, 0.1417, 0.0745, 0.1610, 0.1201, 0.0524, 0.1089], device='cuda:0'), in_proj_covar=tensor([0.0507, 0.0656, 0.0580, 0.0597, 0.0783, 0.0697, 0.0532, 0.0457], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0003], device='cuda:0') 2023-05-17 02:05:43,115 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=270969.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 02:05:49,332 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=270978.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 02:05:52,000 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.5703, 4.5101, 4.4645, 4.1042, 4.1203, 4.5438, 4.3156, 4.1676], device='cuda:0'), covar=tensor([0.0788, 0.0999, 0.0642, 0.1376, 0.2384, 0.0733, 0.1285, 0.0965], device='cuda:0'), in_proj_covar=tensor([0.0591, 0.0530, 0.0486, 0.0598, 0.0396, 0.0684, 0.0727, 0.0541], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002], device='cuda:0') 2023-05-17 02:05:52,500 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.421e+02 3.376e+02 3.838e+02 4.530e+02 1.360e+03, threshold=7.675e+02, percent-clipped=2.0 2023-05-17 02:05:53,292 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=270984.0, num_to_drop=1, layers_to_drop={0} 2023-05-17 02:05:58,458 INFO [finetune.py:992] (0/2) Epoch 14, batch 11950, loss[loss=0.1865, simple_loss=0.2793, pruned_loss=0.0468, over 10085.00 frames. ], tot_loss[loss=0.2276, simple_loss=0.3079, pruned_loss=0.07367, over 1705361.12 frames. ], batch size: 69, lr: 3.54e-03, grad_scale: 16.0 2023-05-17 02:06:15,216 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([1.9532, 2.2073, 2.0927, 2.1487, 1.8910, 1.9341, 2.0837, 1.6787], device='cuda:0'), covar=tensor([0.0349, 0.0229, 0.0249, 0.0222, 0.0384, 0.0293, 0.0189, 0.0465], device='cuda:0'), in_proj_covar=tensor([0.0190, 0.0160, 0.0162, 0.0185, 0.0197, 0.0195, 0.0168, 0.0201], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-17 02:06:29,247 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=4.96 vs. limit=5.0 2023-05-17 02:06:34,148 INFO [finetune.py:992] (0/2) Epoch 14, batch 12000, loss[loss=0.1723, simple_loss=0.268, pruned_loss=0.03826, over 10065.00 frames. ], tot_loss[loss=0.2223, simple_loss=0.3036, pruned_loss=0.07051, over 1696037.61 frames. ], batch size: 68, lr: 3.54e-03, grad_scale: 16.0 2023-05-17 02:06:34,153 INFO [finetune.py:1017] (0/2) Computing validation loss 2023-05-17 02:06:52,782 INFO [finetune.py:1026] (0/2) Epoch 14, validation: loss=0.2855, simple_loss=0.361, pruned_loss=0.105, over 1020973.00 frames. 2023-05-17 02:06:52,783 INFO [finetune.py:1027] (0/2) Maximum memory allocated so far is 12827MB 2023-05-17 02:06:59,494 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=271052.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 02:07:08,351 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=2.02 vs. limit=2.0 2023-05-17 02:07:09,460 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.5602, 4.5206, 4.4353, 4.0403, 4.1073, 4.5270, 4.3146, 4.0907], device='cuda:0'), covar=tensor([0.0758, 0.0860, 0.0616, 0.1413, 0.2296, 0.0739, 0.1308, 0.1080], device='cuda:0'), in_proj_covar=tensor([0.0589, 0.0528, 0.0485, 0.0596, 0.0395, 0.0682, 0.0723, 0.0539], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002], device='cuda:0') 2023-05-17 02:07:20,940 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=271082.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 02:07:21,521 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.959e+02 2.853e+02 3.352e+02 4.102e+02 7.378e+02, threshold=6.704e+02, percent-clipped=0.0 2023-05-17 02:07:25,071 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.8076, 3.5806, 3.5820, 3.7157, 3.7320, 3.8323, 3.7231, 2.6579], device='cuda:0'), covar=tensor([0.0084, 0.0098, 0.0138, 0.0077, 0.0063, 0.0096, 0.0085, 0.0725], device='cuda:0'), in_proj_covar=tensor([0.0066, 0.0077, 0.0081, 0.0071, 0.0059, 0.0089, 0.0079, 0.0095], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-17 02:07:27,504 INFO [finetune.py:992] (0/2) Epoch 14, batch 12050, loss[loss=0.2214, simple_loss=0.3067, pruned_loss=0.06808, over 12275.00 frames. ], tot_loss[loss=0.2174, simple_loss=0.2995, pruned_loss=0.06766, over 1690063.66 frames. ], batch size: 37, lr: 3.54e-03, grad_scale: 16.0 2023-05-17 02:07:29,633 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=271095.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 02:07:32,750 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=271100.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 02:07:48,402 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.90 vs. limit=2.0 2023-05-17 02:08:00,810 INFO [finetune.py:992] (0/2) Epoch 14, batch 12100, loss[loss=0.209, simple_loss=0.2882, pruned_loss=0.0649, over 6596.00 frames. ], tot_loss[loss=0.216, simple_loss=0.2984, pruned_loss=0.06675, over 1681610.44 frames. ], batch size: 100, lr: 3.54e-03, grad_scale: 32.0 2023-05-17 02:08:01,559 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=271143.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 02:08:27,435 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.397e+02 3.236e+02 3.719e+02 4.278e+02 7.447e+02, threshold=7.437e+02, percent-clipped=1.0 2023-05-17 02:08:33,299 INFO [finetune.py:992] (0/2) Epoch 14, batch 12150, loss[loss=0.1991, simple_loss=0.2964, pruned_loss=0.0509, over 11455.00 frames. ], tot_loss[loss=0.2158, simple_loss=0.2985, pruned_loss=0.06654, over 1697479.79 frames. ], batch size: 48, lr: 3.54e-03, grad_scale: 32.0 2023-05-17 02:08:35,922 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.4781, 4.1409, 4.2087, 4.2984, 4.3158, 4.4373, 4.3817, 2.4780], device='cuda:0'), covar=tensor([0.0080, 0.0087, 0.0111, 0.0075, 0.0052, 0.0099, 0.0078, 0.0951], device='cuda:0'), in_proj_covar=tensor([0.0066, 0.0077, 0.0081, 0.0071, 0.0059, 0.0090, 0.0079, 0.0096], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-17 02:08:50,317 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.8751, 3.7188, 3.8657, 3.6155, 3.7804, 3.6538, 3.8514, 3.4775], device='cuda:0'), covar=tensor([0.0426, 0.0404, 0.0380, 0.0298, 0.0381, 0.0334, 0.0341, 0.1531], device='cuda:0'), in_proj_covar=tensor([0.0247, 0.0251, 0.0273, 0.0251, 0.0251, 0.0249, 0.0228, 0.0200], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-17 02:09:05,323 INFO [finetune.py:992] (0/2) Epoch 14, batch 12200, loss[loss=0.2059, simple_loss=0.2866, pruned_loss=0.06261, over 6701.00 frames. ], tot_loss[loss=0.2174, simple_loss=0.2993, pruned_loss=0.06769, over 1663688.78 frames. ], batch size: 97, lr: 3.54e-03, grad_scale: 32.0 2023-05-17 02:09:05,567 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.3979, 2.9838, 3.6491, 2.3113, 2.6814, 3.0536, 2.9820, 3.1242], device='cuda:0'), covar=tensor([0.0512, 0.1080, 0.0344, 0.1441, 0.1647, 0.1375, 0.1112, 0.1143], device='cuda:0'), in_proj_covar=tensor([0.0225, 0.0228, 0.0237, 0.0178, 0.0227, 0.0280, 0.0215, 0.0251], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:0') 2023-05-17 02:09:17,059 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=271261.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 02:09:17,322 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.45 vs. limit=2.0 2023-05-17 02:09:21,820 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=271269.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 02:09:24,266 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=271273.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 02:09:27,697 INFO [checkpoint.py:75] (0/2) Saving checkpoint to pruned_transducer_stateless7/exp_giga_finetune/epoch-14.pt 2023-05-17 02:09:49,992 INFO [finetune.py:992] (0/2) Epoch 15, batch 0, loss[loss=0.206, simple_loss=0.2941, pruned_loss=0.05894, over 11572.00 frames. ], tot_loss[loss=0.206, simple_loss=0.2941, pruned_loss=0.05894, over 11572.00 frames. ], batch size: 48, lr: 3.53e-03, grad_scale: 32.0 2023-05-17 02:09:49,992 INFO [finetune.py:1017] (0/2) Computing validation loss 2023-05-17 02:10:00,820 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.4992, 2.2533, 3.2895, 3.6191, 3.4451, 3.5940, 3.4646, 2.2825], device='cuda:0'), covar=tensor([0.0070, 0.0428, 0.0162, 0.0062, 0.0105, 0.0092, 0.0107, 0.0522], device='cuda:0'), in_proj_covar=tensor([0.0089, 0.0123, 0.0103, 0.0077, 0.0101, 0.0114, 0.0096, 0.0137], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-17 02:10:07,275 INFO [finetune.py:1026] (0/2) Epoch 15, validation: loss=0.287, simple_loss=0.3614, pruned_loss=0.1063, over 1020973.00 frames. 2023-05-17 02:10:07,276 INFO [finetune.py:1027] (0/2) Maximum memory allocated so far is 12827MB 2023-05-17 02:10:11,512 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.236e+02 3.235e+02 3.835e+02 4.619e+02 1.041e+03, threshold=7.669e+02, percent-clipped=4.0 2023-05-17 02:10:23,352 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 2023-05-17 02:10:35,721 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=271317.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 02:10:39,534 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=271322.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 02:10:42,950 INFO [finetune.py:992] (0/2) Epoch 15, batch 50, loss[loss=0.1556, simple_loss=0.2447, pruned_loss=0.03326, over 12138.00 frames. ], tot_loss[loss=0.1761, simple_loss=0.2658, pruned_loss=0.04322, over 533587.87 frames. ], batch size: 30, lr: 3.53e-03, grad_scale: 32.0 2023-05-17 02:11:12,742 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.92 vs. limit=2.0 2023-05-17 02:11:16,723 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=271375.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 02:11:18,000 INFO [finetune.py:992] (0/2) Epoch 15, batch 100, loss[loss=0.1695, simple_loss=0.2567, pruned_loss=0.04111, over 12163.00 frames. ], tot_loss[loss=0.1756, simple_loss=0.2658, pruned_loss=0.04273, over 937634.47 frames. ], batch size: 29, lr: 3.53e-03, grad_scale: 32.0 2023-05-17 02:11:22,271 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=271382.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 02:11:22,823 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.736e+02 2.781e+02 3.225e+02 3.711e+02 7.745e+02, threshold=6.450e+02, percent-clipped=1.0 2023-05-17 02:11:45,484 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.6424, 2.7942, 4.6306, 4.8251, 2.7258, 2.5150, 2.9245, 2.0166], device='cuda:0'), covar=tensor([0.1896, 0.3643, 0.0518, 0.0455, 0.1560, 0.3051, 0.3265, 0.5155], device='cuda:0'), in_proj_covar=tensor([0.0296, 0.0385, 0.0272, 0.0297, 0.0269, 0.0310, 0.0391, 0.0374], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-17 02:11:54,575 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=271426.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 02:11:55,132 INFO [finetune.py:992] (0/2) Epoch 15, batch 150, loss[loss=0.1445, simple_loss=0.2379, pruned_loss=0.0256, over 11859.00 frames. ], tot_loss[loss=0.1738, simple_loss=0.2642, pruned_loss=0.04167, over 1258471.84 frames. ], batch size: 26, lr: 3.53e-03, grad_scale: 32.0 2023-05-17 02:11:57,425 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=271430.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 02:12:01,750 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=271436.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 02:12:24,909 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=4.93 vs. limit=5.0 2023-05-17 02:12:30,904 INFO [finetune.py:992] (0/2) Epoch 15, batch 200, loss[loss=0.1827, simple_loss=0.2741, pruned_loss=0.04571, over 12121.00 frames. ], tot_loss[loss=0.1715, simple_loss=0.2621, pruned_loss=0.0404, over 1504100.63 frames. ], batch size: 39, lr: 3.53e-03, grad_scale: 32.0 2023-05-17 02:12:35,271 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.803e+02 2.755e+02 3.140e+02 3.726e+02 5.198e+02, threshold=6.280e+02, percent-clipped=0.0 2023-05-17 02:12:38,363 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=271487.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 02:13:06,815 INFO [finetune.py:992] (0/2) Epoch 15, batch 250, loss[loss=0.1464, simple_loss=0.2337, pruned_loss=0.02957, over 12160.00 frames. ], tot_loss[loss=0.1703, simple_loss=0.2608, pruned_loss=0.03992, over 1698682.15 frames. ], batch size: 29, lr: 3.53e-03, grad_scale: 32.0 2023-05-17 02:13:39,728 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.0356, 2.2413, 3.6296, 3.0787, 3.4184, 3.0800, 2.4826, 3.5034], device='cuda:0'), covar=tensor([0.0186, 0.0500, 0.0152, 0.0278, 0.0150, 0.0243, 0.0444, 0.0150], device='cuda:0'), in_proj_covar=tensor([0.0178, 0.0203, 0.0186, 0.0184, 0.0213, 0.0165, 0.0196, 0.0184], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-17 02:13:41,081 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=271573.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 02:13:42,200 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2023-05-17 02:13:43,950 INFO [finetune.py:992] (0/2) Epoch 15, batch 300, loss[loss=0.1698, simple_loss=0.2684, pruned_loss=0.03559, over 12099.00 frames. ], tot_loss[loss=0.1688, simple_loss=0.2592, pruned_loss=0.03916, over 1855792.82 frames. ], batch size: 33, lr: 3.53e-03, grad_scale: 32.0 2023-05-17 02:13:48,292 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.910e+02 2.692e+02 3.140e+02 3.693e+02 6.015e+02, threshold=6.279e+02, percent-clipped=0.0 2023-05-17 02:14:12,956 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=271617.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 02:14:15,901 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=271621.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 02:14:20,122 INFO [finetune.py:992] (0/2) Epoch 15, batch 350, loss[loss=0.1829, simple_loss=0.279, pruned_loss=0.04342, over 12153.00 frames. ], tot_loss[loss=0.1686, simple_loss=0.2588, pruned_loss=0.03916, over 1972510.25 frames. ], batch size: 34, lr: 3.53e-03, grad_scale: 32.0 2023-05-17 02:14:42,808 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.9000, 3.6102, 5.2359, 2.8286, 2.9696, 3.8819, 3.5085, 3.8213], device='cuda:0'), covar=tensor([0.0381, 0.1094, 0.0284, 0.1167, 0.1908, 0.1686, 0.1169, 0.1320], device='cuda:0'), in_proj_covar=tensor([0.0234, 0.0236, 0.0247, 0.0184, 0.0235, 0.0291, 0.0222, 0.0261], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-17 02:14:46,337 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.1539, 2.6012, 3.7908, 3.1920, 3.5362, 3.2321, 2.6772, 3.6444], device='cuda:0'), covar=tensor([0.0166, 0.0389, 0.0146, 0.0256, 0.0153, 0.0201, 0.0406, 0.0130], device='cuda:0'), in_proj_covar=tensor([0.0178, 0.0202, 0.0186, 0.0184, 0.0213, 0.0165, 0.0196, 0.0184], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-17 02:14:55,446 INFO [finetune.py:992] (0/2) Epoch 15, batch 400, loss[loss=0.1748, simple_loss=0.2541, pruned_loss=0.04775, over 11778.00 frames. ], tot_loss[loss=0.1674, simple_loss=0.2572, pruned_loss=0.03884, over 2070222.90 frames. ], batch size: 26, lr: 3.53e-03, grad_scale: 32.0 2023-05-17 02:14:59,683 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.800e+02 2.614e+02 3.141e+02 3.533e+02 5.574e+02, threshold=6.283e+02, percent-clipped=0.0 2023-05-17 02:15:26,048 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.2188, 6.0817, 5.7231, 5.6268, 6.1482, 5.3862, 5.6490, 5.5625], device='cuda:0'), covar=tensor([0.1465, 0.0881, 0.1144, 0.1807, 0.0855, 0.2091, 0.1707, 0.1205], device='cuda:0'), in_proj_covar=tensor([0.0354, 0.0496, 0.0400, 0.0449, 0.0461, 0.0426, 0.0389, 0.0382], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-17 02:15:32,304 INFO [finetune.py:992] (0/2) Epoch 15, batch 450, loss[loss=0.1701, simple_loss=0.2531, pruned_loss=0.04355, over 12333.00 frames. ], tot_loss[loss=0.1679, simple_loss=0.2574, pruned_loss=0.03919, over 2133122.92 frames. ], batch size: 31, lr: 3.53e-03, grad_scale: 32.0 2023-05-17 02:15:35,307 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=271731.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 02:15:59,758 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.5336, 3.4992, 3.1664, 3.2130, 2.8294, 2.6668, 3.4899, 2.4869], device='cuda:0'), covar=tensor([0.0392, 0.0183, 0.0198, 0.0187, 0.0431, 0.0357, 0.0159, 0.0466], device='cuda:0'), in_proj_covar=tensor([0.0187, 0.0158, 0.0160, 0.0183, 0.0195, 0.0191, 0.0167, 0.0198], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-17 02:16:07,212 INFO [finetune.py:992] (0/2) Epoch 15, batch 500, loss[loss=0.1806, simple_loss=0.281, pruned_loss=0.04006, over 11577.00 frames. ], tot_loss[loss=0.1678, simple_loss=0.2573, pruned_loss=0.0391, over 2185644.75 frames. ], batch size: 48, lr: 3.53e-03, grad_scale: 32.0 2023-05-17 02:16:08,163 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.1713, 2.5088, 3.7553, 3.1287, 3.4784, 3.1879, 2.5602, 3.5802], device='cuda:0'), covar=tensor([0.0162, 0.0412, 0.0146, 0.0290, 0.0185, 0.0229, 0.0428, 0.0168], device='cuda:0'), in_proj_covar=tensor([0.0181, 0.0206, 0.0189, 0.0187, 0.0217, 0.0168, 0.0200, 0.0187], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-17 02:16:10,970 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=271782.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 02:16:11,641 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.669e+02 2.663e+02 3.335e+02 3.915e+02 7.207e+02, threshold=6.671e+02, percent-clipped=2.0 2023-05-17 02:16:31,262 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.9008, 5.8662, 5.6758, 5.0934, 5.0828, 5.7920, 5.3668, 5.1864], device='cuda:0'), covar=tensor([0.0863, 0.0979, 0.0725, 0.1749, 0.0761, 0.0761, 0.1564, 0.1168], device='cuda:0'), in_proj_covar=tensor([0.0618, 0.0551, 0.0507, 0.0627, 0.0411, 0.0712, 0.0764, 0.0562], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002], device='cuda:0') 2023-05-17 02:16:43,425 INFO [finetune.py:992] (0/2) Epoch 15, batch 550, loss[loss=0.1662, simple_loss=0.2644, pruned_loss=0.03397, over 12365.00 frames. ], tot_loss[loss=0.1672, simple_loss=0.2569, pruned_loss=0.03879, over 2232682.16 frames. ], batch size: 38, lr: 3.53e-03, grad_scale: 32.0 2023-05-17 02:16:48,022 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.36 vs. limit=2.0 2023-05-17 02:17:04,171 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=271854.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 02:17:11,840 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.37 vs. limit=2.0 2023-05-17 02:17:20,546 INFO [finetune.py:992] (0/2) Epoch 15, batch 600, loss[loss=0.155, simple_loss=0.2411, pruned_loss=0.03445, over 12187.00 frames. ], tot_loss[loss=0.1666, simple_loss=0.2566, pruned_loss=0.03826, over 2268314.18 frames. ], batch size: 31, lr: 3.53e-03, grad_scale: 32.0 2023-05-17 02:17:24,789 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.766e+02 2.643e+02 2.962e+02 3.585e+02 7.418e+02, threshold=5.924e+02, percent-clipped=1.0 2023-05-17 02:17:47,587 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=271915.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 02:17:48,979 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=271917.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 02:17:56,079 INFO [finetune.py:992] (0/2) Epoch 15, batch 650, loss[loss=0.1547, simple_loss=0.2423, pruned_loss=0.03348, over 12124.00 frames. ], tot_loss[loss=0.1666, simple_loss=0.2563, pruned_loss=0.03845, over 2281668.80 frames. ], batch size: 30, lr: 3.53e-03, grad_scale: 32.0 2023-05-17 02:18:01,213 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=271934.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 02:18:13,249 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.6765, 2.7281, 4.7463, 4.9464, 3.0134, 2.5981, 2.9336, 2.0038], device='cuda:0'), covar=tensor([0.1654, 0.3326, 0.0417, 0.0345, 0.1215, 0.2685, 0.2999, 0.4717], device='cuda:0'), in_proj_covar=tensor([0.0297, 0.0388, 0.0274, 0.0300, 0.0271, 0.0311, 0.0391, 0.0377], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-17 02:18:16,670 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.4729, 4.8232, 3.0877, 2.9455, 4.1721, 2.7546, 4.0097, 3.3616], device='cuda:0'), covar=tensor([0.0714, 0.0526, 0.1146, 0.1529, 0.0296, 0.1486, 0.0510, 0.0860], device='cuda:0'), in_proj_covar=tensor([0.0191, 0.0255, 0.0180, 0.0203, 0.0142, 0.0185, 0.0198, 0.0177], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-17 02:18:22,925 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=271965.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 02:18:31,320 INFO [finetune.py:992] (0/2) Epoch 15, batch 700, loss[loss=0.1496, simple_loss=0.2433, pruned_loss=0.02795, over 12168.00 frames. ], tot_loss[loss=0.1667, simple_loss=0.2564, pruned_loss=0.03852, over 2302373.56 frames. ], batch size: 31, lr: 3.53e-03, grad_scale: 32.0 2023-05-17 02:18:36,823 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.814e+02 2.488e+02 2.987e+02 3.609e+02 6.450e+02, threshold=5.974e+02, percent-clipped=2.0 2023-05-17 02:18:45,577 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=271995.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 02:18:49,283 INFO [checkpoint.py:75] (0/2) Saving checkpoint to pruned_transducer_stateless7/exp_giga_finetune/checkpoint-172000.pt 2023-05-17 02:18:59,572 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.6255, 4.9085, 4.3384, 5.2623, 4.7088, 2.9874, 4.3608, 3.1118], device='cuda:0'), covar=tensor([0.0645, 0.0675, 0.1232, 0.0466, 0.1008, 0.1703, 0.1078, 0.3271], device='cuda:0'), in_proj_covar=tensor([0.0307, 0.0377, 0.0360, 0.0317, 0.0366, 0.0276, 0.0348, 0.0366], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-17 02:19:11,341 INFO [finetune.py:992] (0/2) Epoch 15, batch 750, loss[loss=0.1903, simple_loss=0.2762, pruned_loss=0.0522, over 11386.00 frames. ], tot_loss[loss=0.1662, simple_loss=0.2555, pruned_loss=0.03843, over 2319995.26 frames. ], batch size: 55, lr: 3.53e-03, grad_scale: 32.0 2023-05-17 02:19:14,369 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=272031.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 02:19:43,431 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=2.57 vs. limit=5.0 2023-05-17 02:19:47,240 INFO [finetune.py:992] (0/2) Epoch 15, batch 800, loss[loss=0.1876, simple_loss=0.2803, pruned_loss=0.04744, over 12149.00 frames. ], tot_loss[loss=0.1653, simple_loss=0.2546, pruned_loss=0.03796, over 2341641.01 frames. ], batch size: 36, lr: 3.53e-03, grad_scale: 32.0 2023-05-17 02:19:48,792 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=272079.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 02:19:50,957 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=272082.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 02:19:51,535 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.751e+02 2.501e+02 2.779e+02 3.240e+02 5.466e+02, threshold=5.558e+02, percent-clipped=0.0 2023-05-17 02:20:18,019 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.2368, 3.5900, 3.5878, 4.0407, 2.7583, 3.5398, 2.5295, 3.4740], device='cuda:0'), covar=tensor([0.1690, 0.0807, 0.0949, 0.0667, 0.1247, 0.0743, 0.1927, 0.1113], device='cuda:0'), in_proj_covar=tensor([0.0233, 0.0270, 0.0300, 0.0359, 0.0246, 0.0247, 0.0264, 0.0378], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-17 02:20:23,526 INFO [finetune.py:992] (0/2) Epoch 15, batch 850, loss[loss=0.1363, simple_loss=0.2215, pruned_loss=0.02554, over 11805.00 frames. ], tot_loss[loss=0.165, simple_loss=0.2547, pruned_loss=0.03766, over 2351121.95 frames. ], batch size: 26, lr: 3.53e-03, grad_scale: 32.0 2023-05-17 02:20:23,728 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=272127.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 02:20:25,654 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=272130.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 02:21:00,055 INFO [finetune.py:992] (0/2) Epoch 15, batch 900, loss[loss=0.1743, simple_loss=0.2632, pruned_loss=0.04269, over 12136.00 frames. ], tot_loss[loss=0.1648, simple_loss=0.2546, pruned_loss=0.03751, over 2358532.87 frames. ], batch size: 34, lr: 3.53e-03, grad_scale: 16.0 2023-05-17 02:21:02,978 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=272181.0, num_to_drop=1, layers_to_drop={0} 2023-05-17 02:21:04,686 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2023-05-17 02:21:04,985 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.001e+02 2.661e+02 3.135e+02 3.668e+02 6.921e+02, threshold=6.270e+02, percent-clipped=3.0 2023-05-17 02:21:08,154 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=272188.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 02:21:17,297 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.35 vs. limit=2.0 2023-05-17 02:21:24,211 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=272210.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 02:21:33,553 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=272223.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 02:21:36,215 INFO [finetune.py:992] (0/2) Epoch 15, batch 950, loss[loss=0.1478, simple_loss=0.249, pruned_loss=0.02328, over 12154.00 frames. ], tot_loss[loss=0.1639, simple_loss=0.2534, pruned_loss=0.03724, over 2369399.26 frames. ], batch size: 34, lr: 3.53e-03, grad_scale: 16.0 2023-05-17 02:21:47,230 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=272242.0, num_to_drop=1, layers_to_drop={3} 2023-05-17 02:21:50,577 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.0484, 4.9474, 4.7423, 4.8753, 4.5259, 5.0077, 5.0110, 5.1794], device='cuda:0'), covar=tensor([0.0229, 0.0146, 0.0242, 0.0360, 0.0800, 0.0294, 0.0155, 0.0185], device='cuda:0'), in_proj_covar=tensor([0.0195, 0.0195, 0.0189, 0.0243, 0.0238, 0.0215, 0.0175, 0.0229], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003, 0.0005], device='cuda:0') 2023-05-17 02:21:50,620 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=272247.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 02:22:00,219 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.78 vs. limit=2.0 2023-05-17 02:22:00,531 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.4394, 2.4424, 3.0752, 4.3274, 2.2236, 4.2609, 4.4436, 4.4483], device='cuda:0'), covar=tensor([0.0136, 0.1461, 0.0598, 0.0139, 0.1512, 0.0291, 0.0160, 0.0106], device='cuda:0'), in_proj_covar=tensor([0.0124, 0.0205, 0.0184, 0.0120, 0.0192, 0.0180, 0.0175, 0.0124], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-17 02:22:12,392 INFO [finetune.py:992] (0/2) Epoch 15, batch 1000, loss[loss=0.1454, simple_loss=0.2359, pruned_loss=0.02744, over 11821.00 frames. ], tot_loss[loss=0.1638, simple_loss=0.2532, pruned_loss=0.03724, over 2364384.19 frames. ], batch size: 26, lr: 3.53e-03, grad_scale: 16.0 2023-05-17 02:22:18,067 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.836e+02 2.601e+02 3.157e+02 3.834e+02 6.545e+02, threshold=6.314e+02, percent-clipped=1.0 2023-05-17 02:22:18,274 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=272284.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 02:22:22,488 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=272290.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 02:22:30,115 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.73 vs. limit=2.0 2023-05-17 02:22:35,489 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=272308.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 02:22:48,885 INFO [finetune.py:992] (0/2) Epoch 15, batch 1050, loss[loss=0.1512, simple_loss=0.2418, pruned_loss=0.03027, over 12020.00 frames. ], tot_loss[loss=0.1643, simple_loss=0.2539, pruned_loss=0.0374, over 2364482.33 frames. ], batch size: 31, lr: 3.53e-03, grad_scale: 16.0 2023-05-17 02:23:21,796 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=272373.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 02:23:24,537 INFO [finetune.py:992] (0/2) Epoch 15, batch 1100, loss[loss=0.1578, simple_loss=0.2536, pruned_loss=0.03101, over 12169.00 frames. ], tot_loss[loss=0.1642, simple_loss=0.2536, pruned_loss=0.03736, over 2367280.81 frames. ], batch size: 36, lr: 3.52e-03, grad_scale: 16.0 2023-05-17 02:23:29,568 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.747e+02 2.645e+02 3.121e+02 3.561e+02 5.261e+02, threshold=6.242e+02, percent-clipped=0.0 2023-05-17 02:24:01,152 INFO [finetune.py:992] (0/2) Epoch 15, batch 1150, loss[loss=0.1668, simple_loss=0.2639, pruned_loss=0.03489, over 12120.00 frames. ], tot_loss[loss=0.1647, simple_loss=0.2544, pruned_loss=0.03747, over 2365391.45 frames. ], batch size: 38, lr: 3.52e-03, grad_scale: 16.0 2023-05-17 02:24:07,078 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=272434.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 02:24:37,835 INFO [finetune.py:992] (0/2) Epoch 15, batch 1200, loss[loss=0.1784, simple_loss=0.269, pruned_loss=0.04392, over 12024.00 frames. ], tot_loss[loss=0.1657, simple_loss=0.2556, pruned_loss=0.0379, over 2370647.21 frames. ], batch size: 40, lr: 3.52e-03, grad_scale: 16.0 2023-05-17 02:24:42,206 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=272483.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 02:24:42,804 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.853e+02 2.776e+02 3.345e+02 4.010e+02 1.303e+03, threshold=6.691e+02, percent-clipped=6.0 2023-05-17 02:25:01,787 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=272510.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 02:25:13,703 INFO [finetune.py:992] (0/2) Epoch 15, batch 1250, loss[loss=0.1537, simple_loss=0.2486, pruned_loss=0.02944, over 12141.00 frames. ], tot_loss[loss=0.1658, simple_loss=0.2555, pruned_loss=0.03804, over 2371030.31 frames. ], batch size: 34, lr: 3.52e-03, grad_scale: 16.0 2023-05-17 02:25:20,936 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=272537.0, num_to_drop=1, layers_to_drop={0} 2023-05-17 02:25:35,698 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=272558.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 02:25:50,016 INFO [finetune.py:992] (0/2) Epoch 15, batch 1300, loss[loss=0.1379, simple_loss=0.2248, pruned_loss=0.0255, over 12421.00 frames. ], tot_loss[loss=0.1652, simple_loss=0.2552, pruned_loss=0.0376, over 2381929.29 frames. ], batch size: 32, lr: 3.52e-03, grad_scale: 16.0 2023-05-17 02:25:51,606 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=272579.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 02:25:55,540 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.40 vs. limit=2.0 2023-05-17 02:25:55,666 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.530e+02 2.581e+02 3.049e+02 3.397e+02 7.875e+02, threshold=6.098e+02, percent-clipped=1.0 2023-05-17 02:26:00,051 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=272590.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 02:26:09,219 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=272603.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 02:26:16,400 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=272613.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 02:26:23,009 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.8397, 3.4437, 5.2535, 2.6808, 2.8913, 3.8470, 3.2648, 3.8037], device='cuda:0'), covar=tensor([0.0429, 0.1170, 0.0296, 0.1231, 0.1982, 0.1562, 0.1358, 0.1171], device='cuda:0'), in_proj_covar=tensor([0.0240, 0.0240, 0.0253, 0.0187, 0.0239, 0.0298, 0.0227, 0.0266], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-17 02:26:26,346 INFO [finetune.py:992] (0/2) Epoch 15, batch 1350, loss[loss=0.1633, simple_loss=0.2493, pruned_loss=0.03861, over 12192.00 frames. ], tot_loss[loss=0.1655, simple_loss=0.2555, pruned_loss=0.03777, over 2373035.76 frames. ], batch size: 29, lr: 3.52e-03, grad_scale: 16.0 2023-05-17 02:26:31,567 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.52 vs. limit=2.0 2023-05-17 02:26:32,971 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.8353, 3.7865, 3.3453, 3.2493, 3.0123, 2.8951, 3.8893, 2.5653], device='cuda:0'), covar=tensor([0.0384, 0.0128, 0.0211, 0.0233, 0.0416, 0.0407, 0.0127, 0.0463], device='cuda:0'), in_proj_covar=tensor([0.0193, 0.0162, 0.0163, 0.0189, 0.0202, 0.0198, 0.0171, 0.0204], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-17 02:26:34,275 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=272638.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 02:26:37,880 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.0912, 5.7875, 5.4026, 5.3749, 5.9486, 5.4737, 5.3369, 5.4414], device='cuda:0'), covar=tensor([0.1640, 0.1126, 0.1269, 0.2112, 0.1011, 0.1924, 0.2260, 0.1174], device='cuda:0'), in_proj_covar=tensor([0.0354, 0.0498, 0.0400, 0.0451, 0.0463, 0.0429, 0.0395, 0.0383], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-17 02:26:39,844 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.35 vs. limit=2.0 2023-05-17 02:27:00,042 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=272674.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 02:27:01,944 INFO [finetune.py:992] (0/2) Epoch 15, batch 1400, loss[loss=0.1933, simple_loss=0.2843, pruned_loss=0.05117, over 12105.00 frames. ], tot_loss[loss=0.165, simple_loss=0.2547, pruned_loss=0.03762, over 2378622.96 frames. ], batch size: 38, lr: 3.52e-03, grad_scale: 16.0 2023-05-17 02:27:04,356 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.4633, 2.6270, 3.1231, 4.3307, 2.3923, 4.3674, 4.4911, 4.4417], device='cuda:0'), covar=tensor([0.0131, 0.1193, 0.0561, 0.0136, 0.1338, 0.0228, 0.0118, 0.0095], device='cuda:0'), in_proj_covar=tensor([0.0124, 0.0205, 0.0183, 0.0120, 0.0191, 0.0180, 0.0175, 0.0124], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-17 02:27:06,528 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.5309, 2.7953, 3.5974, 4.5370, 3.9131, 4.5565, 3.8124, 3.4190], device='cuda:0'), covar=tensor([0.0043, 0.0364, 0.0163, 0.0037, 0.0154, 0.0067, 0.0163, 0.0324], device='cuda:0'), in_proj_covar=tensor([0.0094, 0.0127, 0.0108, 0.0079, 0.0108, 0.0118, 0.0101, 0.0142], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-17 02:27:07,030 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.908e+02 2.607e+02 3.137e+02 3.946e+02 6.819e+02, threshold=6.274e+02, percent-clipped=2.0 2023-05-17 02:27:26,883 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.41 vs. limit=2.0 2023-05-17 02:27:31,028 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.90 vs. limit=2.0 2023-05-17 02:27:31,433 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.5467, 5.3102, 5.4571, 5.4934, 5.0557, 5.1442, 4.8936, 5.4728], device='cuda:0'), covar=tensor([0.0815, 0.0679, 0.0845, 0.0588, 0.2172, 0.1372, 0.0605, 0.1007], device='cuda:0'), in_proj_covar=tensor([0.0551, 0.0710, 0.0626, 0.0644, 0.0860, 0.0759, 0.0575, 0.0489], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0005, 0.0005, 0.0004, 0.0003], device='cuda:0') 2023-05-17 02:27:38,492 INFO [finetune.py:992] (0/2) Epoch 15, batch 1450, loss[loss=0.1348, simple_loss=0.2112, pruned_loss=0.02915, over 12004.00 frames. ], tot_loss[loss=0.1638, simple_loss=0.2533, pruned_loss=0.03715, over 2378667.27 frames. ], batch size: 28, lr: 3.52e-03, grad_scale: 16.0 2023-05-17 02:27:40,048 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=272729.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 02:27:44,674 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=4.21 vs. limit=5.0 2023-05-17 02:28:15,194 INFO [finetune.py:992] (0/2) Epoch 15, batch 1500, loss[loss=0.1443, simple_loss=0.2268, pruned_loss=0.03093, over 11812.00 frames. ], tot_loss[loss=0.1643, simple_loss=0.2538, pruned_loss=0.03738, over 2386749.51 frames. ], batch size: 26, lr: 3.52e-03, grad_scale: 16.0 2023-05-17 02:28:19,513 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=272783.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 02:28:20,070 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.871e+02 2.683e+02 3.061e+02 3.641e+02 1.079e+03, threshold=6.122e+02, percent-clipped=1.0 2023-05-17 02:28:34,302 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.2776, 3.9596, 4.1367, 4.3800, 2.9047, 3.8964, 2.6963, 4.1014], device='cuda:0'), covar=tensor([0.1558, 0.0748, 0.0790, 0.0674, 0.1230, 0.0665, 0.1756, 0.1045], device='cuda:0'), in_proj_covar=tensor([0.0231, 0.0268, 0.0299, 0.0357, 0.0244, 0.0247, 0.0262, 0.0376], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-17 02:28:51,259 INFO [finetune.py:992] (0/2) Epoch 15, batch 1550, loss[loss=0.1809, simple_loss=0.2818, pruned_loss=0.04006, over 11367.00 frames. ], tot_loss[loss=0.1639, simple_loss=0.2537, pruned_loss=0.03707, over 2379892.90 frames. ], batch size: 55, lr: 3.52e-03, grad_scale: 16.0 2023-05-17 02:28:54,177 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=272831.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 02:28:58,427 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=272837.0, num_to_drop=1, layers_to_drop={1} 2023-05-17 02:29:27,291 INFO [finetune.py:992] (0/2) Epoch 15, batch 1600, loss[loss=0.1503, simple_loss=0.2409, pruned_loss=0.02982, over 12099.00 frames. ], tot_loss[loss=0.1639, simple_loss=0.2538, pruned_loss=0.03703, over 2383214.95 frames. ], batch size: 33, lr: 3.52e-03, grad_scale: 16.0 2023-05-17 02:29:28,853 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=272879.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 02:29:32,359 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.933e+02 2.828e+02 3.239e+02 3.866e+02 7.933e+02, threshold=6.477e+02, percent-clipped=1.0 2023-05-17 02:29:33,841 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=272885.0, num_to_drop=1, layers_to_drop={0} 2023-05-17 02:29:46,854 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=272903.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 02:30:03,980 INFO [finetune.py:992] (0/2) Epoch 15, batch 1650, loss[loss=0.1399, simple_loss=0.2183, pruned_loss=0.03082, over 12279.00 frames. ], tot_loss[loss=0.1642, simple_loss=0.2538, pruned_loss=0.03726, over 2384675.69 frames. ], batch size: 28, lr: 3.52e-03, grad_scale: 16.0 2023-05-17 02:30:04,052 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=272927.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 02:30:16,415 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.2850, 4.9408, 5.3219, 4.6244, 4.9995, 4.6790, 5.3299, 5.0249], device='cuda:0'), covar=tensor([0.0296, 0.0419, 0.0278, 0.0272, 0.0388, 0.0350, 0.0207, 0.0350], device='cuda:0'), in_proj_covar=tensor([0.0266, 0.0270, 0.0296, 0.0268, 0.0271, 0.0271, 0.0244, 0.0218], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-17 02:30:21,358 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=272951.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 02:30:28,751 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.63 vs. limit=2.0 2023-05-17 02:30:34,200 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=272969.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 02:30:39,925 INFO [finetune.py:992] (0/2) Epoch 15, batch 1700, loss[loss=0.1747, simple_loss=0.2635, pruned_loss=0.04299, over 12288.00 frames. ], tot_loss[loss=0.1648, simple_loss=0.2547, pruned_loss=0.0375, over 2378369.69 frames. ], batch size: 37, lr: 3.52e-03, grad_scale: 16.0 2023-05-17 02:30:44,985 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.547e+02 2.674e+02 3.132e+02 3.940e+02 7.043e+02, threshold=6.265e+02, percent-clipped=2.0 2023-05-17 02:31:17,001 INFO [finetune.py:992] (0/2) Epoch 15, batch 1750, loss[loss=0.1412, simple_loss=0.2281, pruned_loss=0.0271, over 12179.00 frames. ], tot_loss[loss=0.1646, simple_loss=0.2544, pruned_loss=0.03735, over 2382624.35 frames. ], batch size: 31, lr: 3.52e-03, grad_scale: 16.0 2023-05-17 02:31:18,618 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=273029.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 02:31:27,902 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.0394, 5.7691, 5.4279, 5.3901, 5.9306, 5.1736, 5.3452, 5.3441], device='cuda:0'), covar=tensor([0.1580, 0.1038, 0.1061, 0.1833, 0.0946, 0.2139, 0.2069, 0.1342], device='cuda:0'), in_proj_covar=tensor([0.0359, 0.0503, 0.0401, 0.0454, 0.0466, 0.0436, 0.0398, 0.0389], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-17 02:31:48,418 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([6.1584, 6.0897, 5.8762, 5.2870, 5.2129, 6.0177, 5.6732, 5.3618], device='cuda:0'), covar=tensor([0.0784, 0.1184, 0.0636, 0.1604, 0.0724, 0.0794, 0.1665, 0.1119], device='cuda:0'), in_proj_covar=tensor([0.0629, 0.0568, 0.0518, 0.0640, 0.0423, 0.0731, 0.0785, 0.0575], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-05-17 02:31:49,268 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.7482, 4.4108, 4.4253, 4.6324, 3.3438, 4.4105, 2.8570, 4.4066], device='cuda:0'), covar=tensor([0.1279, 0.0569, 0.0726, 0.0649, 0.1038, 0.0470, 0.1612, 0.1230], device='cuda:0'), in_proj_covar=tensor([0.0231, 0.0269, 0.0299, 0.0359, 0.0244, 0.0247, 0.0263, 0.0378], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-17 02:31:53,343 INFO [finetune.py:992] (0/2) Epoch 15, batch 1800, loss[loss=0.1632, simple_loss=0.2565, pruned_loss=0.03493, over 10734.00 frames. ], tot_loss[loss=0.1646, simple_loss=0.2543, pruned_loss=0.03743, over 2377593.21 frames. ], batch size: 69, lr: 3.52e-03, grad_scale: 16.0 2023-05-17 02:31:53,408 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=273077.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 02:31:58,070 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.716e+02 2.669e+02 3.049e+02 3.861e+02 7.728e+02, threshold=6.099e+02, percent-clipped=2.0 2023-05-17 02:31:58,361 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.7978, 3.5595, 5.2820, 2.5905, 2.8107, 3.8912, 3.2381, 3.8571], device='cuda:0'), covar=tensor([0.0458, 0.1090, 0.0291, 0.1297, 0.1938, 0.1477, 0.1321, 0.1131], device='cuda:0'), in_proj_covar=tensor([0.0242, 0.0241, 0.0256, 0.0188, 0.0241, 0.0299, 0.0228, 0.0268], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-17 02:32:21,798 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.0823, 4.8654, 4.9738, 5.0459, 4.6951, 4.7219, 4.5438, 4.9568], device='cuda:0'), covar=tensor([0.0751, 0.0652, 0.0903, 0.0511, 0.1792, 0.1268, 0.0556, 0.1078], device='cuda:0'), in_proj_covar=tensor([0.0547, 0.0706, 0.0622, 0.0638, 0.0854, 0.0753, 0.0571, 0.0488], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0005, 0.0005, 0.0003, 0.0003], device='cuda:0') 2023-05-17 02:32:27,400 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.3790, 4.1226, 4.1752, 4.5262, 3.2075, 4.1187, 2.6481, 4.1706], device='cuda:0'), covar=tensor([0.1669, 0.0785, 0.0963, 0.0629, 0.1103, 0.0612, 0.1957, 0.1535], device='cuda:0'), in_proj_covar=tensor([0.0233, 0.0271, 0.0302, 0.0362, 0.0247, 0.0250, 0.0265, 0.0382], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-17 02:32:28,572 INFO [finetune.py:992] (0/2) Epoch 15, batch 1850, loss[loss=0.1416, simple_loss=0.2201, pruned_loss=0.03157, over 12350.00 frames. ], tot_loss[loss=0.1644, simple_loss=0.2542, pruned_loss=0.03735, over 2379657.79 frames. ], batch size: 30, lr: 3.52e-03, grad_scale: 16.0 2023-05-17 02:32:53,506 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.1092, 4.6210, 4.6957, 4.8196, 4.7794, 4.8879, 4.8250, 2.5551], device='cuda:0'), covar=tensor([0.0081, 0.0065, 0.0094, 0.0067, 0.0042, 0.0101, 0.0074, 0.0874], device='cuda:0'), in_proj_covar=tensor([0.0069, 0.0080, 0.0084, 0.0074, 0.0061, 0.0093, 0.0083, 0.0098], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-17 02:33:04,601 INFO [finetune.py:992] (0/2) Epoch 15, batch 1900, loss[loss=0.1693, simple_loss=0.2692, pruned_loss=0.03471, over 12349.00 frames. ], tot_loss[loss=0.1653, simple_loss=0.2551, pruned_loss=0.03771, over 2383099.96 frames. ], batch size: 36, lr: 3.52e-03, grad_scale: 16.0 2023-05-17 02:33:10,121 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.894e+02 2.728e+02 3.072e+02 3.546e+02 6.596e+02, threshold=6.144e+02, percent-clipped=0.0 2023-05-17 02:33:12,070 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=4.73 vs. limit=5.0 2023-05-17 02:33:33,091 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.39 vs. limit=2.0 2023-05-17 02:33:41,222 INFO [finetune.py:992] (0/2) Epoch 15, batch 1950, loss[loss=0.1635, simple_loss=0.2494, pruned_loss=0.0388, over 12256.00 frames. ], tot_loss[loss=0.1652, simple_loss=0.2547, pruned_loss=0.03788, over 2381383.42 frames. ], batch size: 32, lr: 3.52e-03, grad_scale: 16.0 2023-05-17 02:33:55,839 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.9827, 4.2899, 3.6890, 4.6673, 4.1334, 2.7298, 3.9439, 2.8154], device='cuda:0'), covar=tensor([0.0936, 0.0909, 0.1661, 0.0519, 0.1310, 0.1842, 0.1228, 0.3688], device='cuda:0'), in_proj_covar=tensor([0.0307, 0.0381, 0.0360, 0.0319, 0.0368, 0.0275, 0.0348, 0.0367], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-17 02:34:11,148 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=273269.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 02:34:16,717 INFO [finetune.py:992] (0/2) Epoch 15, batch 2000, loss[loss=0.1609, simple_loss=0.2512, pruned_loss=0.03534, over 11654.00 frames. ], tot_loss[loss=0.1654, simple_loss=0.2548, pruned_loss=0.03801, over 2378113.40 frames. ], batch size: 48, lr: 3.52e-03, grad_scale: 16.0 2023-05-17 02:34:21,753 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.866e+02 2.695e+02 3.081e+02 3.617e+02 1.680e+03, threshold=6.163e+02, percent-clipped=6.0 2023-05-17 02:34:46,007 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=273317.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 02:34:53,121 INFO [finetune.py:992] (0/2) Epoch 15, batch 2050, loss[loss=0.1769, simple_loss=0.2714, pruned_loss=0.04119, over 12346.00 frames. ], tot_loss[loss=0.1648, simple_loss=0.2543, pruned_loss=0.03765, over 2385464.42 frames. ], batch size: 36, lr: 3.52e-03, grad_scale: 16.0 2023-05-17 02:35:29,951 INFO [finetune.py:992] (0/2) Epoch 15, batch 2100, loss[loss=0.1791, simple_loss=0.2721, pruned_loss=0.04302, over 10567.00 frames. ], tot_loss[loss=0.1643, simple_loss=0.2536, pruned_loss=0.03746, over 2384842.45 frames. ], batch size: 69, lr: 3.52e-03, grad_scale: 16.0 2023-05-17 02:35:34,746 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.954e+02 2.558e+02 2.830e+02 3.304e+02 5.961e+02, threshold=5.659e+02, percent-clipped=0.0 2023-05-17 02:35:52,888 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=273409.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 02:36:05,474 INFO [finetune.py:992] (0/2) Epoch 15, batch 2150, loss[loss=0.1556, simple_loss=0.2362, pruned_loss=0.0375, over 12189.00 frames. ], tot_loss[loss=0.1649, simple_loss=0.2543, pruned_loss=0.03777, over 2390685.29 frames. ], batch size: 29, lr: 3.52e-03, grad_scale: 16.0 2023-05-17 02:36:09,986 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.2074, 6.0468, 5.7243, 5.6465, 6.1254, 5.4163, 5.5296, 5.6503], device='cuda:0'), covar=tensor([0.1585, 0.0996, 0.1078, 0.1803, 0.0896, 0.2193, 0.2262, 0.1168], device='cuda:0'), in_proj_covar=tensor([0.0359, 0.0506, 0.0405, 0.0457, 0.0469, 0.0437, 0.0402, 0.0392], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-17 02:36:11,271 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.38 vs. limit=2.0 2023-05-17 02:36:11,593 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=273435.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 02:36:15,038 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.0285, 5.9094, 5.5498, 5.4645, 5.9733, 5.2805, 5.4208, 5.4960], device='cuda:0'), covar=tensor([0.1587, 0.0905, 0.1220, 0.1811, 0.0899, 0.1954, 0.1914, 0.1048], device='cuda:0'), in_proj_covar=tensor([0.0359, 0.0506, 0.0405, 0.0457, 0.0469, 0.0438, 0.0402, 0.0392], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-17 02:36:37,282 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=273470.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 02:36:42,245 INFO [finetune.py:992] (0/2) Epoch 15, batch 2200, loss[loss=0.1783, simple_loss=0.2654, pruned_loss=0.04554, over 12144.00 frames. ], tot_loss[loss=0.1642, simple_loss=0.2535, pruned_loss=0.03747, over 2390582.01 frames. ], batch size: 39, lr: 3.52e-03, grad_scale: 16.0 2023-05-17 02:36:47,160 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.882e+02 2.612e+02 3.130e+02 3.605e+02 1.162e+03, threshold=6.261e+02, percent-clipped=3.0 2023-05-17 02:36:56,544 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=273496.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 02:37:18,402 INFO [finetune.py:992] (0/2) Epoch 15, batch 2250, loss[loss=0.163, simple_loss=0.2507, pruned_loss=0.03767, over 12304.00 frames. ], tot_loss[loss=0.1641, simple_loss=0.2534, pruned_loss=0.03736, over 2393502.67 frames. ], batch size: 34, lr: 3.52e-03, grad_scale: 16.0 2023-05-17 02:37:53,926 INFO [finetune.py:992] (0/2) Epoch 15, batch 2300, loss[loss=0.2126, simple_loss=0.295, pruned_loss=0.06513, over 8515.00 frames. ], tot_loss[loss=0.1643, simple_loss=0.2539, pruned_loss=0.03735, over 2385048.27 frames. ], batch size: 97, lr: 3.52e-03, grad_scale: 16.0 2023-05-17 02:37:58,824 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.834e+02 2.471e+02 2.860e+02 3.398e+02 1.097e+03, threshold=5.720e+02, percent-clipped=1.0 2023-05-17 02:38:14,630 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=4.90 vs. limit=5.0 2023-05-17 02:38:17,792 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.2379, 2.3355, 2.9686, 4.0952, 2.0858, 4.1918, 4.2308, 4.2422], device='cuda:0'), covar=tensor([0.0179, 0.1291, 0.0574, 0.0191, 0.1421, 0.0267, 0.0200, 0.0143], device='cuda:0'), in_proj_covar=tensor([0.0125, 0.0208, 0.0186, 0.0122, 0.0192, 0.0183, 0.0179, 0.0125], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-17 02:38:30,532 INFO [finetune.py:992] (0/2) Epoch 15, batch 2350, loss[loss=0.1529, simple_loss=0.2381, pruned_loss=0.03386, over 12012.00 frames. ], tot_loss[loss=0.1638, simple_loss=0.2535, pruned_loss=0.03707, over 2384363.24 frames. ], batch size: 28, lr: 3.52e-03, grad_scale: 16.0 2023-05-17 02:38:57,375 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.8614, 5.6641, 5.3092, 5.2211, 5.7827, 5.1871, 5.1685, 5.2528], device='cuda:0'), covar=tensor([0.1518, 0.1151, 0.1295, 0.2085, 0.0991, 0.1928, 0.2252, 0.1144], device='cuda:0'), in_proj_covar=tensor([0.0357, 0.0503, 0.0402, 0.0454, 0.0465, 0.0434, 0.0399, 0.0389], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-17 02:39:06,585 INFO [finetune.py:992] (0/2) Epoch 15, batch 2400, loss[loss=0.1453, simple_loss=0.2223, pruned_loss=0.03419, over 12281.00 frames. ], tot_loss[loss=0.1642, simple_loss=0.2541, pruned_loss=0.0371, over 2382061.19 frames. ], batch size: 28, lr: 3.51e-03, grad_scale: 16.0 2023-05-17 02:39:11,642 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.980e+02 2.708e+02 3.183e+02 3.628e+02 6.615e+02, threshold=6.367e+02, percent-clipped=2.0 2023-05-17 02:39:42,456 INFO [finetune.py:992] (0/2) Epoch 15, batch 2450, loss[loss=0.172, simple_loss=0.266, pruned_loss=0.03899, over 11188.00 frames. ], tot_loss[loss=0.1638, simple_loss=0.2537, pruned_loss=0.03694, over 2385748.05 frames. ], batch size: 55, lr: 3.51e-03, grad_scale: 16.0 2023-05-17 02:39:48,969 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.1013, 5.9818, 5.5807, 5.5263, 6.0596, 5.3213, 5.6434, 5.5135], device='cuda:0'), covar=tensor([0.1399, 0.0866, 0.1050, 0.1922, 0.0861, 0.1983, 0.1701, 0.1122], device='cuda:0'), in_proj_covar=tensor([0.0358, 0.0505, 0.0403, 0.0454, 0.0467, 0.0434, 0.0400, 0.0390], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-17 02:40:08,147 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.4410, 2.4769, 3.1116, 4.2941, 2.2541, 4.3716, 4.3932, 4.3915], device='cuda:0'), covar=tensor([0.0132, 0.1402, 0.0557, 0.0153, 0.1490, 0.0255, 0.0165, 0.0121], device='cuda:0'), in_proj_covar=tensor([0.0125, 0.0206, 0.0184, 0.0121, 0.0192, 0.0181, 0.0178, 0.0125], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-17 02:40:10,246 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=273765.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 02:40:18,793 INFO [finetune.py:992] (0/2) Epoch 15, batch 2500, loss[loss=0.2307, simple_loss=0.3116, pruned_loss=0.07487, over 7937.00 frames. ], tot_loss[loss=0.1644, simple_loss=0.2543, pruned_loss=0.03724, over 2383037.24 frames. ], batch size: 98, lr: 3.51e-03, grad_scale: 16.0 2023-05-17 02:40:21,864 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.7555, 2.8974, 4.5489, 4.6749, 2.8234, 2.7497, 2.9668, 2.2100], device='cuda:0'), covar=tensor([0.1567, 0.2957, 0.0472, 0.0455, 0.1364, 0.2381, 0.2679, 0.3890], device='cuda:0'), in_proj_covar=tensor([0.0303, 0.0396, 0.0280, 0.0307, 0.0278, 0.0318, 0.0397, 0.0382], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-17 02:40:24,381 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.876e+02 2.648e+02 3.049e+02 3.649e+02 6.935e+02, threshold=6.097e+02, percent-clipped=1.0 2023-05-17 02:40:29,431 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=273791.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 02:40:55,333 INFO [finetune.py:992] (0/2) Epoch 15, batch 2550, loss[loss=0.1591, simple_loss=0.2467, pruned_loss=0.03569, over 12158.00 frames. ], tot_loss[loss=0.1646, simple_loss=0.2547, pruned_loss=0.03725, over 2387714.21 frames. ], batch size: 34, lr: 3.51e-03, grad_scale: 16.0 2023-05-17 02:40:59,803 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.5651, 5.4799, 5.3677, 4.7643, 4.8634, 5.6130, 4.7937, 4.9792], device='cuda:0'), covar=tensor([0.1572, 0.1813, 0.1334, 0.2866, 0.1460, 0.1520, 0.3705, 0.2071], device='cuda:0'), in_proj_covar=tensor([0.0641, 0.0569, 0.0526, 0.0653, 0.0428, 0.0740, 0.0802, 0.0585], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0004, 0.0004, 0.0003], device='cuda:0') 2023-05-17 02:41:31,342 INFO [finetune.py:992] (0/2) Epoch 15, batch 2600, loss[loss=0.1671, simple_loss=0.2589, pruned_loss=0.03765, over 12352.00 frames. ], tot_loss[loss=0.1648, simple_loss=0.2545, pruned_loss=0.03754, over 2380354.90 frames. ], batch size: 35, lr: 3.51e-03, grad_scale: 16.0 2023-05-17 02:41:36,387 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.902e+02 2.651e+02 3.107e+02 3.567e+02 6.473e+02, threshold=6.214e+02, percent-clipped=1.0 2023-05-17 02:41:39,793 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.38 vs. limit=2.0 2023-05-17 02:41:50,787 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 2023-05-17 02:42:07,967 INFO [finetune.py:992] (0/2) Epoch 15, batch 2650, loss[loss=0.1505, simple_loss=0.2349, pruned_loss=0.03308, over 12124.00 frames. ], tot_loss[loss=0.1641, simple_loss=0.2538, pruned_loss=0.0372, over 2382927.85 frames. ], batch size: 30, lr: 3.51e-03, grad_scale: 16.0 2023-05-17 02:42:43,085 INFO [finetune.py:992] (0/2) Epoch 15, batch 2700, loss[loss=0.1862, simple_loss=0.2732, pruned_loss=0.04958, over 11317.00 frames. ], tot_loss[loss=0.1649, simple_loss=0.2543, pruned_loss=0.03778, over 2382545.89 frames. ], batch size: 55, lr: 3.51e-03, grad_scale: 16.0 2023-05-17 02:42:43,607 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.86 vs. limit=2.0 2023-05-17 02:42:48,121 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.877e+02 2.625e+02 3.117e+02 3.846e+02 1.086e+03, threshold=6.234e+02, percent-clipped=6.0 2023-05-17 02:42:59,805 INFO [checkpoint.py:75] (0/2) Saving checkpoint to pruned_transducer_stateless7/exp_giga_finetune/checkpoint-174000.pt 2023-05-17 02:43:03,569 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=274001.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 02:43:05,485 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.2400, 5.0115, 5.1792, 5.1681, 4.6728, 4.6974, 4.6657, 5.0363], device='cuda:0'), covar=tensor([0.0878, 0.0992, 0.0982, 0.0885, 0.3121, 0.1890, 0.0689, 0.1349], device='cuda:0'), in_proj_covar=tensor([0.0555, 0.0722, 0.0633, 0.0653, 0.0875, 0.0769, 0.0580, 0.0495], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0005, 0.0004, 0.0004, 0.0005, 0.0005, 0.0004, 0.0003], device='cuda:0') 2023-05-17 02:43:13,187 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.5663, 5.3692, 5.5136, 5.5181, 5.1746, 5.1762, 4.9612, 5.4762], device='cuda:0'), covar=tensor([0.0749, 0.0634, 0.0749, 0.0666, 0.2027, 0.1475, 0.0527, 0.0956], device='cuda:0'), in_proj_covar=tensor([0.0555, 0.0722, 0.0633, 0.0654, 0.0875, 0.0770, 0.0580, 0.0495], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0005, 0.0004, 0.0004, 0.0005, 0.0005, 0.0004, 0.0003], device='cuda:0') 2023-05-17 02:43:21,526 INFO [finetune.py:992] (0/2) Epoch 15, batch 2750, loss[loss=0.1294, simple_loss=0.2175, pruned_loss=0.02062, over 11330.00 frames. ], tot_loss[loss=0.1648, simple_loss=0.2544, pruned_loss=0.03757, over 2380908.61 frames. ], batch size: 25, lr: 3.51e-03, grad_scale: 16.0 2023-05-17 02:43:33,842 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.9391, 4.6179, 4.7874, 4.8805, 4.6283, 4.8970, 4.7127, 2.2875], device='cuda:0'), covar=tensor([0.0124, 0.0078, 0.0096, 0.0072, 0.0055, 0.0109, 0.0094, 0.1009], device='cuda:0'), in_proj_covar=tensor([0.0070, 0.0081, 0.0085, 0.0075, 0.0062, 0.0094, 0.0084, 0.0099], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-17 02:43:47,448 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=274062.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 02:43:48,333 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.62 vs. limit=2.0 2023-05-17 02:43:49,492 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=274065.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 02:43:58,699 INFO [finetune.py:992] (0/2) Epoch 15, batch 2800, loss[loss=0.1972, simple_loss=0.2717, pruned_loss=0.06137, over 8317.00 frames. ], tot_loss[loss=0.1647, simple_loss=0.2541, pruned_loss=0.03766, over 2380375.91 frames. ], batch size: 98, lr: 3.51e-03, grad_scale: 16.0 2023-05-17 02:44:03,733 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.932e+02 2.628e+02 3.055e+02 3.704e+02 7.126e+02, threshold=6.111e+02, percent-clipped=1.0 2023-05-17 02:44:08,932 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=274091.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 02:44:24,282 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=274113.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 02:44:32,526 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2023-05-17 02:44:34,250 INFO [finetune.py:992] (0/2) Epoch 15, batch 2850, loss[loss=0.2202, simple_loss=0.3016, pruned_loss=0.06937, over 7915.00 frames. ], tot_loss[loss=0.165, simple_loss=0.2545, pruned_loss=0.03774, over 2372285.90 frames. ], batch size: 98, lr: 3.51e-03, grad_scale: 16.0 2023-05-17 02:44:37,226 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=274131.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 02:44:39,570 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.47 vs. limit=2.0 2023-05-17 02:44:42,786 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=274139.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 02:45:09,684 INFO [finetune.py:992] (0/2) Epoch 15, batch 2900, loss[loss=0.1536, simple_loss=0.2342, pruned_loss=0.03655, over 12343.00 frames. ], tot_loss[loss=0.1646, simple_loss=0.2543, pruned_loss=0.0374, over 2378949.44 frames. ], batch size: 31, lr: 3.51e-03, grad_scale: 32.0 2023-05-17 02:45:14,749 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.761e+02 2.580e+02 2.952e+02 3.474e+02 5.749e+02, threshold=5.904e+02, percent-clipped=0.0 2023-05-17 02:45:20,584 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=274192.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 02:45:47,195 INFO [finetune.py:992] (0/2) Epoch 15, batch 2950, loss[loss=0.1473, simple_loss=0.2269, pruned_loss=0.03381, over 11840.00 frames. ], tot_loss[loss=0.1637, simple_loss=0.2533, pruned_loss=0.03704, over 2379569.50 frames. ], batch size: 26, lr: 3.51e-03, grad_scale: 32.0 2023-05-17 02:45:57,137 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.0803, 4.8851, 4.9575, 5.1165, 4.8220, 5.0952, 4.9480, 2.7771], device='cuda:0'), covar=tensor([0.0099, 0.0058, 0.0078, 0.0048, 0.0039, 0.0090, 0.0067, 0.0692], device='cuda:0'), in_proj_covar=tensor([0.0070, 0.0081, 0.0085, 0.0075, 0.0062, 0.0094, 0.0084, 0.0099], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-17 02:46:01,310 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.5304, 5.1141, 5.5358, 4.8606, 5.1327, 4.9531, 5.5459, 5.1665], device='cuda:0'), covar=tensor([0.0255, 0.0356, 0.0236, 0.0234, 0.0390, 0.0319, 0.0198, 0.0270], device='cuda:0'), in_proj_covar=tensor([0.0269, 0.0272, 0.0298, 0.0270, 0.0274, 0.0271, 0.0246, 0.0218], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-17 02:46:16,891 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=274269.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 02:46:22,444 INFO [finetune.py:992] (0/2) Epoch 15, batch 3000, loss[loss=0.1487, simple_loss=0.2281, pruned_loss=0.03464, over 12004.00 frames. ], tot_loss[loss=0.1632, simple_loss=0.2527, pruned_loss=0.0368, over 2386528.41 frames. ], batch size: 28, lr: 3.51e-03, grad_scale: 32.0 2023-05-17 02:46:22,445 INFO [finetune.py:1017] (0/2) Computing validation loss 2023-05-17 02:46:40,668 INFO [finetune.py:1026] (0/2) Epoch 15, validation: loss=0.3142, simple_loss=0.3917, pruned_loss=0.1184, over 1020973.00 frames. 2023-05-17 02:46:40,669 INFO [finetune.py:1027] (0/2) Maximum memory allocated so far is 12827MB 2023-05-17 02:46:45,564 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.919e+02 2.680e+02 3.036e+02 3.627e+02 7.186e+02, threshold=6.071e+02, percent-clipped=2.0 2023-05-17 02:46:59,884 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=274303.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 02:47:17,333 INFO [finetune.py:992] (0/2) Epoch 15, batch 3050, loss[loss=0.153, simple_loss=0.2414, pruned_loss=0.03234, over 12297.00 frames. ], tot_loss[loss=0.1624, simple_loss=0.2518, pruned_loss=0.03649, over 2385713.90 frames. ], batch size: 33, lr: 3.51e-03, grad_scale: 32.0 2023-05-17 02:47:19,690 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=274330.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 02:47:39,043 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=274357.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 02:47:44,046 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=274364.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 02:47:44,387 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=4.61 vs. limit=5.0 2023-05-17 02:47:53,129 INFO [finetune.py:992] (0/2) Epoch 15, batch 3100, loss[loss=0.1715, simple_loss=0.2679, pruned_loss=0.0375, over 10421.00 frames. ], tot_loss[loss=0.1622, simple_loss=0.2517, pruned_loss=0.03633, over 2383997.22 frames. ], batch size: 68, lr: 3.51e-03, grad_scale: 32.0 2023-05-17 02:47:58,032 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.719e+02 2.566e+02 2.991e+02 3.624e+02 6.819e+02, threshold=5.983e+02, percent-clipped=2.0 2023-05-17 02:48:17,676 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.9755, 5.8383, 5.4473, 5.4319, 5.9580, 5.3303, 5.4442, 5.4379], device='cuda:0'), covar=tensor([0.1665, 0.1031, 0.1187, 0.1874, 0.0977, 0.2076, 0.2071, 0.1135], device='cuda:0'), in_proj_covar=tensor([0.0363, 0.0512, 0.0409, 0.0459, 0.0471, 0.0443, 0.0403, 0.0394], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-17 02:48:19,300 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.8000, 2.6100, 3.9407, 4.1220, 2.8749, 2.5803, 2.7853, 2.2019], device='cuda:0'), covar=tensor([0.1590, 0.2958, 0.0626, 0.0500, 0.1232, 0.2522, 0.2881, 0.4155], device='cuda:0'), in_proj_covar=tensor([0.0301, 0.0393, 0.0279, 0.0305, 0.0277, 0.0316, 0.0394, 0.0379], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-17 02:48:28,795 INFO [finetune.py:992] (0/2) Epoch 15, batch 3150, loss[loss=0.1744, simple_loss=0.2708, pruned_loss=0.03898, over 12134.00 frames. ], tot_loss[loss=0.163, simple_loss=0.2524, pruned_loss=0.03678, over 2379825.07 frames. ], batch size: 38, lr: 3.51e-03, grad_scale: 32.0 2023-05-17 02:48:58,130 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 2023-05-17 02:49:05,476 INFO [finetune.py:992] (0/2) Epoch 15, batch 3200, loss[loss=0.1816, simple_loss=0.2665, pruned_loss=0.04836, over 12118.00 frames. ], tot_loss[loss=0.1632, simple_loss=0.2526, pruned_loss=0.03695, over 2380513.70 frames. ], batch size: 33, lr: 3.51e-03, grad_scale: 32.0 2023-05-17 02:49:10,430 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.982e+02 2.855e+02 3.190e+02 3.704e+02 5.907e+02, threshold=6.379e+02, percent-clipped=0.0 2023-05-17 02:49:12,614 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=274487.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 02:49:31,879 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.3631, 2.4381, 3.5928, 4.3277, 3.7436, 4.3310, 3.8466, 3.0596], device='cuda:0'), covar=tensor([0.0047, 0.0445, 0.0150, 0.0045, 0.0132, 0.0077, 0.0120, 0.0383], device='cuda:0'), in_proj_covar=tensor([0.0092, 0.0124, 0.0106, 0.0079, 0.0107, 0.0117, 0.0099, 0.0140], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-17 02:49:40,957 INFO [finetune.py:992] (0/2) Epoch 15, batch 3250, loss[loss=0.1569, simple_loss=0.2548, pruned_loss=0.02956, over 12164.00 frames. ], tot_loss[loss=0.164, simple_loss=0.2531, pruned_loss=0.03743, over 2376768.52 frames. ], batch size: 36, lr: 3.51e-03, grad_scale: 32.0 2023-05-17 02:49:44,015 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.6161, 3.6644, 3.3806, 3.2800, 3.0231, 2.8214, 3.8138, 2.5344], device='cuda:0'), covar=tensor([0.0374, 0.0154, 0.0188, 0.0196, 0.0379, 0.0415, 0.0112, 0.0475], device='cuda:0'), in_proj_covar=tensor([0.0196, 0.0166, 0.0168, 0.0190, 0.0205, 0.0201, 0.0174, 0.0209], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-17 02:49:50,331 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=274540.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 02:50:17,315 INFO [finetune.py:992] (0/2) Epoch 15, batch 3300, loss[loss=0.2164, simple_loss=0.2875, pruned_loss=0.07262, over 8056.00 frames. ], tot_loss[loss=0.1641, simple_loss=0.2531, pruned_loss=0.03758, over 2368396.32 frames. ], batch size: 100, lr: 3.51e-03, grad_scale: 32.0 2023-05-17 02:50:22,370 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.741e+02 2.627e+02 3.016e+02 3.479e+02 6.691e+02, threshold=6.031e+02, percent-clipped=1.0 2023-05-17 02:50:34,865 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=274601.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 02:50:43,358 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.4922, 4.8600, 3.2993, 2.9579, 4.1174, 2.8078, 4.0387, 3.4299], device='cuda:0'), covar=tensor([0.0656, 0.0413, 0.0902, 0.1320, 0.0253, 0.1215, 0.0491, 0.0750], device='cuda:0'), in_proj_covar=tensor([0.0194, 0.0262, 0.0180, 0.0204, 0.0146, 0.0187, 0.0201, 0.0178], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-17 02:50:52,259 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=274625.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 02:50:53,643 INFO [finetune.py:992] (0/2) Epoch 15, batch 3350, loss[loss=0.1839, simple_loss=0.2769, pruned_loss=0.04543, over 12112.00 frames. ], tot_loss[loss=0.1642, simple_loss=0.2532, pruned_loss=0.03758, over 2375372.46 frames. ], batch size: 38, lr: 3.51e-03, grad_scale: 32.0 2023-05-17 02:51:14,862 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.8213, 3.8281, 3.4583, 3.4638, 3.1021, 3.0490, 3.9613, 2.6109], device='cuda:0'), covar=tensor([0.0399, 0.0171, 0.0196, 0.0209, 0.0393, 0.0349, 0.0112, 0.0461], device='cuda:0'), in_proj_covar=tensor([0.0197, 0.0166, 0.0169, 0.0191, 0.0205, 0.0203, 0.0175, 0.0210], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-17 02:51:15,453 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=274657.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 02:51:16,824 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=274659.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 02:51:29,372 INFO [finetune.py:992] (0/2) Epoch 15, batch 3400, loss[loss=0.16, simple_loss=0.2542, pruned_loss=0.03294, over 12103.00 frames. ], tot_loss[loss=0.1645, simple_loss=0.2534, pruned_loss=0.0378, over 2375740.75 frames. ], batch size: 33, lr: 3.51e-03, grad_scale: 32.0 2023-05-17 02:51:34,338 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.652e+02 2.752e+02 3.196e+02 3.746e+02 7.387e+02, threshold=6.392e+02, percent-clipped=3.0 2023-05-17 02:51:34,582 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.4563, 4.8478, 3.1599, 2.9631, 4.0590, 2.6392, 4.0553, 3.2813], device='cuda:0'), covar=tensor([0.0668, 0.0451, 0.1005, 0.1324, 0.0274, 0.1295, 0.0421, 0.0819], device='cuda:0'), in_proj_covar=tensor([0.0194, 0.0263, 0.0181, 0.0204, 0.0146, 0.0187, 0.0201, 0.0178], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-17 02:51:49,560 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=274705.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 02:51:52,531 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.2735, 5.0917, 5.2295, 5.2566, 4.8580, 4.9203, 4.6762, 5.1739], device='cuda:0'), covar=tensor([0.0836, 0.0697, 0.0844, 0.0713, 0.2068, 0.1568, 0.0650, 0.1129], device='cuda:0'), in_proj_covar=tensor([0.0557, 0.0721, 0.0634, 0.0651, 0.0873, 0.0771, 0.0581, 0.0495], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0005, 0.0004, 0.0004, 0.0005, 0.0005, 0.0004, 0.0003], device='cuda:0') 2023-05-17 02:52:02,636 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=4.43 vs. limit=5.0 2023-05-17 02:52:05,807 INFO [finetune.py:992] (0/2) Epoch 15, batch 3450, loss[loss=0.1578, simple_loss=0.2484, pruned_loss=0.03361, over 12192.00 frames. ], tot_loss[loss=0.164, simple_loss=0.2529, pruned_loss=0.03756, over 2375331.87 frames. ], batch size: 35, lr: 3.51e-03, grad_scale: 32.0 2023-05-17 02:52:28,473 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.1435, 5.9929, 5.6526, 5.5022, 6.0404, 5.2559, 5.5425, 5.5890], device='cuda:0'), covar=tensor([0.1523, 0.0881, 0.1169, 0.1960, 0.0886, 0.2414, 0.1839, 0.1177], device='cuda:0'), in_proj_covar=tensor([0.0360, 0.0507, 0.0406, 0.0456, 0.0468, 0.0441, 0.0400, 0.0390], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-17 02:52:30,798 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=3.20 vs. limit=5.0 2023-05-17 02:52:41,746 INFO [finetune.py:992] (0/2) Epoch 15, batch 3500, loss[loss=0.1921, simple_loss=0.2636, pruned_loss=0.06032, over 8015.00 frames. ], tot_loss[loss=0.1642, simple_loss=0.2531, pruned_loss=0.03766, over 2372808.72 frames. ], batch size: 98, lr: 3.51e-03, grad_scale: 32.0 2023-05-17 02:52:46,625 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.822e+02 2.661e+02 2.952e+02 3.432e+02 7.718e+02, threshold=5.904e+02, percent-clipped=1.0 2023-05-17 02:52:48,928 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=274787.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 02:52:59,233 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.4482, 4.2675, 4.2874, 4.6125, 3.0746, 4.0920, 2.9107, 4.2421], device='cuda:0'), covar=tensor([0.1528, 0.0669, 0.0932, 0.0658, 0.1297, 0.0606, 0.1744, 0.1303], device='cuda:0'), in_proj_covar=tensor([0.0230, 0.0268, 0.0299, 0.0361, 0.0244, 0.0245, 0.0262, 0.0378], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-17 02:53:17,159 INFO [finetune.py:992] (0/2) Epoch 15, batch 3550, loss[loss=0.1493, simple_loss=0.2297, pruned_loss=0.03449, over 12336.00 frames. ], tot_loss[loss=0.1641, simple_loss=0.2531, pruned_loss=0.03756, over 2377506.71 frames. ], batch size: 30, lr: 3.51e-03, grad_scale: 32.0 2023-05-17 02:53:22,879 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=274835.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 02:53:22,963 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.4443, 5.2715, 5.3915, 5.4295, 5.0024, 5.1097, 4.8625, 5.3912], device='cuda:0'), covar=tensor([0.0772, 0.0677, 0.0764, 0.0595, 0.2146, 0.1287, 0.0546, 0.0974], device='cuda:0'), in_proj_covar=tensor([0.0558, 0.0720, 0.0631, 0.0651, 0.0874, 0.0768, 0.0582, 0.0494], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0005, 0.0004, 0.0004, 0.0005, 0.0005, 0.0004, 0.0003], device='cuda:0') 2023-05-17 02:53:53,484 INFO [finetune.py:992] (0/2) Epoch 15, batch 3600, loss[loss=0.1606, simple_loss=0.2504, pruned_loss=0.03545, over 12253.00 frames. ], tot_loss[loss=0.1645, simple_loss=0.2535, pruned_loss=0.03773, over 2371168.08 frames. ], batch size: 32, lr: 3.51e-03, grad_scale: 32.0 2023-05-17 02:53:58,375 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.915e+02 2.835e+02 3.190e+02 3.643e+02 6.667e+02, threshold=6.380e+02, percent-clipped=3.0 2023-05-17 02:54:05,712 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=274894.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 02:54:07,085 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=274896.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 02:54:28,447 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=274925.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 02:54:29,715 INFO [finetune.py:992] (0/2) Epoch 15, batch 3650, loss[loss=0.1709, simple_loss=0.265, pruned_loss=0.03841, over 12150.00 frames. ], tot_loss[loss=0.1643, simple_loss=0.2534, pruned_loss=0.03755, over 2378106.70 frames. ], batch size: 39, lr: 3.51e-03, grad_scale: 16.0 2023-05-17 02:54:46,334 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.4270, 2.8052, 3.0401, 4.2045, 2.2716, 4.3100, 4.3237, 4.4312], device='cuda:0'), covar=tensor([0.0121, 0.1090, 0.0564, 0.0160, 0.1377, 0.0283, 0.0157, 0.0097], device='cuda:0'), in_proj_covar=tensor([0.0126, 0.0206, 0.0184, 0.0122, 0.0191, 0.0183, 0.0179, 0.0125], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-17 02:54:49,948 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=274955.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 02:54:52,798 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=274959.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 02:55:00,996 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-05-17 02:55:02,668 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=274973.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 02:55:05,401 INFO [finetune.py:992] (0/2) Epoch 15, batch 3700, loss[loss=0.1656, simple_loss=0.2588, pruned_loss=0.03623, over 11271.00 frames. ], tot_loss[loss=0.1642, simple_loss=0.2537, pruned_loss=0.03734, over 2378279.56 frames. ], batch size: 55, lr: 3.50e-03, grad_scale: 16.0 2023-05-17 02:55:11,054 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.800e+02 2.616e+02 2.961e+02 3.537e+02 1.203e+03, threshold=5.922e+02, percent-clipped=3.0 2023-05-17 02:55:27,130 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=275007.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 02:55:27,155 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.8412, 4.5962, 4.2092, 4.2073, 4.6595, 3.9839, 4.2668, 4.0017], device='cuda:0'), covar=tensor([0.1674, 0.1084, 0.1466, 0.2056, 0.1081, 0.2333, 0.1673, 0.1409], device='cuda:0'), in_proj_covar=tensor([0.0359, 0.0503, 0.0405, 0.0455, 0.0467, 0.0441, 0.0400, 0.0390], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-17 02:55:31,625 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.1737, 3.5714, 3.5987, 4.0012, 2.8410, 3.5239, 2.6089, 3.4689], device='cuda:0'), covar=tensor([0.1610, 0.0792, 0.0934, 0.0652, 0.1122, 0.0702, 0.1684, 0.0978], device='cuda:0'), in_proj_covar=tensor([0.0230, 0.0268, 0.0299, 0.0361, 0.0243, 0.0245, 0.0262, 0.0377], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-17 02:55:32,897 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.2319, 4.8394, 4.9771, 5.0479, 4.8890, 5.0727, 4.9469, 2.4841], device='cuda:0'), covar=tensor([0.0094, 0.0075, 0.0084, 0.0064, 0.0046, 0.0115, 0.0083, 0.0859], device='cuda:0'), in_proj_covar=tensor([0.0071, 0.0081, 0.0085, 0.0076, 0.0063, 0.0095, 0.0084, 0.0100], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-17 02:55:41,940 INFO [finetune.py:992] (0/2) Epoch 15, batch 3750, loss[loss=0.1732, simple_loss=0.2652, pruned_loss=0.04064, over 11564.00 frames. ], tot_loss[loss=0.1644, simple_loss=0.2537, pruned_loss=0.03751, over 2376959.21 frames. ], batch size: 48, lr: 3.50e-03, grad_scale: 16.0 2023-05-17 02:56:08,133 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.7085, 4.5799, 4.6075, 4.5927, 4.2851, 4.7127, 4.7222, 4.8981], device='cuda:0'), covar=tensor([0.0268, 0.0183, 0.0205, 0.0375, 0.0747, 0.0391, 0.0189, 0.0190], device='cuda:0'), in_proj_covar=tensor([0.0202, 0.0204, 0.0198, 0.0257, 0.0249, 0.0227, 0.0184, 0.0235], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0005, 0.0004, 0.0004, 0.0003, 0.0005], device='cuda:0') 2023-05-17 02:56:13,752 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=3.40 vs. limit=5.0 2023-05-17 02:56:17,617 INFO [finetune.py:992] (0/2) Epoch 15, batch 3800, loss[loss=0.169, simple_loss=0.2623, pruned_loss=0.03783, over 11371.00 frames. ], tot_loss[loss=0.165, simple_loss=0.2542, pruned_loss=0.03788, over 2379625.50 frames. ], batch size: 55, lr: 3.50e-03, grad_scale: 16.0 2023-05-17 02:56:23,263 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.098e+02 2.835e+02 3.309e+02 3.845e+02 7.855e+02, threshold=6.617e+02, percent-clipped=6.0 2023-05-17 02:56:31,324 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.1745, 4.7745, 5.1519, 4.4749, 4.8143, 4.5809, 5.1795, 4.7824], device='cuda:0'), covar=tensor([0.0297, 0.0386, 0.0323, 0.0298, 0.0446, 0.0347, 0.0213, 0.0414], device='cuda:0'), in_proj_covar=tensor([0.0268, 0.0272, 0.0300, 0.0270, 0.0273, 0.0271, 0.0246, 0.0218], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-17 02:56:52,559 INFO [finetune.py:992] (0/2) Epoch 15, batch 3850, loss[loss=0.1372, simple_loss=0.219, pruned_loss=0.02768, over 11999.00 frames. ], tot_loss[loss=0.1658, simple_loss=0.2546, pruned_loss=0.03854, over 2377160.37 frames. ], batch size: 28, lr: 3.50e-03, grad_scale: 16.0 2023-05-17 02:57:15,588 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=275158.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 02:57:28,948 INFO [finetune.py:992] (0/2) Epoch 15, batch 3900, loss[loss=0.1339, simple_loss=0.2189, pruned_loss=0.0244, over 12181.00 frames. ], tot_loss[loss=0.1653, simple_loss=0.254, pruned_loss=0.03826, over 2376889.43 frames. ], batch size: 29, lr: 3.50e-03, grad_scale: 16.0 2023-05-17 02:57:34,722 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.858e+02 2.712e+02 3.194e+02 3.668e+02 1.051e+03, threshold=6.388e+02, percent-clipped=2.0 2023-05-17 02:57:37,781 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.6393, 2.9278, 3.3052, 4.4548, 2.4924, 4.6319, 4.5575, 4.7349], device='cuda:0'), covar=tensor([0.0127, 0.1193, 0.0516, 0.0173, 0.1348, 0.0231, 0.0190, 0.0092], device='cuda:0'), in_proj_covar=tensor([0.0125, 0.0206, 0.0184, 0.0122, 0.0192, 0.0183, 0.0179, 0.0125], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-17 02:57:42,892 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=275196.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 02:57:55,895 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.5512, 5.2007, 5.5455, 4.8840, 5.1904, 4.9514, 5.5586, 5.1567], device='cuda:0'), covar=tensor([0.0234, 0.0306, 0.0248, 0.0255, 0.0359, 0.0321, 0.0175, 0.0277], device='cuda:0'), in_proj_covar=tensor([0.0269, 0.0272, 0.0299, 0.0270, 0.0273, 0.0271, 0.0245, 0.0218], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-17 02:58:00,345 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=275219.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 02:58:05,752 INFO [finetune.py:992] (0/2) Epoch 15, batch 3950, loss[loss=0.1545, simple_loss=0.25, pruned_loss=0.02951, over 11558.00 frames. ], tot_loss[loss=0.1648, simple_loss=0.2538, pruned_loss=0.03793, over 2367485.63 frames. ], batch size: 48, lr: 3.50e-03, grad_scale: 16.0 2023-05-17 02:58:17,889 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=275244.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 02:58:22,303 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=275250.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 02:58:32,597 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.6250, 3.8769, 3.4186, 3.3474, 3.0958, 2.8833, 3.8227, 2.5882], device='cuda:0'), covar=tensor([0.0450, 0.0117, 0.0181, 0.0208, 0.0382, 0.0391, 0.0154, 0.0484], device='cuda:0'), in_proj_covar=tensor([0.0196, 0.0165, 0.0168, 0.0190, 0.0204, 0.0203, 0.0175, 0.0209], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-17 02:58:41,637 INFO [finetune.py:992] (0/2) Epoch 15, batch 4000, loss[loss=0.1552, simple_loss=0.2408, pruned_loss=0.03483, over 12034.00 frames. ], tot_loss[loss=0.1649, simple_loss=0.2542, pruned_loss=0.03781, over 2372926.00 frames. ], batch size: 31, lr: 3.50e-03, grad_scale: 16.0 2023-05-17 02:58:47,338 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.964e+02 2.716e+02 3.114e+02 3.647e+02 6.314e+02, threshold=6.229e+02, percent-clipped=0.0 2023-05-17 02:58:56,242 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.1755, 2.4641, 3.5836, 4.1656, 3.7109, 4.1427, 3.7990, 3.1015], device='cuda:0'), covar=tensor([0.0051, 0.0388, 0.0128, 0.0049, 0.0119, 0.0073, 0.0110, 0.0321], device='cuda:0'), in_proj_covar=tensor([0.0091, 0.0123, 0.0105, 0.0078, 0.0107, 0.0116, 0.0099, 0.0140], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-17 02:59:03,991 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=275307.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 02:59:18,109 INFO [finetune.py:992] (0/2) Epoch 15, batch 4050, loss[loss=0.1571, simple_loss=0.2471, pruned_loss=0.03354, over 12064.00 frames. ], tot_loss[loss=0.1649, simple_loss=0.2543, pruned_loss=0.03779, over 2370245.07 frames. ], batch size: 40, lr: 3.50e-03, grad_scale: 16.0 2023-05-17 02:59:32,673 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=275347.0, num_to_drop=1, layers_to_drop={0} 2023-05-17 02:59:48,551 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=275368.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 02:59:52,281 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.12 vs. limit=2.0 2023-05-17 02:59:54,691 INFO [finetune.py:992] (0/2) Epoch 15, batch 4100, loss[loss=0.1711, simple_loss=0.2696, pruned_loss=0.03634, over 11860.00 frames. ], tot_loss[loss=0.1646, simple_loss=0.254, pruned_loss=0.03766, over 2371765.41 frames. ], batch size: 44, lr: 3.50e-03, grad_scale: 16.0 2023-05-17 02:59:59,379 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2023-05-17 03:00:00,457 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.568e+02 2.590e+02 3.056e+02 3.712e+02 6.933e+02, threshold=6.111e+02, percent-clipped=2.0 2023-05-17 03:00:00,618 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=275385.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 03:00:17,299 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=275408.0, num_to_drop=1, layers_to_drop={0} 2023-05-17 03:00:30,584 INFO [finetune.py:992] (0/2) Epoch 15, batch 4150, loss[loss=0.1503, simple_loss=0.2401, pruned_loss=0.03029, over 12421.00 frames. ], tot_loss[loss=0.1639, simple_loss=0.253, pruned_loss=0.03737, over 2369505.78 frames. ], batch size: 32, lr: 3.50e-03, grad_scale: 16.0 2023-05-17 03:00:31,043 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.63 vs. limit=2.0 2023-05-17 03:00:32,430 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=3.49 vs. limit=5.0 2023-05-17 03:00:45,058 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=275446.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 03:00:54,704 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.37 vs. limit=2.0 2023-05-17 03:01:06,512 INFO [finetune.py:992] (0/2) Epoch 15, batch 4200, loss[loss=0.1562, simple_loss=0.2392, pruned_loss=0.03658, over 12326.00 frames. ], tot_loss[loss=0.1641, simple_loss=0.253, pruned_loss=0.0376, over 2366163.39 frames. ], batch size: 30, lr: 3.50e-03, grad_scale: 16.0 2023-05-17 03:01:12,128 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.724e+02 2.645e+02 3.044e+02 3.736e+02 5.843e+02, threshold=6.088e+02, percent-clipped=0.0 2023-05-17 03:01:19,106 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=4.96 vs. limit=5.0 2023-05-17 03:01:33,383 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=275514.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 03:01:42,469 INFO [finetune.py:992] (0/2) Epoch 15, batch 4250, loss[loss=0.1814, simple_loss=0.2662, pruned_loss=0.04831, over 11284.00 frames. ], tot_loss[loss=0.1643, simple_loss=0.2534, pruned_loss=0.03761, over 2378012.36 frames. ], batch size: 55, lr: 3.50e-03, grad_scale: 16.0 2023-05-17 03:01:56,999 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.8642, 3.8625, 3.3442, 3.3037, 3.0041, 2.9295, 3.8858, 2.5370], device='cuda:0'), covar=tensor([0.0391, 0.0138, 0.0235, 0.0211, 0.0436, 0.0366, 0.0120, 0.0506], device='cuda:0'), in_proj_covar=tensor([0.0195, 0.0164, 0.0168, 0.0189, 0.0204, 0.0202, 0.0175, 0.0208], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-17 03:01:59,001 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=275550.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 03:02:17,795 INFO [finetune.py:992] (0/2) Epoch 15, batch 4300, loss[loss=0.1783, simple_loss=0.2694, pruned_loss=0.04358, over 11288.00 frames. ], tot_loss[loss=0.1651, simple_loss=0.2542, pruned_loss=0.03803, over 2371324.50 frames. ], batch size: 55, lr: 3.50e-03, grad_scale: 16.0 2023-05-17 03:02:24,276 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.647e+02 2.684e+02 3.230e+02 3.791e+02 8.652e+02, threshold=6.460e+02, percent-clipped=1.0 2023-05-17 03:02:33,648 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=275598.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 03:02:54,704 INFO [finetune.py:992] (0/2) Epoch 15, batch 4350, loss[loss=0.147, simple_loss=0.2382, pruned_loss=0.02788, over 12120.00 frames. ], tot_loss[loss=0.1651, simple_loss=0.2543, pruned_loss=0.03795, over 2370008.25 frames. ], batch size: 30, lr: 3.50e-03, grad_scale: 16.0 2023-05-17 03:03:05,898 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.53 vs. limit=2.0 2023-05-17 03:03:21,598 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=275663.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 03:03:24,508 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.2154, 4.0204, 4.0297, 4.3535, 2.9345, 3.9932, 2.7100, 4.0932], device='cuda:0'), covar=tensor([0.1679, 0.0734, 0.0914, 0.0654, 0.1274, 0.0628, 0.1775, 0.1180], device='cuda:0'), in_proj_covar=tensor([0.0231, 0.0269, 0.0301, 0.0363, 0.0244, 0.0247, 0.0264, 0.0380], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-17 03:03:31,330 INFO [finetune.py:992] (0/2) Epoch 15, batch 4400, loss[loss=0.1673, simple_loss=0.2611, pruned_loss=0.03674, over 11300.00 frames. ], tot_loss[loss=0.1649, simple_loss=0.254, pruned_loss=0.03785, over 2372295.45 frames. ], batch size: 55, lr: 3.50e-03, grad_scale: 16.0 2023-05-17 03:03:32,892 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.2951, 2.7867, 3.8201, 3.1893, 3.6891, 3.3755, 2.7972, 3.8060], device='cuda:0'), covar=tensor([0.0139, 0.0317, 0.0184, 0.0244, 0.0144, 0.0173, 0.0337, 0.0109], device='cuda:0'), in_proj_covar=tensor([0.0186, 0.0209, 0.0197, 0.0192, 0.0224, 0.0171, 0.0204, 0.0193], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-17 03:03:36,991 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.030e+02 2.679e+02 3.186e+02 3.610e+02 8.620e+02, threshold=6.372e+02, percent-clipped=1.0 2023-05-17 03:03:50,032 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=275703.0, num_to_drop=1, layers_to_drop={1} 2023-05-17 03:04:00,116 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.4510, 2.4067, 3.1285, 4.2806, 2.2257, 4.3544, 4.4753, 4.5033], device='cuda:0'), covar=tensor([0.0115, 0.1434, 0.0530, 0.0146, 0.1449, 0.0232, 0.0146, 0.0099], device='cuda:0'), in_proj_covar=tensor([0.0126, 0.0207, 0.0185, 0.0122, 0.0192, 0.0184, 0.0178, 0.0125], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-17 03:04:06,878 INFO [finetune.py:992] (0/2) Epoch 15, batch 4450, loss[loss=0.1605, simple_loss=0.2442, pruned_loss=0.03838, over 12253.00 frames. ], tot_loss[loss=0.1651, simple_loss=0.254, pruned_loss=0.03815, over 2357452.90 frames. ], batch size: 32, lr: 3.50e-03, grad_scale: 16.0 2023-05-17 03:04:09,817 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.1958, 3.9697, 3.9299, 4.0894, 4.1260, 4.2194, 3.9877, 2.3486], device='cuda:0'), covar=tensor([0.0175, 0.0130, 0.0195, 0.0132, 0.0081, 0.0189, 0.0150, 0.1076], device='cuda:0'), in_proj_covar=tensor([0.0071, 0.0081, 0.0084, 0.0075, 0.0062, 0.0094, 0.0084, 0.0099], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-17 03:04:17,559 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=275741.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 03:04:25,491 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.1237, 5.0038, 4.8967, 4.9687, 4.6382, 5.0513, 5.1098, 5.2436], device='cuda:0'), covar=tensor([0.0201, 0.0130, 0.0177, 0.0333, 0.0690, 0.0294, 0.0129, 0.0174], device='cuda:0'), in_proj_covar=tensor([0.0199, 0.0202, 0.0196, 0.0256, 0.0246, 0.0225, 0.0183, 0.0234], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0005, 0.0004, 0.0004, 0.0003, 0.0005], device='cuda:0') 2023-05-17 03:04:34,279 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.46 vs. limit=2.0 2023-05-17 03:04:37,312 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.37 vs. limit=2.0 2023-05-17 03:04:40,187 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.39 vs. limit=2.0 2023-05-17 03:04:43,217 INFO [finetune.py:992] (0/2) Epoch 15, batch 4500, loss[loss=0.1509, simple_loss=0.2473, pruned_loss=0.02722, over 12280.00 frames. ], tot_loss[loss=0.1656, simple_loss=0.2545, pruned_loss=0.03834, over 2353355.62 frames. ], batch size: 37, lr: 3.50e-03, grad_scale: 16.0 2023-05-17 03:04:48,940 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.793e+02 2.591e+02 3.128e+02 3.902e+02 6.297e+02, threshold=6.256e+02, percent-clipped=0.0 2023-05-17 03:05:08,268 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.7664, 5.7147, 5.5061, 4.9998, 5.0163, 5.6604, 5.2554, 5.0248], device='cuda:0'), covar=tensor([0.0806, 0.0997, 0.0785, 0.1730, 0.0907, 0.0748, 0.1611, 0.1145], device='cuda:0'), in_proj_covar=tensor([0.0643, 0.0574, 0.0530, 0.0652, 0.0430, 0.0743, 0.0801, 0.0586], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0004, 0.0004, 0.0003], device='cuda:0') 2023-05-17 03:05:11,184 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=275814.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 03:05:20,519 INFO [finetune.py:992] (0/2) Epoch 15, batch 4550, loss[loss=0.1554, simple_loss=0.2477, pruned_loss=0.03154, over 12150.00 frames. ], tot_loss[loss=0.1654, simple_loss=0.2546, pruned_loss=0.03812, over 2360729.68 frames. ], batch size: 34, lr: 3.50e-03, grad_scale: 16.0 2023-05-17 03:05:22,138 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.3292, 2.4147, 3.6901, 4.1932, 3.7328, 4.3053, 3.7376, 2.9952], device='cuda:0'), covar=tensor([0.0042, 0.0443, 0.0129, 0.0059, 0.0127, 0.0080, 0.0161, 0.0383], device='cuda:0'), in_proj_covar=tensor([0.0092, 0.0124, 0.0106, 0.0079, 0.0108, 0.0117, 0.0101, 0.0141], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-17 03:05:44,025 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=275860.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 03:05:44,319 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.61 vs. limit=2.0 2023-05-17 03:05:45,346 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=275862.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 03:05:55,993 INFO [finetune.py:992] (0/2) Epoch 15, batch 4600, loss[loss=0.1632, simple_loss=0.2586, pruned_loss=0.03385, over 12114.00 frames. ], tot_loss[loss=0.1661, simple_loss=0.2554, pruned_loss=0.03843, over 2360077.37 frames. ], batch size: 38, lr: 3.50e-03, grad_scale: 16.0 2023-05-17 03:06:02,265 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.883e+02 2.630e+02 3.077e+02 3.861e+02 7.958e+02, threshold=6.155e+02, percent-clipped=3.0 2023-05-17 03:06:02,435 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.6173, 4.3312, 4.6221, 4.0472, 4.3544, 4.0941, 4.6246, 4.3533], device='cuda:0'), covar=tensor([0.0359, 0.0388, 0.0332, 0.0322, 0.0425, 0.0393, 0.0244, 0.0585], device='cuda:0'), in_proj_covar=tensor([0.0270, 0.0273, 0.0299, 0.0270, 0.0275, 0.0272, 0.0245, 0.0218], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-17 03:06:28,046 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=275921.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 03:06:32,369 INFO [finetune.py:992] (0/2) Epoch 15, batch 4650, loss[loss=0.1334, simple_loss=0.2131, pruned_loss=0.02686, over 12346.00 frames. ], tot_loss[loss=0.1658, simple_loss=0.2551, pruned_loss=0.03829, over 2362052.40 frames. ], batch size: 30, lr: 3.50e-03, grad_scale: 16.0 2023-05-17 03:06:36,897 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.4675, 5.2111, 5.4421, 5.4123, 4.9688, 5.0995, 4.8620, 5.2958], device='cuda:0'), covar=tensor([0.0793, 0.0671, 0.0864, 0.0619, 0.2204, 0.1350, 0.0589, 0.1261], device='cuda:0'), in_proj_covar=tensor([0.0559, 0.0732, 0.0639, 0.0656, 0.0884, 0.0776, 0.0590, 0.0499], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0005, 0.0004, 0.0004, 0.0005, 0.0005, 0.0004, 0.0003], device='cuda:0') 2023-05-17 03:06:51,226 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=275952.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 03:06:58,854 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=275963.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 03:07:08,579 INFO [finetune.py:992] (0/2) Epoch 15, batch 4700, loss[loss=0.1581, simple_loss=0.2505, pruned_loss=0.03288, over 12323.00 frames. ], tot_loss[loss=0.1656, simple_loss=0.2552, pruned_loss=0.038, over 2370970.64 frames. ], batch size: 34, lr: 3.50e-03, grad_scale: 16.0 2023-05-17 03:07:14,349 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.801e+02 2.857e+02 3.354e+02 4.097e+02 8.987e+02, threshold=6.708e+02, percent-clipped=2.0 2023-05-17 03:07:25,632 INFO [checkpoint.py:75] (0/2) Saving checkpoint to pruned_transducer_stateless7/exp_giga_finetune/checkpoint-176000.pt 2023-05-17 03:07:31,000 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=276003.0, num_to_drop=1, layers_to_drop={1} 2023-05-17 03:07:31,793 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.4813, 4.8545, 3.1176, 2.7638, 4.1602, 2.9727, 4.0332, 3.5659], device='cuda:0'), covar=tensor([0.0688, 0.0587, 0.1106, 0.1522, 0.0332, 0.1234, 0.0535, 0.0709], device='cuda:0'), in_proj_covar=tensor([0.0193, 0.0262, 0.0180, 0.0203, 0.0144, 0.0186, 0.0200, 0.0177], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-17 03:07:35,922 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.9487, 3.4006, 5.3136, 2.7135, 2.8663, 3.9135, 3.5597, 3.8460], device='cuda:0'), covar=tensor([0.0467, 0.1275, 0.0307, 0.1313, 0.2124, 0.1582, 0.1223, 0.1252], device='cuda:0'), in_proj_covar=tensor([0.0239, 0.0240, 0.0257, 0.0186, 0.0239, 0.0296, 0.0227, 0.0267], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-17 03:07:36,431 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=276011.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 03:07:38,062 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=276013.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 03:07:48,316 INFO [finetune.py:992] (0/2) Epoch 15, batch 4750, loss[loss=0.1587, simple_loss=0.2477, pruned_loss=0.03485, over 12113.00 frames. ], tot_loss[loss=0.1653, simple_loss=0.2546, pruned_loss=0.03803, over 2366969.35 frames. ], batch size: 38, lr: 3.50e-03, grad_scale: 16.0 2023-05-17 03:07:58,509 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=276041.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 03:08:05,529 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=276051.0, num_to_drop=1, layers_to_drop={1} 2023-05-17 03:08:08,026 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.37 vs. limit=2.0 2023-05-17 03:08:21,646 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.38 vs. limit=2.0 2023-05-17 03:08:23,960 INFO [finetune.py:992] (0/2) Epoch 15, batch 4800, loss[loss=0.1608, simple_loss=0.2578, pruned_loss=0.03184, over 11749.00 frames. ], tot_loss[loss=0.1657, simple_loss=0.2548, pruned_loss=0.03833, over 2360134.76 frames. ], batch size: 44, lr: 3.50e-03, grad_scale: 16.0 2023-05-17 03:08:30,242 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.960e+02 2.861e+02 3.478e+02 4.262e+02 1.522e+03, threshold=6.956e+02, percent-clipped=2.0 2023-05-17 03:08:33,194 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=276089.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 03:08:41,921 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.6993, 2.8201, 3.3774, 4.4845, 2.3187, 4.6620, 4.6734, 4.7049], device='cuda:0'), covar=tensor([0.0110, 0.1215, 0.0445, 0.0166, 0.1370, 0.0204, 0.0144, 0.0107], device='cuda:0'), in_proj_covar=tensor([0.0125, 0.0205, 0.0184, 0.0122, 0.0191, 0.0183, 0.0178, 0.0124], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-17 03:09:00,429 INFO [finetune.py:992] (0/2) Epoch 15, batch 4850, loss[loss=0.1596, simple_loss=0.254, pruned_loss=0.03256, over 12114.00 frames. ], tot_loss[loss=0.1659, simple_loss=0.255, pruned_loss=0.03841, over 2362559.85 frames. ], batch size: 33, lr: 3.50e-03, grad_scale: 16.0 2023-05-17 03:09:36,887 INFO [finetune.py:992] (0/2) Epoch 15, batch 4900, loss[loss=0.1496, simple_loss=0.2415, pruned_loss=0.02888, over 12121.00 frames. ], tot_loss[loss=0.1662, simple_loss=0.2552, pruned_loss=0.03859, over 2352615.75 frames. ], batch size: 33, lr: 3.50e-03, grad_scale: 16.0 2023-05-17 03:09:42,528 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.073e+02 2.675e+02 3.087e+02 3.649e+02 6.820e+02, threshold=6.174e+02, percent-clipped=0.0 2023-05-17 03:10:05,206 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=276216.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 03:10:13,651 INFO [finetune.py:992] (0/2) Epoch 15, batch 4950, loss[loss=0.1676, simple_loss=0.2554, pruned_loss=0.03994, over 12366.00 frames. ], tot_loss[loss=0.1657, simple_loss=0.2549, pruned_loss=0.03822, over 2363482.18 frames. ], batch size: 36, lr: 3.50e-03, grad_scale: 16.0 2023-05-17 03:10:21,203 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.37 vs. limit=2.0 2023-05-17 03:10:49,085 INFO [finetune.py:992] (0/2) Epoch 15, batch 5000, loss[loss=0.1569, simple_loss=0.25, pruned_loss=0.03189, over 12351.00 frames. ], tot_loss[loss=0.1656, simple_loss=0.255, pruned_loss=0.03814, over 2357913.32 frames. ], batch size: 35, lr: 3.50e-03, grad_scale: 16.0 2023-05-17 03:10:54,695 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.563e+02 2.604e+02 3.078e+02 3.726e+02 1.471e+03, threshold=6.157e+02, percent-clipped=5.0 2023-05-17 03:11:11,865 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=276308.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 03:11:18,262 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=276317.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 03:11:25,116 INFO [finetune.py:992] (0/2) Epoch 15, batch 5050, loss[loss=0.1741, simple_loss=0.2614, pruned_loss=0.0434, over 12082.00 frames. ], tot_loss[loss=0.1659, simple_loss=0.2555, pruned_loss=0.03815, over 2356110.46 frames. ], batch size: 32, lr: 3.49e-03, grad_scale: 16.0 2023-05-17 03:11:29,596 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.3610, 4.8935, 5.3270, 4.6203, 4.9631, 4.6686, 5.3878, 5.0429], device='cuda:0'), covar=tensor([0.0311, 0.0450, 0.0336, 0.0290, 0.0480, 0.0358, 0.0222, 0.0317], device='cuda:0'), in_proj_covar=tensor([0.0269, 0.0274, 0.0298, 0.0270, 0.0274, 0.0272, 0.0245, 0.0218], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-17 03:11:54,583 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.2267, 4.6445, 4.0696, 4.9008, 4.4947, 2.8445, 4.0669, 2.9505], device='cuda:0'), covar=tensor([0.0842, 0.0775, 0.1623, 0.0551, 0.1177, 0.1772, 0.1226, 0.3519], device='cuda:0'), in_proj_covar=tensor([0.0308, 0.0381, 0.0360, 0.0325, 0.0368, 0.0275, 0.0347, 0.0368], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-17 03:12:01,390 INFO [finetune.py:992] (0/2) Epoch 15, batch 5100, loss[loss=0.1668, simple_loss=0.2514, pruned_loss=0.04107, over 10617.00 frames. ], tot_loss[loss=0.1655, simple_loss=0.2549, pruned_loss=0.03806, over 2361993.71 frames. ], batch size: 68, lr: 3.49e-03, grad_scale: 16.0 2023-05-17 03:12:02,218 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=276378.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 03:12:06,900 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.891e+02 2.657e+02 3.057e+02 3.698e+02 1.370e+03, threshold=6.113e+02, percent-clipped=2.0 2023-05-17 03:12:07,833 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.0320, 4.6158, 4.8636, 4.8622, 4.8132, 4.8806, 4.7465, 2.6633], device='cuda:0'), covar=tensor([0.0169, 0.0108, 0.0131, 0.0123, 0.0068, 0.0216, 0.0116, 0.0936], device='cuda:0'), in_proj_covar=tensor([0.0071, 0.0081, 0.0085, 0.0076, 0.0062, 0.0096, 0.0085, 0.0100], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-17 03:12:37,396 INFO [finetune.py:992] (0/2) Epoch 15, batch 5150, loss[loss=0.1562, simple_loss=0.2451, pruned_loss=0.03364, over 12346.00 frames. ], tot_loss[loss=0.1649, simple_loss=0.2543, pruned_loss=0.03773, over 2372932.83 frames. ], batch size: 31, lr: 3.49e-03, grad_scale: 16.0 2023-05-17 03:13:13,941 INFO [finetune.py:992] (0/2) Epoch 15, batch 5200, loss[loss=0.1392, simple_loss=0.2217, pruned_loss=0.02838, over 12168.00 frames. ], tot_loss[loss=0.1652, simple_loss=0.2546, pruned_loss=0.03794, over 2362251.62 frames. ], batch size: 29, lr: 3.49e-03, grad_scale: 16.0 2023-05-17 03:13:19,583 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.834e+02 2.618e+02 2.916e+02 3.289e+02 6.315e+02, threshold=5.832e+02, percent-clipped=1.0 2023-05-17 03:13:30,709 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.8981, 2.9952, 4.8285, 4.9143, 2.9669, 2.7326, 3.1296, 2.3500], device='cuda:0'), covar=tensor([0.1624, 0.2956, 0.0414, 0.0400, 0.1363, 0.2464, 0.2815, 0.3948], device='cuda:0'), in_proj_covar=tensor([0.0302, 0.0393, 0.0281, 0.0306, 0.0277, 0.0317, 0.0395, 0.0379], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-17 03:13:35,050 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.8531, 2.9591, 4.7544, 4.7957, 2.8822, 2.6756, 3.0731, 2.3094], device='cuda:0'), covar=tensor([0.1635, 0.2862, 0.0391, 0.0408, 0.1374, 0.2476, 0.2790, 0.3987], device='cuda:0'), in_proj_covar=tensor([0.0303, 0.0393, 0.0281, 0.0306, 0.0277, 0.0317, 0.0396, 0.0379], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-17 03:13:42,121 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=276516.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 03:13:50,500 INFO [finetune.py:992] (0/2) Epoch 15, batch 5250, loss[loss=0.1451, simple_loss=0.2365, pruned_loss=0.02688, over 12026.00 frames. ], tot_loss[loss=0.1647, simple_loss=0.2542, pruned_loss=0.03759, over 2361176.48 frames. ], batch size: 31, lr: 3.49e-03, grad_scale: 16.0 2023-05-17 03:14:04,819 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=3.15 vs. limit=5.0 2023-05-17 03:14:11,137 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=4.30 vs. limit=5.0 2023-05-17 03:14:16,600 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=276564.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 03:14:25,995 INFO [finetune.py:992] (0/2) Epoch 15, batch 5300, loss[loss=0.1453, simple_loss=0.2274, pruned_loss=0.03158, over 12186.00 frames. ], tot_loss[loss=0.1641, simple_loss=0.2537, pruned_loss=0.03723, over 2372095.87 frames. ], batch size: 29, lr: 3.49e-03, grad_scale: 16.0 2023-05-17 03:14:31,808 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.897e+02 2.612e+02 3.140e+02 3.525e+02 6.336e+02, threshold=6.281e+02, percent-clipped=1.0 2023-05-17 03:14:49,510 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=276608.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 03:14:51,758 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=276611.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 03:15:02,959 INFO [finetune.py:992] (0/2) Epoch 15, batch 5350, loss[loss=0.1548, simple_loss=0.2411, pruned_loss=0.03427, over 12283.00 frames. ], tot_loss[loss=0.1643, simple_loss=0.2541, pruned_loss=0.03729, over 2370473.39 frames. ], batch size: 33, lr: 3.49e-03, grad_scale: 16.0 2023-05-17 03:15:22,486 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.2577, 4.8609, 5.2559, 4.5591, 4.9168, 4.6463, 5.2944, 5.0649], device='cuda:0'), covar=tensor([0.0286, 0.0400, 0.0314, 0.0278, 0.0418, 0.0319, 0.0217, 0.0326], device='cuda:0'), in_proj_covar=tensor([0.0268, 0.0274, 0.0298, 0.0269, 0.0273, 0.0269, 0.0245, 0.0217], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-17 03:15:23,782 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=276656.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 03:15:33,382 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=276668.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 03:15:36,160 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=276672.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 03:15:36,685 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=276673.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 03:15:39,301 INFO [finetune.py:992] (0/2) Epoch 15, batch 5400, loss[loss=0.1735, simple_loss=0.2685, pruned_loss=0.03932, over 11537.00 frames. ], tot_loss[loss=0.1645, simple_loss=0.2542, pruned_loss=0.03739, over 2374285.65 frames. ], batch size: 48, lr: 3.49e-03, grad_scale: 16.0 2023-05-17 03:15:44,947 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.816e+02 2.736e+02 3.127e+02 3.839e+02 6.677e+02, threshold=6.254e+02, percent-clipped=1.0 2023-05-17 03:15:55,845 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.3386, 4.9631, 5.1997, 5.2375, 5.0297, 5.1969, 5.0963, 3.0757], device='cuda:0'), covar=tensor([0.0077, 0.0053, 0.0063, 0.0047, 0.0040, 0.0090, 0.0067, 0.0591], device='cuda:0'), in_proj_covar=tensor([0.0072, 0.0081, 0.0085, 0.0076, 0.0062, 0.0096, 0.0085, 0.0100], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-17 03:16:03,018 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.2728, 4.5623, 4.0207, 4.9015, 4.5341, 3.0092, 4.1329, 2.9289], device='cuda:0'), covar=tensor([0.0833, 0.0913, 0.1435, 0.0488, 0.1134, 0.1651, 0.1107, 0.3704], device='cuda:0'), in_proj_covar=tensor([0.0308, 0.0383, 0.0361, 0.0327, 0.0370, 0.0275, 0.0349, 0.0369], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-17 03:16:14,457 INFO [finetune.py:992] (0/2) Epoch 15, batch 5450, loss[loss=0.1829, simple_loss=0.2789, pruned_loss=0.04346, over 12146.00 frames. ], tot_loss[loss=0.1649, simple_loss=0.2545, pruned_loss=0.03767, over 2371268.18 frames. ], batch size: 34, lr: 3.49e-03, grad_scale: 16.0 2023-05-17 03:16:16,029 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=276729.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 03:16:50,371 INFO [finetune.py:992] (0/2) Epoch 15, batch 5500, loss[loss=0.1844, simple_loss=0.2681, pruned_loss=0.05031, over 12107.00 frames. ], tot_loss[loss=0.1652, simple_loss=0.2548, pruned_loss=0.03779, over 2373091.69 frames. ], batch size: 38, lr: 3.49e-03, grad_scale: 16.0 2023-05-17 03:16:56,046 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.612e+02 2.745e+02 3.154e+02 3.777e+02 7.064e+02, threshold=6.307e+02, percent-clipped=2.0 2023-05-17 03:16:59,832 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.9426, 3.9058, 3.9742, 4.0337, 3.7771, 3.8082, 3.6851, 3.9231], device='cuda:0'), covar=tensor([0.1380, 0.0810, 0.1519, 0.0702, 0.1768, 0.1431, 0.0696, 0.1008], device='cuda:0'), in_proj_covar=tensor([0.0561, 0.0733, 0.0641, 0.0657, 0.0880, 0.0775, 0.0591, 0.0502], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0005, 0.0004, 0.0004, 0.0005, 0.0005, 0.0004, 0.0003], device='cuda:0') 2023-05-17 03:16:59,867 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=276790.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 03:17:04,167 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.65 vs. limit=2.0 2023-05-17 03:17:23,322 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.6672, 2.8439, 5.2439, 2.6072, 2.6141, 4.2269, 3.1471, 4.1962], device='cuda:0'), covar=tensor([0.0481, 0.1623, 0.0345, 0.1396, 0.2151, 0.1210, 0.1494, 0.0979], device='cuda:0'), in_proj_covar=tensor([0.0238, 0.0239, 0.0256, 0.0184, 0.0238, 0.0295, 0.0226, 0.0268], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-17 03:17:26,679 INFO [finetune.py:992] (0/2) Epoch 15, batch 5550, loss[loss=0.1832, simple_loss=0.2706, pruned_loss=0.04791, over 12362.00 frames. ], tot_loss[loss=0.166, simple_loss=0.2553, pruned_loss=0.03835, over 2368184.94 frames. ], batch size: 38, lr: 3.49e-03, grad_scale: 16.0 2023-05-17 03:17:43,768 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.5217, 2.6535, 3.7941, 4.5204, 3.9063, 4.6060, 3.8775, 3.2288], device='cuda:0'), covar=tensor([0.0043, 0.0377, 0.0137, 0.0041, 0.0125, 0.0066, 0.0131, 0.0347], device='cuda:0'), in_proj_covar=tensor([0.0091, 0.0124, 0.0106, 0.0079, 0.0109, 0.0117, 0.0101, 0.0141], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-17 03:17:43,799 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=276851.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 03:17:52,581 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=276863.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 03:18:02,414 INFO [finetune.py:992] (0/2) Epoch 15, batch 5600, loss[loss=0.1688, simple_loss=0.2619, pruned_loss=0.03787, over 12115.00 frames. ], tot_loss[loss=0.1655, simple_loss=0.2549, pruned_loss=0.03804, over 2375022.36 frames. ], batch size: 33, lr: 3.49e-03, grad_scale: 16.0 2023-05-17 03:18:08,764 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.820e+02 2.770e+02 3.218e+02 3.855e+02 6.712e+02, threshold=6.436e+02, percent-clipped=1.0 2023-05-17 03:18:37,009 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=276924.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 03:18:37,647 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.5987, 2.5120, 3.8306, 4.6375, 4.0079, 4.6388, 3.9560, 3.2683], device='cuda:0'), covar=tensor([0.0042, 0.0441, 0.0129, 0.0033, 0.0131, 0.0067, 0.0130, 0.0387], device='cuda:0'), in_proj_covar=tensor([0.0092, 0.0125, 0.0107, 0.0080, 0.0109, 0.0118, 0.0101, 0.0143], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-17 03:18:38,817 INFO [finetune.py:992] (0/2) Epoch 15, batch 5650, loss[loss=0.1839, simple_loss=0.2732, pruned_loss=0.04734, over 12343.00 frames. ], tot_loss[loss=0.166, simple_loss=0.2554, pruned_loss=0.0383, over 2365694.49 frames. ], batch size: 36, lr: 3.49e-03, grad_scale: 32.0 2023-05-17 03:18:41,743 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=276931.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 03:19:07,761 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=276967.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 03:19:12,037 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=276973.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 03:19:14,634 INFO [finetune.py:992] (0/2) Epoch 15, batch 5700, loss[loss=0.1575, simple_loss=0.2465, pruned_loss=0.03428, over 12105.00 frames. ], tot_loss[loss=0.1658, simple_loss=0.2552, pruned_loss=0.03821, over 2368555.13 frames. ], batch size: 32, lr: 3.49e-03, grad_scale: 32.0 2023-05-17 03:19:17,865 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.91 vs. limit=2.0 2023-05-17 03:19:20,244 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.920e+02 2.620e+02 3.127e+02 4.121e+02 6.231e+02, threshold=6.254e+02, percent-clipped=0.0 2023-05-17 03:19:25,529 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=276992.0, num_to_drop=1, layers_to_drop={1} 2023-05-17 03:19:39,413 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=277011.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 03:19:46,444 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=277021.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 03:19:48,482 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=277024.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 03:19:50,577 INFO [finetune.py:992] (0/2) Epoch 15, batch 5750, loss[loss=0.1715, simple_loss=0.266, pruned_loss=0.03856, over 12346.00 frames. ], tot_loss[loss=0.1658, simple_loss=0.2554, pruned_loss=0.03816, over 2368229.71 frames. ], batch size: 36, lr: 3.49e-03, grad_scale: 32.0 2023-05-17 03:20:23,209 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=277072.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 03:20:26,342 INFO [finetune.py:992] (0/2) Epoch 15, batch 5800, loss[loss=0.1826, simple_loss=0.2833, pruned_loss=0.041, over 10466.00 frames. ], tot_loss[loss=0.1664, simple_loss=0.2561, pruned_loss=0.03841, over 2366063.94 frames. ], batch size: 68, lr: 3.49e-03, grad_scale: 32.0 2023-05-17 03:20:31,929 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.611e+02 2.608e+02 3.112e+02 3.705e+02 8.677e+02, threshold=6.225e+02, percent-clipped=5.0 2023-05-17 03:21:02,532 INFO [finetune.py:992] (0/2) Epoch 15, batch 5850, loss[loss=0.1821, simple_loss=0.27, pruned_loss=0.0471, over 12360.00 frames. ], tot_loss[loss=0.1663, simple_loss=0.2557, pruned_loss=0.03847, over 2364410.55 frames. ], batch size: 36, lr: 3.49e-03, grad_scale: 32.0 2023-05-17 03:21:16,043 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=277146.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 03:21:38,768 INFO [finetune.py:992] (0/2) Epoch 15, batch 5900, loss[loss=0.1432, simple_loss=0.2279, pruned_loss=0.02931, over 11828.00 frames. ], tot_loss[loss=0.1658, simple_loss=0.2554, pruned_loss=0.03814, over 2370995.63 frames. ], batch size: 26, lr: 3.49e-03, grad_scale: 32.0 2023-05-17 03:21:38,951 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=277177.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 03:21:44,304 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.759e+02 2.580e+02 3.106e+02 3.557e+02 8.858e+02, threshold=6.212e+02, percent-clipped=2.0 2023-05-17 03:21:46,895 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-05-17 03:22:09,444 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=277219.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 03:22:09,528 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=277219.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 03:22:15,091 INFO [finetune.py:992] (0/2) Epoch 15, batch 5950, loss[loss=0.1604, simple_loss=0.2588, pruned_loss=0.03103, over 12304.00 frames. ], tot_loss[loss=0.1654, simple_loss=0.2551, pruned_loss=0.03789, over 2375044.35 frames. ], batch size: 34, lr: 3.49e-03, grad_scale: 32.0 2023-05-17 03:22:15,243 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.2700, 5.0964, 5.2172, 5.2556, 4.8644, 4.9129, 4.7342, 5.2205], device='cuda:0'), covar=tensor([0.0794, 0.0692, 0.0960, 0.0626, 0.2069, 0.1509, 0.0523, 0.1033], device='cuda:0'), in_proj_covar=tensor([0.0561, 0.0730, 0.0640, 0.0651, 0.0879, 0.0773, 0.0588, 0.0499], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0005, 0.0004, 0.0004, 0.0005, 0.0005, 0.0004, 0.0003], device='cuda:0') 2023-05-17 03:22:23,065 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=277238.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 03:22:44,500 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=277267.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 03:22:51,369 INFO [finetune.py:992] (0/2) Epoch 15, batch 6000, loss[loss=0.1468, simple_loss=0.2361, pruned_loss=0.02872, over 12273.00 frames. ], tot_loss[loss=0.1653, simple_loss=0.2549, pruned_loss=0.03781, over 2384720.01 frames. ], batch size: 33, lr: 3.49e-03, grad_scale: 32.0 2023-05-17 03:22:51,370 INFO [finetune.py:1017] (0/2) Computing validation loss 2023-05-17 03:23:00,369 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.6285, 2.0498, 2.8225, 3.6061, 2.0313, 3.7166, 3.1199, 3.6670], device='cuda:0'), covar=tensor([0.0129, 0.1332, 0.0498, 0.0146, 0.1339, 0.0198, 0.0403, 0.0135], device='cuda:0'), in_proj_covar=tensor([0.0125, 0.0208, 0.0186, 0.0124, 0.0195, 0.0185, 0.0180, 0.0127], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-17 03:23:09,832 INFO [finetune.py:1026] (0/2) Epoch 15, validation: loss=0.3187, simple_loss=0.3936, pruned_loss=0.1219, over 1020973.00 frames. 2023-05-17 03:23:09,833 INFO [finetune.py:1027] (0/2) Maximum memory allocated so far is 12827MB 2023-05-17 03:23:11,383 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.7472, 4.1411, 3.7178, 4.4675, 4.0261, 2.6805, 3.7103, 2.7338], device='cuda:0'), covar=tensor([0.1099, 0.0989, 0.1649, 0.0635, 0.1383, 0.1867, 0.1310, 0.3938], device='cuda:0'), in_proj_covar=tensor([0.0313, 0.0388, 0.0367, 0.0331, 0.0376, 0.0279, 0.0353, 0.0373], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-17 03:23:12,044 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=277280.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 03:23:15,956 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.739e+02 2.528e+02 2.885e+02 3.414e+02 8.880e+02, threshold=5.771e+02, percent-clipped=2.0 2023-05-17 03:23:17,451 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=277287.0, num_to_drop=1, layers_to_drop={2} 2023-05-17 03:23:30,638 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=277305.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 03:23:37,657 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=277315.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 03:23:44,192 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=277324.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 03:23:46,205 INFO [finetune.py:992] (0/2) Epoch 15, batch 6050, loss[loss=0.1504, simple_loss=0.2463, pruned_loss=0.02723, over 12145.00 frames. ], tot_loss[loss=0.1659, simple_loss=0.2556, pruned_loss=0.03808, over 2375708.09 frames. ], batch size: 34, lr: 3.49e-03, grad_scale: 32.0 2023-05-17 03:24:14,381 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=277366.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 03:24:14,919 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=277367.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 03:24:18,480 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=277372.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 03:24:21,949 INFO [finetune.py:992] (0/2) Epoch 15, batch 6100, loss[loss=0.1952, simple_loss=0.2834, pruned_loss=0.05355, over 12100.00 frames. ], tot_loss[loss=0.1664, simple_loss=0.256, pruned_loss=0.03835, over 2373593.87 frames. ], batch size: 42, lr: 3.49e-03, grad_scale: 32.0 2023-05-17 03:24:23,608 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.1455, 2.4772, 3.7079, 3.1234, 3.4785, 3.2065, 2.4925, 3.5575], device='cuda:0'), covar=tensor([0.0143, 0.0389, 0.0151, 0.0268, 0.0160, 0.0196, 0.0393, 0.0155], device='cuda:0'), in_proj_covar=tensor([0.0182, 0.0206, 0.0194, 0.0189, 0.0221, 0.0169, 0.0199, 0.0190], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-17 03:24:27,652 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.956e+02 2.661e+02 3.167e+02 3.859e+02 7.157e+02, threshold=6.333e+02, percent-clipped=3.0 2023-05-17 03:24:31,211 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.77 vs. limit=2.0 2023-05-17 03:24:39,535 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.6159, 5.4780, 5.5969, 5.6421, 5.2315, 5.3117, 5.0911, 5.6336], device='cuda:0'), covar=tensor([0.0756, 0.0590, 0.0693, 0.0515, 0.1996, 0.1181, 0.0507, 0.0729], device='cuda:0'), in_proj_covar=tensor([0.0566, 0.0729, 0.0640, 0.0654, 0.0883, 0.0774, 0.0589, 0.0499], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0005, 0.0004, 0.0004, 0.0005, 0.0005, 0.0004, 0.0003], device='cuda:0') 2023-05-17 03:24:49,896 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.12 vs. limit=2.0 2023-05-17 03:24:52,178 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=4.73 vs. limit=5.0 2023-05-17 03:24:55,299 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.2849, 5.2051, 5.0489, 5.1630, 4.8210, 5.2625, 5.2413, 5.3748], device='cuda:0'), covar=tensor([0.0188, 0.0126, 0.0170, 0.0258, 0.0659, 0.0281, 0.0135, 0.0146], device='cuda:0'), in_proj_covar=tensor([0.0201, 0.0204, 0.0199, 0.0257, 0.0249, 0.0227, 0.0184, 0.0237], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0005, 0.0004, 0.0004, 0.0003, 0.0005], device='cuda:0') 2023-05-17 03:24:58,111 INFO [finetune.py:992] (0/2) Epoch 15, batch 6150, loss[loss=0.1663, simple_loss=0.2626, pruned_loss=0.03501, over 12052.00 frames. ], tot_loss[loss=0.1675, simple_loss=0.2572, pruned_loss=0.03891, over 2372004.92 frames. ], batch size: 40, lr: 3.49e-03, grad_scale: 32.0 2023-05-17 03:25:12,052 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=277446.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 03:25:15,707 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-05-17 03:25:20,688 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([1.9812, 3.8105, 4.0001, 4.4043, 2.7001, 3.6399, 2.5542, 3.8953], device='cuda:0'), covar=tensor([0.1861, 0.0869, 0.0881, 0.0663, 0.1467, 0.0797, 0.1999, 0.1236], device='cuda:0'), in_proj_covar=tensor([0.0229, 0.0268, 0.0301, 0.0362, 0.0244, 0.0247, 0.0263, 0.0379], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-17 03:25:33,879 INFO [finetune.py:992] (0/2) Epoch 15, batch 6200, loss[loss=0.1414, simple_loss=0.2254, pruned_loss=0.02869, over 12171.00 frames. ], tot_loss[loss=0.1677, simple_loss=0.2573, pruned_loss=0.0391, over 2373809.87 frames. ], batch size: 29, lr: 3.49e-03, grad_scale: 32.0 2023-05-17 03:25:39,644 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.885e+02 2.720e+02 3.197e+02 3.558e+02 7.007e+02, threshold=6.393e+02, percent-clipped=2.0 2023-05-17 03:25:46,779 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=277494.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 03:25:47,698 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.2784, 4.6533, 4.0792, 4.9106, 4.5081, 2.9515, 4.2263, 3.1100], device='cuda:0'), covar=tensor([0.0888, 0.0757, 0.1618, 0.0642, 0.1339, 0.1720, 0.1048, 0.3290], device='cuda:0'), in_proj_covar=tensor([0.0310, 0.0384, 0.0365, 0.0328, 0.0373, 0.0276, 0.0351, 0.0369], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-17 03:26:04,887 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=277519.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 03:26:10,496 INFO [finetune.py:992] (0/2) Epoch 15, batch 6250, loss[loss=0.1871, simple_loss=0.2735, pruned_loss=0.0503, over 10453.00 frames. ], tot_loss[loss=0.1664, simple_loss=0.2563, pruned_loss=0.03827, over 2380191.04 frames. ], batch size: 69, lr: 3.49e-03, grad_scale: 32.0 2023-05-17 03:26:14,861 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=277533.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 03:26:29,876 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=277554.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 03:26:39,172 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=277567.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 03:26:41,411 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=277570.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 03:26:44,805 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=277575.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 03:26:46,187 INFO [finetune.py:992] (0/2) Epoch 15, batch 6300, loss[loss=0.1675, simple_loss=0.257, pruned_loss=0.03903, over 11854.00 frames. ], tot_loss[loss=0.1665, simple_loss=0.2563, pruned_loss=0.03831, over 2375466.82 frames. ], batch size: 44, lr: 3.49e-03, grad_scale: 32.0 2023-05-17 03:26:52,618 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.776e+02 2.629e+02 3.057e+02 3.498e+02 6.273e+02, threshold=6.115e+02, percent-clipped=0.0 2023-05-17 03:26:54,307 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=277587.0, num_to_drop=1, layers_to_drop={2} 2023-05-17 03:26:55,676 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.3231, 4.9288, 5.3177, 4.7424, 4.9946, 4.6702, 5.3339, 5.0871], device='cuda:0'), covar=tensor([0.0271, 0.0357, 0.0269, 0.0235, 0.0389, 0.0352, 0.0231, 0.0267], device='cuda:0'), in_proj_covar=tensor([0.0267, 0.0272, 0.0296, 0.0266, 0.0270, 0.0268, 0.0244, 0.0217], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-17 03:27:12,886 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.3082, 5.1270, 5.3195, 5.3237, 4.9824, 4.9946, 4.7216, 5.2948], device='cuda:0'), covar=tensor([0.0823, 0.0699, 0.0724, 0.0604, 0.1740, 0.1240, 0.0589, 0.0919], device='cuda:0'), in_proj_covar=tensor([0.0566, 0.0731, 0.0639, 0.0656, 0.0878, 0.0772, 0.0589, 0.0499], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0005, 0.0004, 0.0004, 0.0005, 0.0005, 0.0004, 0.0003], device='cuda:0') 2023-05-17 03:27:14,363 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=277615.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 03:27:22,492 INFO [finetune.py:992] (0/2) Epoch 15, batch 6350, loss[loss=0.1793, simple_loss=0.2749, pruned_loss=0.04188, over 12080.00 frames. ], tot_loss[loss=0.1675, simple_loss=0.2574, pruned_loss=0.03885, over 2355756.32 frames. ], batch size: 42, lr: 3.49e-03, grad_scale: 32.0 2023-05-17 03:27:26,103 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=277631.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 03:27:28,803 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=277635.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 03:27:47,796 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=277661.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 03:27:52,240 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=277667.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 03:27:59,278 INFO [finetune.py:992] (0/2) Epoch 15, batch 6400, loss[loss=0.1667, simple_loss=0.2581, pruned_loss=0.03768, over 11205.00 frames. ], tot_loss[loss=0.1665, simple_loss=0.2564, pruned_loss=0.03829, over 2361177.59 frames. ], batch size: 55, lr: 3.48e-03, grad_scale: 32.0 2023-05-17 03:28:04,861 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.508e+02 2.566e+02 2.970e+02 3.495e+02 7.369e+02, threshold=5.939e+02, percent-clipped=2.0 2023-05-17 03:28:09,238 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=277691.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 03:28:26,382 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=277715.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 03:28:35,617 INFO [finetune.py:992] (0/2) Epoch 15, batch 6450, loss[loss=0.1534, simple_loss=0.242, pruned_loss=0.03244, over 12247.00 frames. ], tot_loss[loss=0.1656, simple_loss=0.2555, pruned_loss=0.03785, over 2375260.99 frames. ], batch size: 32, lr: 3.48e-03, grad_scale: 32.0 2023-05-17 03:28:47,326 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.8503, 2.6327, 3.5837, 3.6164, 2.9524, 2.6648, 2.7154, 2.4364], device='cuda:0'), covar=tensor([0.1317, 0.2691, 0.0651, 0.0577, 0.0983, 0.2176, 0.2527, 0.3462], device='cuda:0'), in_proj_covar=tensor([0.0300, 0.0390, 0.0280, 0.0304, 0.0275, 0.0315, 0.0392, 0.0376], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-17 03:28:53,731 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=277752.0, num_to_drop=1, layers_to_drop={3} 2023-05-17 03:29:12,049 INFO [finetune.py:992] (0/2) Epoch 15, batch 6500, loss[loss=0.1513, simple_loss=0.2551, pruned_loss=0.02369, over 12111.00 frames. ], tot_loss[loss=0.1665, simple_loss=0.2562, pruned_loss=0.03838, over 2371669.81 frames. ], batch size: 33, lr: 3.48e-03, grad_scale: 32.0 2023-05-17 03:29:14,391 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.1715, 5.0709, 4.9273, 5.1019, 4.7206, 5.2679, 5.2570, 5.3229], device='cuda:0'), covar=tensor([0.0235, 0.0134, 0.0184, 0.0269, 0.0674, 0.0267, 0.0126, 0.0148], device='cuda:0'), in_proj_covar=tensor([0.0202, 0.0204, 0.0200, 0.0258, 0.0250, 0.0228, 0.0185, 0.0237], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0005, 0.0004, 0.0004, 0.0003, 0.0005], device='cuda:0') 2023-05-17 03:29:17,672 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.750e+02 2.648e+02 3.126e+02 3.804e+02 5.907e+02, threshold=6.252e+02, percent-clipped=0.0 2023-05-17 03:29:43,910 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.1041, 2.4723, 3.6402, 3.0590, 3.4791, 3.2632, 2.4920, 3.5544], device='cuda:0'), covar=tensor([0.0156, 0.0409, 0.0171, 0.0281, 0.0192, 0.0179, 0.0399, 0.0160], device='cuda:0'), in_proj_covar=tensor([0.0184, 0.0209, 0.0197, 0.0193, 0.0225, 0.0171, 0.0202, 0.0194], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-17 03:29:47,791 INFO [finetune.py:992] (0/2) Epoch 15, batch 6550, loss[loss=0.1664, simple_loss=0.268, pruned_loss=0.03242, over 12154.00 frames. ], tot_loss[loss=0.1672, simple_loss=0.2571, pruned_loss=0.0387, over 2370279.05 frames. ], batch size: 34, lr: 3.48e-03, grad_scale: 32.0 2023-05-17 03:29:52,105 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=277833.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 03:30:05,119 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.9774, 3.3559, 5.3139, 2.7973, 2.9052, 3.7953, 3.4450, 3.7360], device='cuda:0'), covar=tensor([0.0376, 0.1211, 0.0317, 0.1181, 0.1889, 0.1652, 0.1218, 0.1266], device='cuda:0'), in_proj_covar=tensor([0.0239, 0.0239, 0.0259, 0.0184, 0.0239, 0.0297, 0.0226, 0.0269], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-17 03:30:18,829 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.38 vs. limit=2.0 2023-05-17 03:30:22,603 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=277875.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 03:30:23,957 INFO [finetune.py:992] (0/2) Epoch 15, batch 6600, loss[loss=0.1522, simple_loss=0.2369, pruned_loss=0.03374, over 12286.00 frames. ], tot_loss[loss=0.1673, simple_loss=0.2572, pruned_loss=0.03872, over 2373200.22 frames. ], batch size: 33, lr: 3.48e-03, grad_scale: 32.0 2023-05-17 03:30:26,952 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=277881.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 03:30:29,661 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.932e+02 2.688e+02 3.224e+02 3.921e+02 8.533e+02, threshold=6.448e+02, percent-clipped=2.0 2023-05-17 03:30:47,650 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=277910.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 03:30:57,470 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=277923.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 03:30:59,569 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=277926.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 03:31:00,250 INFO [finetune.py:992] (0/2) Epoch 15, batch 6650, loss[loss=0.1632, simple_loss=0.2603, pruned_loss=0.03306, over 12187.00 frames. ], tot_loss[loss=0.167, simple_loss=0.2567, pruned_loss=0.03863, over 2376469.77 frames. ], batch size: 35, lr: 3.48e-03, grad_scale: 32.0 2023-05-17 03:31:01,777 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.3126, 5.1487, 4.9675, 5.0954, 4.8593, 5.2363, 5.2650, 5.3068], device='cuda:0'), covar=tensor([0.0176, 0.0132, 0.0173, 0.0332, 0.0634, 0.0250, 0.0129, 0.0178], device='cuda:0'), in_proj_covar=tensor([0.0202, 0.0205, 0.0199, 0.0258, 0.0250, 0.0228, 0.0185, 0.0238], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0005, 0.0004, 0.0004, 0.0003, 0.0005], device='cuda:0') 2023-05-17 03:31:01,808 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.7236, 2.9776, 3.8385, 4.7077, 4.0142, 4.7525, 4.1077, 3.6842], device='cuda:0'), covar=tensor([0.0037, 0.0358, 0.0155, 0.0034, 0.0122, 0.0070, 0.0108, 0.0295], device='cuda:0'), in_proj_covar=tensor([0.0091, 0.0125, 0.0107, 0.0080, 0.0109, 0.0117, 0.0100, 0.0141], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-17 03:31:17,439 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.7533, 4.5897, 4.5276, 4.6239, 4.2452, 4.7518, 4.7720, 4.7814], device='cuda:0'), covar=tensor([0.0241, 0.0180, 0.0217, 0.0431, 0.0808, 0.0372, 0.0171, 0.0256], device='cuda:0'), in_proj_covar=tensor([0.0201, 0.0204, 0.0199, 0.0258, 0.0250, 0.0228, 0.0185, 0.0237], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0005, 0.0004, 0.0004, 0.0003, 0.0005], device='cuda:0') 2023-05-17 03:31:21,615 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.2693, 5.1053, 4.9689, 5.0630, 4.8118, 5.2407, 5.2453, 5.3457], device='cuda:0'), covar=tensor([0.0232, 0.0134, 0.0182, 0.0325, 0.0678, 0.0329, 0.0118, 0.0160], device='cuda:0'), in_proj_covar=tensor([0.0202, 0.0204, 0.0199, 0.0258, 0.0250, 0.0228, 0.0185, 0.0237], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0005, 0.0004, 0.0004, 0.0003, 0.0005], device='cuda:0') 2023-05-17 03:31:24,586 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=277961.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 03:31:31,730 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=277971.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 03:31:35,919 INFO [finetune.py:992] (0/2) Epoch 15, batch 6700, loss[loss=0.1494, simple_loss=0.2362, pruned_loss=0.03132, over 12131.00 frames. ], tot_loss[loss=0.1666, simple_loss=0.2563, pruned_loss=0.03842, over 2374334.60 frames. ], batch size: 30, lr: 3.48e-03, grad_scale: 32.0 2023-05-17 03:31:41,604 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.948e+02 2.527e+02 3.046e+02 3.825e+02 1.241e+03, threshold=6.091e+02, percent-clipped=2.0 2023-05-17 03:31:52,858 INFO [checkpoint.py:75] (0/2) Saving checkpoint to pruned_transducer_stateless7/exp_giga_finetune/checkpoint-178000.pt 2023-05-17 03:32:01,651 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=4.80 vs. limit=5.0 2023-05-17 03:32:02,670 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=278009.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 03:32:15,937 INFO [finetune.py:992] (0/2) Epoch 15, batch 6750, loss[loss=0.1596, simple_loss=0.2434, pruned_loss=0.03792, over 12087.00 frames. ], tot_loss[loss=0.166, simple_loss=0.2557, pruned_loss=0.03811, over 2377237.65 frames. ], batch size: 32, lr: 3.48e-03, grad_scale: 32.0 2023-05-17 03:32:19,724 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=278032.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 03:32:20,406 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.4777, 2.2343, 3.1475, 4.2796, 2.0575, 4.3164, 4.4109, 4.5024], device='cuda:0'), covar=tensor([0.0090, 0.1452, 0.0540, 0.0181, 0.1509, 0.0260, 0.0141, 0.0099], device='cuda:0'), in_proj_covar=tensor([0.0126, 0.0210, 0.0188, 0.0125, 0.0196, 0.0186, 0.0183, 0.0128], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002], device='cuda:0') 2023-05-17 03:32:30,270 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=278047.0, num_to_drop=1, layers_to_drop={1} 2023-05-17 03:32:52,487 INFO [finetune.py:992] (0/2) Epoch 15, batch 6800, loss[loss=0.141, simple_loss=0.2233, pruned_loss=0.0293, over 12359.00 frames. ], tot_loss[loss=0.1659, simple_loss=0.2557, pruned_loss=0.038, over 2375063.19 frames. ], batch size: 30, lr: 3.48e-03, grad_scale: 32.0 2023-05-17 03:32:58,188 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.665e+02 2.542e+02 3.281e+02 4.073e+02 8.715e+02, threshold=6.562e+02, percent-clipped=2.0 2023-05-17 03:33:27,868 INFO [finetune.py:992] (0/2) Epoch 15, batch 6850, loss[loss=0.1471, simple_loss=0.2277, pruned_loss=0.03326, over 12302.00 frames. ], tot_loss[loss=0.1661, simple_loss=0.2556, pruned_loss=0.03828, over 2374765.93 frames. ], batch size: 28, lr: 3.48e-03, grad_scale: 32.0 2023-05-17 03:33:50,106 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.99 vs. limit=2.0 2023-05-17 03:33:52,425 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.3625, 4.0987, 4.1931, 4.2824, 4.2359, 4.3382, 4.2658, 2.5676], device='cuda:0'), covar=tensor([0.0106, 0.0098, 0.0104, 0.0074, 0.0059, 0.0098, 0.0090, 0.0761], device='cuda:0'), in_proj_covar=tensor([0.0072, 0.0082, 0.0085, 0.0075, 0.0063, 0.0095, 0.0085, 0.0100], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-17 03:34:04,483 INFO [finetune.py:992] (0/2) Epoch 15, batch 6900, loss[loss=0.1603, simple_loss=0.2451, pruned_loss=0.03771, over 11793.00 frames. ], tot_loss[loss=0.1666, simple_loss=0.2559, pruned_loss=0.03865, over 2369188.22 frames. ], batch size: 44, lr: 3.48e-03, grad_scale: 32.0 2023-05-17 03:34:10,168 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.907e+02 2.747e+02 3.327e+02 4.069e+02 6.730e+02, threshold=6.654e+02, percent-clipped=1.0 2023-05-17 03:34:27,645 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.5205, 5.1159, 5.5182, 4.8512, 5.1774, 4.9072, 5.5466, 5.2091], device='cuda:0'), covar=tensor([0.0243, 0.0347, 0.0254, 0.0224, 0.0370, 0.0337, 0.0198, 0.0215], device='cuda:0'), in_proj_covar=tensor([0.0269, 0.0276, 0.0298, 0.0268, 0.0273, 0.0271, 0.0246, 0.0219], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-17 03:34:29,021 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=278210.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 03:34:36,941 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.74 vs. limit=2.0 2023-05-17 03:34:40,260 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=278226.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 03:34:40,753 INFO [finetune.py:992] (0/2) Epoch 15, batch 6950, loss[loss=0.1547, simple_loss=0.2527, pruned_loss=0.02841, over 12349.00 frames. ], tot_loss[loss=0.1662, simple_loss=0.2555, pruned_loss=0.03847, over 2363446.68 frames. ], batch size: 36, lr: 3.48e-03, grad_scale: 32.0 2023-05-17 03:34:50,338 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.64 vs. limit=2.0 2023-05-17 03:35:02,921 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=278258.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 03:35:13,125 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.2677, 4.4433, 2.8225, 2.5122, 3.8299, 2.5851, 3.8376, 3.0238], device='cuda:0'), covar=tensor([0.0729, 0.0573, 0.1072, 0.1513, 0.0320, 0.1336, 0.0493, 0.0916], device='cuda:0'), in_proj_covar=tensor([0.0190, 0.0259, 0.0179, 0.0201, 0.0143, 0.0185, 0.0199, 0.0176], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-17 03:35:14,321 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=278274.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 03:35:16,398 INFO [finetune.py:992] (0/2) Epoch 15, batch 7000, loss[loss=0.1472, simple_loss=0.229, pruned_loss=0.03272, over 12355.00 frames. ], tot_loss[loss=0.1655, simple_loss=0.2551, pruned_loss=0.03796, over 2370470.61 frames. ], batch size: 30, lr: 3.48e-03, grad_scale: 32.0 2023-05-17 03:35:22,218 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.626e+02 2.518e+02 3.074e+02 3.919e+02 6.846e+02, threshold=6.148e+02, percent-clipped=1.0 2023-05-17 03:35:50,414 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=278323.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 03:35:53,083 INFO [finetune.py:992] (0/2) Epoch 15, batch 7050, loss[loss=0.1444, simple_loss=0.2354, pruned_loss=0.02667, over 12327.00 frames. ], tot_loss[loss=0.1651, simple_loss=0.2549, pruned_loss=0.0377, over 2376073.91 frames. ], batch size: 31, lr: 3.48e-03, grad_scale: 32.0 2023-05-17 03:35:53,166 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=278327.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 03:35:54,882 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2023-05-17 03:36:08,141 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=278347.0, num_to_drop=1, layers_to_drop={2} 2023-05-17 03:36:27,378 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=278374.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 03:36:29,294 INFO [finetune.py:992] (0/2) Epoch 15, batch 7100, loss[loss=0.1825, simple_loss=0.2624, pruned_loss=0.05128, over 12114.00 frames. ], tot_loss[loss=0.1649, simple_loss=0.2542, pruned_loss=0.03785, over 2378839.82 frames. ], batch size: 33, lr: 3.48e-03, grad_scale: 32.0 2023-05-17 03:36:34,471 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=278384.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 03:36:34,967 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.829e+02 2.567e+02 2.960e+02 3.611e+02 5.466e+02, threshold=5.920e+02, percent-clipped=0.0 2023-05-17 03:36:42,061 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=278395.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 03:36:50,873 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.6122, 4.5002, 4.5927, 4.6287, 4.3426, 4.3768, 4.2050, 4.5684], device='cuda:0'), covar=tensor([0.0859, 0.0703, 0.1079, 0.0677, 0.1808, 0.1331, 0.0633, 0.0977], device='cuda:0'), in_proj_covar=tensor([0.0567, 0.0733, 0.0645, 0.0663, 0.0884, 0.0777, 0.0594, 0.0501], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0005, 0.0004, 0.0004, 0.0005, 0.0005, 0.0004, 0.0003], device='cuda:0') 2023-05-17 03:37:04,848 INFO [finetune.py:992] (0/2) Epoch 15, batch 7150, loss[loss=0.1716, simple_loss=0.2676, pruned_loss=0.03773, over 12396.00 frames. ], tot_loss[loss=0.1658, simple_loss=0.2548, pruned_loss=0.03843, over 2372113.78 frames. ], batch size: 38, lr: 3.48e-03, grad_scale: 32.0 2023-05-17 03:37:10,792 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=278435.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 03:37:34,198 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.2544, 4.4802, 2.7928, 2.3961, 3.9106, 2.5619, 3.8549, 2.9702], device='cuda:0'), covar=tensor([0.0689, 0.0517, 0.1160, 0.1652, 0.0253, 0.1375, 0.0534, 0.0984], device='cuda:0'), in_proj_covar=tensor([0.0188, 0.0257, 0.0177, 0.0200, 0.0142, 0.0183, 0.0198, 0.0174], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-17 03:37:40,934 INFO [finetune.py:992] (0/2) Epoch 15, batch 7200, loss[loss=0.1916, simple_loss=0.2853, pruned_loss=0.04898, over 12112.00 frames. ], tot_loss[loss=0.1659, simple_loss=0.2551, pruned_loss=0.03835, over 2372441.54 frames. ], batch size: 42, lr: 3.48e-03, grad_scale: 32.0 2023-05-17 03:37:46,545 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.886e+02 2.745e+02 3.138e+02 3.930e+02 1.052e+03, threshold=6.276e+02, percent-clipped=4.0 2023-05-17 03:38:05,903 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=278511.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 03:38:17,163 INFO [finetune.py:992] (0/2) Epoch 15, batch 7250, loss[loss=0.1551, simple_loss=0.253, pruned_loss=0.02863, over 11718.00 frames. ], tot_loss[loss=0.166, simple_loss=0.2554, pruned_loss=0.03827, over 2374956.38 frames. ], batch size: 48, lr: 3.48e-03, grad_scale: 16.0 2023-05-17 03:38:49,821 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=278572.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 03:38:53,122 INFO [finetune.py:992] (0/2) Epoch 15, batch 7300, loss[loss=0.1602, simple_loss=0.2544, pruned_loss=0.03301, over 11864.00 frames. ], tot_loss[loss=0.1652, simple_loss=0.2548, pruned_loss=0.03784, over 2379128.81 frames. ], batch size: 44, lr: 3.48e-03, grad_scale: 16.0 2023-05-17 03:38:59,563 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.711e+02 2.602e+02 2.987e+02 3.460e+02 6.466e+02, threshold=5.973e+02, percent-clipped=1.0 2023-05-17 03:39:30,109 INFO [finetune.py:992] (0/2) Epoch 15, batch 7350, loss[loss=0.1655, simple_loss=0.2478, pruned_loss=0.04163, over 12084.00 frames. ], tot_loss[loss=0.1651, simple_loss=0.2548, pruned_loss=0.03773, over 2380293.75 frames. ], batch size: 32, lr: 3.48e-03, grad_scale: 16.0 2023-05-17 03:39:30,320 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.2865, 4.9802, 5.2966, 5.2254, 4.9901, 5.2748, 5.1677, 3.1844], device='cuda:0'), covar=tensor([0.0116, 0.0066, 0.0059, 0.0048, 0.0040, 0.0085, 0.0062, 0.0595], device='cuda:0'), in_proj_covar=tensor([0.0072, 0.0081, 0.0085, 0.0075, 0.0062, 0.0094, 0.0084, 0.0100], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-17 03:39:30,324 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=278627.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 03:40:04,225 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.5688, 2.9172, 3.2537, 4.4158, 2.4442, 4.4852, 4.5611, 4.6132], device='cuda:0'), covar=tensor([0.0131, 0.1154, 0.0542, 0.0187, 0.1341, 0.0224, 0.0166, 0.0108], device='cuda:0'), in_proj_covar=tensor([0.0125, 0.0208, 0.0187, 0.0125, 0.0195, 0.0185, 0.0182, 0.0127], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002], device='cuda:0') 2023-05-17 03:40:04,785 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=278675.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 03:40:06,106 INFO [finetune.py:992] (0/2) Epoch 15, batch 7400, loss[loss=0.1828, simple_loss=0.2801, pruned_loss=0.04279, over 10536.00 frames. ], tot_loss[loss=0.165, simple_loss=0.2546, pruned_loss=0.03769, over 2382870.23 frames. ], batch size: 68, lr: 3.48e-03, grad_scale: 16.0 2023-05-17 03:40:07,641 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=278679.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 03:40:12,476 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.888e+02 2.552e+02 3.075e+02 3.617e+02 5.572e+02, threshold=6.150e+02, percent-clipped=0.0 2023-05-17 03:40:14,130 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.2138, 3.7191, 3.9090, 4.2594, 2.7817, 3.5360, 2.2601, 3.7829], device='cuda:0'), covar=tensor([0.1774, 0.0997, 0.1120, 0.0723, 0.1459, 0.0866, 0.2312, 0.1428], device='cuda:0'), in_proj_covar=tensor([0.0230, 0.0269, 0.0300, 0.0362, 0.0245, 0.0247, 0.0263, 0.0377], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-17 03:40:20,595 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 2023-05-17 03:40:41,671 INFO [finetune.py:992] (0/2) Epoch 15, batch 7450, loss[loss=0.1677, simple_loss=0.2642, pruned_loss=0.03562, over 11590.00 frames. ], tot_loss[loss=0.1645, simple_loss=0.2539, pruned_loss=0.03756, over 2381552.37 frames. ], batch size: 48, lr: 3.48e-03, grad_scale: 16.0 2023-05-17 03:40:43,299 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.9437, 4.6370, 4.7567, 4.9004, 4.6271, 4.8653, 4.8160, 2.6931], device='cuda:0'), covar=tensor([0.0114, 0.0077, 0.0092, 0.0056, 0.0051, 0.0093, 0.0073, 0.0789], device='cuda:0'), in_proj_covar=tensor([0.0072, 0.0082, 0.0085, 0.0075, 0.0062, 0.0095, 0.0084, 0.0100], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-17 03:40:43,909 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=278730.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 03:41:18,053 INFO [finetune.py:992] (0/2) Epoch 15, batch 7500, loss[loss=0.2406, simple_loss=0.319, pruned_loss=0.08112, over 8169.00 frames. ], tot_loss[loss=0.1651, simple_loss=0.2547, pruned_loss=0.03771, over 2378532.59 frames. ], batch size: 98, lr: 3.48e-03, grad_scale: 16.0 2023-05-17 03:41:24,594 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.949e+02 2.649e+02 3.038e+02 3.624e+02 6.631e+02, threshold=6.076e+02, percent-clipped=2.0 2023-05-17 03:41:54,360 INFO [finetune.py:992] (0/2) Epoch 15, batch 7550, loss[loss=0.1686, simple_loss=0.2573, pruned_loss=0.03998, over 10564.00 frames. ], tot_loss[loss=0.1665, simple_loss=0.2559, pruned_loss=0.03849, over 2369213.27 frames. ], batch size: 69, lr: 3.48e-03, grad_scale: 16.0 2023-05-17 03:42:22,756 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=278867.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 03:42:28,548 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.0500, 2.5365, 3.8065, 3.0784, 3.6802, 3.2628, 2.4235, 3.6756], device='cuda:0'), covar=tensor([0.0176, 0.0444, 0.0163, 0.0320, 0.0155, 0.0235, 0.0502, 0.0131], device='cuda:0'), in_proj_covar=tensor([0.0187, 0.0212, 0.0200, 0.0195, 0.0227, 0.0173, 0.0205, 0.0196], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-17 03:42:29,698 INFO [finetune.py:992] (0/2) Epoch 15, batch 7600, loss[loss=0.1508, simple_loss=0.2424, pruned_loss=0.02966, over 12149.00 frames. ], tot_loss[loss=0.166, simple_loss=0.2556, pruned_loss=0.03821, over 2379330.10 frames. ], batch size: 36, lr: 3.48e-03, grad_scale: 16.0 2023-05-17 03:42:36,833 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.660e+02 2.704e+02 3.260e+02 3.972e+02 8.600e+02, threshold=6.520e+02, percent-clipped=2.0 2023-05-17 03:42:43,535 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.1528, 2.2461, 3.1144, 4.0298, 1.8350, 4.1722, 4.2799, 4.3615], device='cuda:0'), covar=tensor([0.0164, 0.1383, 0.0457, 0.0205, 0.1567, 0.0194, 0.0148, 0.0120], device='cuda:0'), in_proj_covar=tensor([0.0125, 0.0209, 0.0187, 0.0125, 0.0196, 0.0186, 0.0182, 0.0127], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002], device='cuda:0') 2023-05-17 03:42:43,558 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.4096, 3.5216, 3.1856, 3.0864, 2.7297, 2.6007, 3.5188, 2.1877], device='cuda:0'), covar=tensor([0.0410, 0.0150, 0.0199, 0.0187, 0.0412, 0.0340, 0.0157, 0.0471], device='cuda:0'), in_proj_covar=tensor([0.0194, 0.0165, 0.0169, 0.0189, 0.0204, 0.0200, 0.0175, 0.0205], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-17 03:42:45,579 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.1336, 5.9709, 5.4998, 5.4888, 6.0549, 5.4466, 5.4308, 5.5154], device='cuda:0'), covar=tensor([0.1368, 0.0960, 0.1199, 0.2087, 0.0988, 0.2024, 0.2043, 0.1274], device='cuda:0'), in_proj_covar=tensor([0.0363, 0.0504, 0.0408, 0.0459, 0.0473, 0.0440, 0.0406, 0.0393], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-17 03:43:06,461 INFO [finetune.py:992] (0/2) Epoch 15, batch 7650, loss[loss=0.1702, simple_loss=0.2637, pruned_loss=0.03832, over 12363.00 frames. ], tot_loss[loss=0.1659, simple_loss=0.2555, pruned_loss=0.0382, over 2377014.90 frames. ], batch size: 35, lr: 3.48e-03, grad_scale: 16.0 2023-05-17 03:43:16,586 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.3689, 5.1325, 5.3011, 5.2830, 4.9053, 5.0161, 4.7861, 5.2252], device='cuda:0'), covar=tensor([0.0665, 0.0628, 0.0884, 0.0691, 0.1962, 0.1268, 0.0579, 0.1003], device='cuda:0'), in_proj_covar=tensor([0.0572, 0.0737, 0.0649, 0.0667, 0.0893, 0.0782, 0.0598, 0.0505], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0005, 0.0004, 0.0004, 0.0005, 0.0005, 0.0004, 0.0003], device='cuda:0') 2023-05-17 03:43:38,162 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.4317, 4.9225, 3.1200, 2.6848, 3.9935, 2.8168, 4.0119, 3.4716], device='cuda:0'), covar=tensor([0.0736, 0.0478, 0.1169, 0.1603, 0.0357, 0.1274, 0.0563, 0.0776], device='cuda:0'), in_proj_covar=tensor([0.0189, 0.0257, 0.0177, 0.0201, 0.0143, 0.0183, 0.0199, 0.0174], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-17 03:43:39,924 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.90 vs. limit=2.0 2023-05-17 03:43:42,940 INFO [finetune.py:992] (0/2) Epoch 15, batch 7700, loss[loss=0.1445, simple_loss=0.2321, pruned_loss=0.02841, over 12133.00 frames. ], tot_loss[loss=0.1663, simple_loss=0.2557, pruned_loss=0.0384, over 2369963.57 frames. ], batch size: 30, lr: 3.48e-03, grad_scale: 16.0 2023-05-17 03:43:44,367 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=278979.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 03:43:48,981 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.059e+02 2.736e+02 3.229e+02 3.704e+02 6.260e+02, threshold=6.459e+02, percent-clipped=0.0 2023-05-17 03:44:18,301 INFO [finetune.py:992] (0/2) Epoch 15, batch 7750, loss[loss=0.1454, simple_loss=0.2257, pruned_loss=0.03251, over 12182.00 frames. ], tot_loss[loss=0.1672, simple_loss=0.2568, pruned_loss=0.03877, over 2369353.17 frames. ], batch size: 29, lr: 3.47e-03, grad_scale: 16.0 2023-05-17 03:44:18,362 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=279027.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 03:44:20,529 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=279030.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 03:44:31,779 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.3819, 5.2036, 5.3639, 5.3585, 4.9752, 5.0710, 4.8536, 5.2451], device='cuda:0'), covar=tensor([0.0708, 0.0621, 0.0852, 0.0595, 0.2055, 0.1281, 0.0571, 0.1078], device='cuda:0'), in_proj_covar=tensor([0.0574, 0.0740, 0.0651, 0.0668, 0.0895, 0.0783, 0.0600, 0.0507], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0005, 0.0004, 0.0004, 0.0005, 0.0005, 0.0004, 0.0003], device='cuda:0') 2023-05-17 03:44:54,127 INFO [finetune.py:992] (0/2) Epoch 15, batch 7800, loss[loss=0.16, simple_loss=0.2514, pruned_loss=0.0343, over 12117.00 frames. ], tot_loss[loss=0.1671, simple_loss=0.2567, pruned_loss=0.03878, over 2376694.60 frames. ], batch size: 33, lr: 3.47e-03, grad_scale: 16.0 2023-05-17 03:44:55,531 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=279078.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 03:45:01,255 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.855e+02 2.663e+02 3.247e+02 3.788e+02 7.442e+02, threshold=6.494e+02, percent-clipped=4.0 2023-05-17 03:45:18,025 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.7592, 4.3674, 4.2339, 4.6506, 3.3146, 4.1011, 2.8087, 4.4516], device='cuda:0'), covar=tensor([0.1319, 0.0582, 0.0831, 0.0600, 0.1108, 0.0558, 0.1657, 0.1114], device='cuda:0'), in_proj_covar=tensor([0.0231, 0.0269, 0.0301, 0.0363, 0.0245, 0.0248, 0.0264, 0.0380], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-17 03:45:20,214 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.5420, 5.1442, 5.5779, 4.8746, 5.1986, 4.9279, 5.5609, 5.2332], device='cuda:0'), covar=tensor([0.0251, 0.0314, 0.0207, 0.0237, 0.0434, 0.0302, 0.0212, 0.0192], device='cuda:0'), in_proj_covar=tensor([0.0270, 0.0275, 0.0299, 0.0268, 0.0272, 0.0271, 0.0245, 0.0219], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-17 03:45:30,584 INFO [finetune.py:992] (0/2) Epoch 15, batch 7850, loss[loss=0.1458, simple_loss=0.2362, pruned_loss=0.02773, over 12244.00 frames. ], tot_loss[loss=0.1669, simple_loss=0.2566, pruned_loss=0.0386, over 2373753.99 frames. ], batch size: 32, lr: 3.47e-03, grad_scale: 16.0 2023-05-17 03:45:33,549 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=279131.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 03:45:33,999 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.33 vs. limit=2.0 2023-05-17 03:45:48,823 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.74 vs. limit=2.0 2023-05-17 03:45:59,261 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=279167.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 03:46:00,119 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.5775, 3.1996, 5.0507, 2.6801, 2.7409, 3.7362, 3.1501, 3.7858], device='cuda:0'), covar=tensor([0.0454, 0.1220, 0.0328, 0.1213, 0.1973, 0.1551, 0.1354, 0.1074], device='cuda:0'), in_proj_covar=tensor([0.0241, 0.0241, 0.0260, 0.0186, 0.0240, 0.0298, 0.0228, 0.0270], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-17 03:46:06,361 INFO [finetune.py:992] (0/2) Epoch 15, batch 7900, loss[loss=0.2027, simple_loss=0.2864, pruned_loss=0.05954, over 12123.00 frames. ], tot_loss[loss=0.1676, simple_loss=0.2572, pruned_loss=0.03906, over 2368489.89 frames. ], batch size: 38, lr: 3.47e-03, grad_scale: 16.0 2023-05-17 03:46:13,338 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.080e+02 2.661e+02 3.094e+02 3.803e+02 8.182e+02, threshold=6.187e+02, percent-clipped=2.0 2023-05-17 03:46:18,011 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=279192.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 03:46:34,680 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=279215.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 03:46:43,838 INFO [finetune.py:992] (0/2) Epoch 15, batch 7950, loss[loss=0.1875, simple_loss=0.2786, pruned_loss=0.04816, over 12035.00 frames. ], tot_loss[loss=0.1678, simple_loss=0.2572, pruned_loss=0.0392, over 2364983.99 frames. ], batch size: 42, lr: 3.47e-03, grad_scale: 8.0 2023-05-17 03:47:19,697 INFO [finetune.py:992] (0/2) Epoch 15, batch 8000, loss[loss=0.1803, simple_loss=0.275, pruned_loss=0.04286, over 11397.00 frames. ], tot_loss[loss=0.1672, simple_loss=0.2565, pruned_loss=0.03892, over 2365035.04 frames. ], batch size: 55, lr: 3.47e-03, grad_scale: 8.0 2023-05-17 03:47:24,117 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.0032, 2.3607, 2.2713, 2.2537, 2.1122, 2.1326, 2.1756, 1.7236], device='cuda:0'), covar=tensor([0.0316, 0.0195, 0.0215, 0.0199, 0.0325, 0.0246, 0.0210, 0.0428], device='cuda:0'), in_proj_covar=tensor([0.0196, 0.0167, 0.0170, 0.0192, 0.0207, 0.0202, 0.0177, 0.0208], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-17 03:47:26,734 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.046e+02 2.777e+02 3.188e+02 3.959e+02 2.988e+03, threshold=6.375e+02, percent-clipped=3.0 2023-05-17 03:47:55,591 INFO [finetune.py:992] (0/2) Epoch 15, batch 8050, loss[loss=0.1441, simple_loss=0.2253, pruned_loss=0.03146, over 11993.00 frames. ], tot_loss[loss=0.1674, simple_loss=0.2567, pruned_loss=0.039, over 2366891.70 frames. ], batch size: 28, lr: 3.47e-03, grad_scale: 8.0 2023-05-17 03:48:31,829 INFO [finetune.py:992] (0/2) Epoch 15, batch 8100, loss[loss=0.1595, simple_loss=0.2455, pruned_loss=0.03672, over 12124.00 frames. ], tot_loss[loss=0.1679, simple_loss=0.257, pruned_loss=0.03942, over 2363599.77 frames. ], batch size: 33, lr: 3.47e-03, grad_scale: 8.0 2023-05-17 03:48:34,127 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.0201, 3.8423, 4.0005, 3.7445, 3.8823, 3.6897, 3.9928, 3.6237], device='cuda:0'), covar=tensor([0.0392, 0.0382, 0.0365, 0.0246, 0.0397, 0.0320, 0.0308, 0.1353], device='cuda:0'), in_proj_covar=tensor([0.0271, 0.0277, 0.0300, 0.0269, 0.0273, 0.0272, 0.0247, 0.0221], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-17 03:48:39,009 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.845e+02 2.813e+02 3.387e+02 3.883e+02 1.132e+03, threshold=6.774e+02, percent-clipped=3.0 2023-05-17 03:49:07,750 INFO [finetune.py:992] (0/2) Epoch 15, batch 8150, loss[loss=0.1651, simple_loss=0.2651, pruned_loss=0.03252, over 12347.00 frames. ], tot_loss[loss=0.169, simple_loss=0.2581, pruned_loss=0.03999, over 2351733.79 frames. ], batch size: 36, lr: 3.47e-03, grad_scale: 8.0 2023-05-17 03:49:20,917 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.7895, 2.9410, 4.7458, 4.8797, 2.7880, 2.7021, 3.0945, 2.3441], device='cuda:0'), covar=tensor([0.1729, 0.2994, 0.0435, 0.0393, 0.1500, 0.2565, 0.2838, 0.3921], device='cuda:0'), in_proj_covar=tensor([0.0301, 0.0392, 0.0282, 0.0305, 0.0276, 0.0317, 0.0394, 0.0377], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-17 03:49:24,214 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=4.65 vs. limit=5.0 2023-05-17 03:49:44,224 INFO [finetune.py:992] (0/2) Epoch 15, batch 8200, loss[loss=0.1821, simple_loss=0.2737, pruned_loss=0.04529, over 12113.00 frames. ], tot_loss[loss=0.1684, simple_loss=0.2574, pruned_loss=0.0397, over 2355757.88 frames. ], batch size: 38, lr: 3.47e-03, grad_scale: 8.0 2023-05-17 03:49:46,776 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.90 vs. limit=2.0 2023-05-17 03:49:51,248 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.767e+02 2.721e+02 3.143e+02 3.749e+02 8.958e+02, threshold=6.285e+02, percent-clipped=2.0 2023-05-17 03:49:51,370 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=279487.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 03:50:20,375 INFO [finetune.py:992] (0/2) Epoch 15, batch 8250, loss[loss=0.1682, simple_loss=0.2487, pruned_loss=0.04386, over 12333.00 frames. ], tot_loss[loss=0.1667, simple_loss=0.2558, pruned_loss=0.03885, over 2367338.46 frames. ], batch size: 30, lr: 3.47e-03, grad_scale: 8.0 2023-05-17 03:50:21,942 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=279529.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 03:50:31,189 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=279542.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 03:50:53,543 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2023-05-17 03:50:55,721 INFO [finetune.py:992] (0/2) Epoch 15, batch 8300, loss[loss=0.1523, simple_loss=0.2381, pruned_loss=0.03323, over 12026.00 frames. ], tot_loss[loss=0.1663, simple_loss=0.2556, pruned_loss=0.03853, over 2373685.77 frames. ], batch size: 31, lr: 3.47e-03, grad_scale: 8.0 2023-05-17 03:51:02,924 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.762e+02 2.526e+02 2.895e+02 3.782e+02 1.118e+03, threshold=5.789e+02, percent-clipped=4.0 2023-05-17 03:51:05,322 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=279590.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 03:51:06,708 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.4589, 2.9258, 3.7864, 2.3564, 2.5752, 3.0849, 2.9269, 3.2076], device='cuda:0'), covar=tensor([0.0593, 0.1080, 0.0461, 0.1179, 0.1720, 0.1347, 0.1224, 0.1074], device='cuda:0'), in_proj_covar=tensor([0.0242, 0.0242, 0.0261, 0.0186, 0.0240, 0.0298, 0.0228, 0.0271], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-17 03:51:14,687 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=279603.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 03:51:32,060 INFO [finetune.py:992] (0/2) Epoch 15, batch 8350, loss[loss=0.1618, simple_loss=0.2581, pruned_loss=0.03277, over 12101.00 frames. ], tot_loss[loss=0.1657, simple_loss=0.2547, pruned_loss=0.03835, over 2374139.67 frames. ], batch size: 38, lr: 3.47e-03, grad_scale: 8.0 2023-05-17 03:51:40,207 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.6566, 2.7810, 4.3621, 4.4739, 2.7597, 2.4952, 2.7841, 2.0609], device='cuda:0'), covar=tensor([0.1719, 0.2898, 0.0520, 0.0499, 0.1396, 0.2653, 0.2895, 0.4308], device='cuda:0'), in_proj_covar=tensor([0.0301, 0.0392, 0.0281, 0.0306, 0.0276, 0.0316, 0.0395, 0.0378], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-17 03:52:06,750 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.84 vs. limit=2.0 2023-05-17 03:52:08,313 INFO [finetune.py:992] (0/2) Epoch 15, batch 8400, loss[loss=0.1453, simple_loss=0.2356, pruned_loss=0.02747, over 11887.00 frames. ], tot_loss[loss=0.1668, simple_loss=0.2558, pruned_loss=0.03892, over 2358629.01 frames. ], batch size: 26, lr: 3.47e-03, grad_scale: 8.0 2023-05-17 03:52:15,368 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.924e+02 2.699e+02 3.158e+02 3.706e+02 8.753e+02, threshold=6.316e+02, percent-clipped=3.0 2023-05-17 03:52:43,833 INFO [finetune.py:992] (0/2) Epoch 15, batch 8450, loss[loss=0.177, simple_loss=0.2659, pruned_loss=0.04404, over 12123.00 frames. ], tot_loss[loss=0.1675, simple_loss=0.2564, pruned_loss=0.0393, over 2350709.60 frames. ], batch size: 39, lr: 3.47e-03, grad_scale: 8.0 2023-05-17 03:52:50,056 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.45 vs. limit=2.0 2023-05-17 03:53:20,573 INFO [finetune.py:992] (0/2) Epoch 15, batch 8500, loss[loss=0.1465, simple_loss=0.2257, pruned_loss=0.03365, over 12307.00 frames. ], tot_loss[loss=0.1674, simple_loss=0.2565, pruned_loss=0.03915, over 2353616.18 frames. ], batch size: 30, lr: 3.47e-03, grad_scale: 8.0 2023-05-17 03:53:27,633 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.155e+02 2.697e+02 3.107e+02 3.683e+02 6.037e+02, threshold=6.214e+02, percent-clipped=0.0 2023-05-17 03:53:27,814 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=279787.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 03:53:57,224 INFO [finetune.py:992] (0/2) Epoch 15, batch 8550, loss[loss=0.215, simple_loss=0.3094, pruned_loss=0.06027, over 11765.00 frames. ], tot_loss[loss=0.1675, simple_loss=0.2569, pruned_loss=0.03907, over 2354098.89 frames. ], batch size: 44, lr: 3.47e-03, grad_scale: 8.0 2023-05-17 03:54:02,841 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=279835.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 03:54:15,037 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.38 vs. limit=2.0 2023-05-17 03:54:15,514 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.1269, 4.5807, 4.0495, 4.8507, 4.4903, 2.8929, 4.1485, 2.9659], device='cuda:0'), covar=tensor([0.0921, 0.0838, 0.1569, 0.0545, 0.1147, 0.1750, 0.1114, 0.3443], device='cuda:0'), in_proj_covar=tensor([0.0311, 0.0386, 0.0366, 0.0330, 0.0372, 0.0277, 0.0349, 0.0372], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-17 03:54:32,562 INFO [finetune.py:992] (0/2) Epoch 15, batch 8600, loss[loss=0.1858, simple_loss=0.2783, pruned_loss=0.04664, over 12025.00 frames. ], tot_loss[loss=0.167, simple_loss=0.2566, pruned_loss=0.03876, over 2364344.79 frames. ], batch size: 40, lr: 3.47e-03, grad_scale: 8.0 2023-05-17 03:54:38,498 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=279885.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 03:54:39,873 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.994e+02 2.663e+02 3.176e+02 3.689e+02 7.221e+02, threshold=6.352e+02, percent-clipped=3.0 2023-05-17 03:54:47,732 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=279898.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 03:54:54,922 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.4196, 2.7162, 3.1738, 4.2676, 2.1640, 4.3081, 4.4267, 4.4267], device='cuda:0'), covar=tensor([0.0125, 0.1196, 0.0548, 0.0181, 0.1478, 0.0258, 0.0159, 0.0121], device='cuda:0'), in_proj_covar=tensor([0.0124, 0.0207, 0.0187, 0.0125, 0.0194, 0.0185, 0.0181, 0.0127], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-17 03:55:00,180 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2023-05-17 03:55:09,155 INFO [finetune.py:992] (0/2) Epoch 15, batch 8650, loss[loss=0.1574, simple_loss=0.2557, pruned_loss=0.0296, over 12036.00 frames. ], tot_loss[loss=0.1662, simple_loss=0.2556, pruned_loss=0.03838, over 2375103.55 frames. ], batch size: 37, lr: 3.47e-03, grad_scale: 8.0 2023-05-17 03:55:45,472 INFO [finetune.py:992] (0/2) Epoch 15, batch 8700, loss[loss=0.2402, simple_loss=0.3128, pruned_loss=0.0838, over 7932.00 frames. ], tot_loss[loss=0.1658, simple_loss=0.2554, pruned_loss=0.03811, over 2371690.30 frames. ], batch size: 98, lr: 3.47e-03, grad_scale: 8.0 2023-05-17 03:55:52,660 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.005e+02 2.577e+02 3.050e+02 3.655e+02 6.083e+02, threshold=6.099e+02, percent-clipped=0.0 2023-05-17 03:56:02,202 INFO [checkpoint.py:75] (0/2) Saving checkpoint to pruned_transducer_stateless7/exp_giga_finetune/checkpoint-180000.pt 2023-05-17 03:56:16,136 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.81 vs. limit=2.0 2023-05-17 03:56:23,860 INFO [finetune.py:992] (0/2) Epoch 15, batch 8750, loss[loss=0.1612, simple_loss=0.2496, pruned_loss=0.03644, over 12286.00 frames. ], tot_loss[loss=0.1659, simple_loss=0.2556, pruned_loss=0.03812, over 2373346.70 frames. ], batch size: 33, lr: 3.47e-03, grad_scale: 8.0 2023-05-17 03:56:26,112 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.4754, 2.6667, 3.6674, 4.4705, 3.8552, 4.4704, 3.9552, 3.1636], device='cuda:0'), covar=tensor([0.0036, 0.0361, 0.0155, 0.0050, 0.0118, 0.0068, 0.0102, 0.0379], device='cuda:0'), in_proj_covar=tensor([0.0091, 0.0123, 0.0107, 0.0080, 0.0107, 0.0118, 0.0101, 0.0140], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-17 03:56:37,563 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=4.18 vs. limit=5.0 2023-05-17 03:56:39,011 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 2023-05-17 03:56:41,718 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=280051.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 03:56:59,743 INFO [finetune.py:992] (0/2) Epoch 15, batch 8800, loss[loss=0.1885, simple_loss=0.2725, pruned_loss=0.05222, over 11620.00 frames. ], tot_loss[loss=0.1661, simple_loss=0.2558, pruned_loss=0.03824, over 2374228.17 frames. ], batch size: 48, lr: 3.47e-03, grad_scale: 8.0 2023-05-17 03:57:07,449 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.044e+02 2.782e+02 3.180e+02 3.650e+02 1.240e+03, threshold=6.359e+02, percent-clipped=2.0 2023-05-17 03:57:08,333 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=280088.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 03:57:14,814 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.5206, 2.5714, 3.2961, 4.3472, 2.2728, 4.4030, 4.5194, 4.6174], device='cuda:0'), covar=tensor([0.0133, 0.1337, 0.0455, 0.0170, 0.1346, 0.0224, 0.0151, 0.0121], device='cuda:0'), in_proj_covar=tensor([0.0125, 0.0210, 0.0189, 0.0126, 0.0196, 0.0187, 0.0183, 0.0128], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002], device='cuda:0') 2023-05-17 03:57:25,371 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=280112.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 03:57:35,704 INFO [finetune.py:992] (0/2) Epoch 15, batch 8850, loss[loss=0.1471, simple_loss=0.2375, pruned_loss=0.02832, over 12073.00 frames. ], tot_loss[loss=0.1652, simple_loss=0.2549, pruned_loss=0.03777, over 2382139.29 frames. ], batch size: 32, lr: 3.47e-03, grad_scale: 8.0 2023-05-17 03:57:51,366 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=280149.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 03:57:56,387 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.7002, 4.2924, 4.4015, 4.5489, 4.4928, 4.6243, 4.5392, 2.6458], device='cuda:0'), covar=tensor([0.0193, 0.0132, 0.0171, 0.0127, 0.0077, 0.0185, 0.0158, 0.0999], device='cuda:0'), in_proj_covar=tensor([0.0072, 0.0082, 0.0085, 0.0076, 0.0062, 0.0096, 0.0084, 0.0100], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-17 03:58:05,033 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.5935, 2.7746, 3.7634, 4.6768, 4.0104, 4.6955, 4.1101, 3.7458], device='cuda:0'), covar=tensor([0.0052, 0.0398, 0.0156, 0.0048, 0.0115, 0.0068, 0.0120, 0.0292], device='cuda:0'), in_proj_covar=tensor([0.0092, 0.0125, 0.0108, 0.0081, 0.0108, 0.0119, 0.0102, 0.0142], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-17 03:58:11,382 INFO [finetune.py:992] (0/2) Epoch 15, batch 8900, loss[loss=0.1742, simple_loss=0.2593, pruned_loss=0.04456, over 12371.00 frames. ], tot_loss[loss=0.1653, simple_loss=0.2552, pruned_loss=0.03772, over 2378785.21 frames. ], batch size: 38, lr: 3.47e-03, grad_scale: 8.0 2023-05-17 03:58:17,133 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=280185.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 03:58:18,477 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.939e+02 2.621e+02 3.017e+02 3.657e+02 5.872e+02, threshold=6.035e+02, percent-clipped=0.0 2023-05-17 03:58:22,257 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.2545, 4.8149, 5.2654, 4.6029, 4.9781, 4.6579, 5.3082, 4.9263], device='cuda:0'), covar=tensor([0.0283, 0.0408, 0.0275, 0.0241, 0.0355, 0.0323, 0.0220, 0.0299], device='cuda:0'), in_proj_covar=tensor([0.0270, 0.0277, 0.0299, 0.0269, 0.0272, 0.0273, 0.0247, 0.0221], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-17 03:58:27,139 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=280198.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 03:58:34,792 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.6133, 2.6038, 3.1553, 4.4055, 2.4566, 4.4467, 4.5804, 4.6196], device='cuda:0'), covar=tensor([0.0110, 0.1296, 0.0538, 0.0175, 0.1323, 0.0241, 0.0163, 0.0109], device='cuda:0'), in_proj_covar=tensor([0.0125, 0.0211, 0.0189, 0.0127, 0.0196, 0.0187, 0.0183, 0.0128], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002], device='cuda:0') 2023-05-17 03:58:39,069 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.5739, 3.5628, 3.1584, 3.2082, 2.9251, 2.7111, 3.5849, 2.3364], device='cuda:0'), covar=tensor([0.0418, 0.0179, 0.0262, 0.0228, 0.0405, 0.0370, 0.0161, 0.0523], device='cuda:0'), in_proj_covar=tensor([0.0194, 0.0167, 0.0170, 0.0191, 0.0204, 0.0202, 0.0176, 0.0206], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-17 03:58:48,061 INFO [finetune.py:992] (0/2) Epoch 15, batch 8950, loss[loss=0.1588, simple_loss=0.2563, pruned_loss=0.03063, over 12143.00 frames. ], tot_loss[loss=0.1656, simple_loss=0.2555, pruned_loss=0.03786, over 2381314.95 frames. ], batch size: 36, lr: 3.47e-03, grad_scale: 8.0 2023-05-17 03:58:52,990 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=280233.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 03:59:02,054 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=280246.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 03:59:12,309 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.11 vs. limit=2.0 2023-05-17 03:59:23,023 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.6283, 4.8772, 3.1431, 2.8510, 4.2320, 3.0193, 4.1315, 3.5959], device='cuda:0'), covar=tensor([0.0687, 0.0571, 0.1100, 0.1466, 0.0316, 0.1170, 0.0508, 0.0676], device='cuda:0'), in_proj_covar=tensor([0.0190, 0.0260, 0.0178, 0.0201, 0.0143, 0.0183, 0.0198, 0.0174], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-17 03:59:24,217 INFO [finetune.py:992] (0/2) Epoch 15, batch 9000, loss[loss=0.1828, simple_loss=0.2781, pruned_loss=0.04376, over 12150.00 frames. ], tot_loss[loss=0.1661, simple_loss=0.2556, pruned_loss=0.03827, over 2372227.51 frames. ], batch size: 38, lr: 3.47e-03, grad_scale: 8.0 2023-05-17 03:59:24,218 INFO [finetune.py:1017] (0/2) Computing validation loss 2023-05-17 03:59:40,556 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.0392, 4.8951, 5.1176, 5.0368, 4.6447, 4.7598, 4.6118, 4.9153], device='cuda:0'), covar=tensor([0.0802, 0.0585, 0.0709, 0.0654, 0.1952, 0.1348, 0.0516, 0.1174], device='cuda:0'), in_proj_covar=tensor([0.0570, 0.0734, 0.0644, 0.0668, 0.0891, 0.0770, 0.0596, 0.0501], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0005, 0.0004, 0.0004, 0.0005, 0.0005, 0.0004, 0.0003], device='cuda:0') 2023-05-17 03:59:42,466 INFO [finetune.py:1026] (0/2) Epoch 15, validation: loss=0.3191, simple_loss=0.3935, pruned_loss=0.1223, over 1020973.00 frames. 2023-05-17 03:59:42,467 INFO [finetune.py:1027] (0/2) Maximum memory allocated so far is 12827MB 2023-05-17 03:59:49,611 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.927e+02 2.626e+02 3.045e+02 3.488e+02 5.350e+02, threshold=6.090e+02, percent-clipped=0.0 2023-05-17 04:00:18,855 INFO [finetune.py:992] (0/2) Epoch 15, batch 9050, loss[loss=0.1692, simple_loss=0.2625, pruned_loss=0.03791, over 12150.00 frames. ], tot_loss[loss=0.166, simple_loss=0.2556, pruned_loss=0.03823, over 2366394.24 frames. ], batch size: 39, lr: 3.47e-03, grad_scale: 8.0 2023-05-17 04:00:19,060 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=280327.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 04:00:32,745 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=3.38 vs. limit=5.0 2023-05-17 04:00:54,712 INFO [finetune.py:992] (0/2) Epoch 15, batch 9100, loss[loss=0.1729, simple_loss=0.2673, pruned_loss=0.03926, over 12115.00 frames. ], tot_loss[loss=0.166, simple_loss=0.2555, pruned_loss=0.03825, over 2367345.54 frames. ], batch size: 38, lr: 3.46e-03, grad_scale: 8.0 2023-05-17 04:01:01,652 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.865e+02 2.720e+02 3.183e+02 3.780e+02 5.897e+02, threshold=6.365e+02, percent-clipped=0.0 2023-05-17 04:01:02,666 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=280388.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 04:01:16,440 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=280407.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 04:01:30,558 INFO [finetune.py:992] (0/2) Epoch 15, batch 9150, loss[loss=0.1772, simple_loss=0.2663, pruned_loss=0.0441, over 12200.00 frames. ], tot_loss[loss=0.1652, simple_loss=0.2545, pruned_loss=0.03797, over 2372554.60 frames. ], batch size: 35, lr: 3.46e-03, grad_scale: 8.0 2023-05-17 04:01:43,449 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=280444.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 04:01:45,249 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.58 vs. limit=2.0 2023-05-17 04:01:48,566 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.8085, 2.4401, 3.6070, 3.7683, 2.9538, 2.5855, 2.6149, 2.3218], device='cuda:0'), covar=tensor([0.1392, 0.2813, 0.0696, 0.0550, 0.1034, 0.2255, 0.3028, 0.4010], device='cuda:0'), in_proj_covar=tensor([0.0301, 0.0389, 0.0281, 0.0305, 0.0275, 0.0315, 0.0395, 0.0376], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-17 04:02:06,331 INFO [finetune.py:992] (0/2) Epoch 15, batch 9200, loss[loss=0.1442, simple_loss=0.2359, pruned_loss=0.02625, over 12107.00 frames. ], tot_loss[loss=0.1654, simple_loss=0.2547, pruned_loss=0.0381, over 2376340.68 frames. ], batch size: 33, lr: 3.46e-03, grad_scale: 8.0 2023-05-17 04:02:14,210 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.924e+02 2.595e+02 3.020e+02 3.681e+02 7.637e+02, threshold=6.041e+02, percent-clipped=2.0 2023-05-17 04:02:36,192 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=280518.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 04:02:42,321 INFO [finetune.py:992] (0/2) Epoch 15, batch 9250, loss[loss=0.1726, simple_loss=0.2622, pruned_loss=0.04146, over 12153.00 frames. ], tot_loss[loss=0.1654, simple_loss=0.2549, pruned_loss=0.038, over 2371986.27 frames. ], batch size: 34, lr: 3.46e-03, grad_scale: 8.0 2023-05-17 04:02:46,697 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=280533.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 04:02:56,498 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=280547.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 04:03:12,417 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-05-17 04:03:17,694 INFO [finetune.py:992] (0/2) Epoch 15, batch 9300, loss[loss=0.1533, simple_loss=0.2374, pruned_loss=0.03457, over 11999.00 frames. ], tot_loss[loss=0.165, simple_loss=0.2544, pruned_loss=0.03776, over 2378042.63 frames. ], batch size: 28, lr: 3.46e-03, grad_scale: 8.0 2023-05-17 04:03:19,257 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=280579.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 04:03:24,706 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.909e+02 2.780e+02 3.170e+02 3.668e+02 7.596e+02, threshold=6.340e+02, percent-clipped=1.0 2023-05-17 04:03:30,505 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=280594.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 04:03:40,701 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=280608.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 04:03:45,016 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=280614.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 04:03:45,404 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.92 vs. limit=2.0 2023-05-17 04:03:54,232 INFO [finetune.py:992] (0/2) Epoch 15, batch 9350, loss[loss=0.2023, simple_loss=0.2921, pruned_loss=0.05619, over 12147.00 frames. ], tot_loss[loss=0.1653, simple_loss=0.2548, pruned_loss=0.03796, over 2377337.79 frames. ], batch size: 39, lr: 3.46e-03, grad_scale: 8.0 2023-05-17 04:04:09,278 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.7342, 2.9871, 4.5631, 4.8465, 2.8071, 2.6961, 3.1075, 2.2981], device='cuda:0'), covar=tensor([0.1646, 0.2819, 0.0502, 0.0408, 0.1413, 0.2364, 0.2663, 0.3932], device='cuda:0'), in_proj_covar=tensor([0.0303, 0.0392, 0.0284, 0.0308, 0.0277, 0.0317, 0.0397, 0.0379], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-17 04:04:29,248 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=280675.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 04:04:30,484 INFO [finetune.py:992] (0/2) Epoch 15, batch 9400, loss[loss=0.1629, simple_loss=0.2582, pruned_loss=0.03379, over 12351.00 frames. ], tot_loss[loss=0.1657, simple_loss=0.2549, pruned_loss=0.03829, over 2366781.02 frames. ], batch size: 36, lr: 3.46e-03, grad_scale: 8.0 2023-05-17 04:04:34,531 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=280683.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 04:04:37,246 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.971e+02 2.825e+02 3.289e+02 3.761e+02 8.250e+02, threshold=6.578e+02, percent-clipped=2.0 2023-05-17 04:04:47,977 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=280702.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 04:04:51,500 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=280707.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 04:05:04,929 INFO [finetune.py:992] (0/2) Epoch 15, batch 9450, loss[loss=0.1676, simple_loss=0.2476, pruned_loss=0.0438, over 12134.00 frames. ], tot_loss[loss=0.1662, simple_loss=0.2556, pruned_loss=0.03841, over 2376321.87 frames. ], batch size: 30, lr: 3.46e-03, grad_scale: 8.0 2023-05-17 04:05:17,869 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=280744.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 04:05:25,461 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=280755.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 04:05:31,228 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=280763.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 04:05:35,528 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.2715, 5.1645, 5.1106, 5.1414, 4.8603, 5.2299, 5.1936, 5.4387], device='cuda:0'), covar=tensor([0.0219, 0.0141, 0.0162, 0.0286, 0.0691, 0.0245, 0.0158, 0.0171], device='cuda:0'), in_proj_covar=tensor([0.0201, 0.0203, 0.0197, 0.0257, 0.0248, 0.0227, 0.0182, 0.0237], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0005, 0.0004, 0.0004, 0.0003, 0.0005], device='cuda:0') 2023-05-17 04:05:41,007 INFO [finetune.py:992] (0/2) Epoch 15, batch 9500, loss[loss=0.1729, simple_loss=0.2623, pruned_loss=0.04178, over 12038.00 frames. ], tot_loss[loss=0.1653, simple_loss=0.2548, pruned_loss=0.03795, over 2378078.38 frames. ], batch size: 40, lr: 3.46e-03, grad_scale: 8.0 2023-05-17 04:05:48,702 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.666e+02 2.589e+02 3.024e+02 3.554e+02 7.083e+02, threshold=6.048e+02, percent-clipped=2.0 2023-05-17 04:05:52,262 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=280792.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 04:06:17,001 INFO [finetune.py:992] (0/2) Epoch 15, batch 9550, loss[loss=0.1735, simple_loss=0.2599, pruned_loss=0.04357, over 12343.00 frames. ], tot_loss[loss=0.1646, simple_loss=0.2541, pruned_loss=0.03759, over 2384272.05 frames. ], batch size: 31, lr: 3.46e-03, grad_scale: 8.0 2023-05-17 04:06:25,496 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.2127, 4.1825, 4.1509, 4.3913, 3.1432, 4.1205, 2.7578, 4.0818], device='cuda:0'), covar=tensor([0.1649, 0.0655, 0.0854, 0.0621, 0.1144, 0.0557, 0.1772, 0.1359], device='cuda:0'), in_proj_covar=tensor([0.0229, 0.0267, 0.0301, 0.0362, 0.0244, 0.0246, 0.0263, 0.0378], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-17 04:06:50,931 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=280874.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 04:06:52,932 INFO [finetune.py:992] (0/2) Epoch 15, batch 9600, loss[loss=0.1761, simple_loss=0.25, pruned_loss=0.0511, over 12272.00 frames. ], tot_loss[loss=0.1652, simple_loss=0.2547, pruned_loss=0.03781, over 2371847.39 frames. ], batch size: 28, lr: 3.46e-03, grad_scale: 8.0 2023-05-17 04:07:00,193 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.801e+02 2.610e+02 3.063e+02 3.778e+02 8.298e+02, threshold=6.127e+02, percent-clipped=2.0 2023-05-17 04:07:01,731 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=280889.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 04:07:11,637 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=280903.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 04:07:29,222 INFO [finetune.py:992] (0/2) Epoch 15, batch 9650, loss[loss=0.1703, simple_loss=0.262, pruned_loss=0.03934, over 11791.00 frames. ], tot_loss[loss=0.1652, simple_loss=0.2547, pruned_loss=0.03788, over 2372277.48 frames. ], batch size: 44, lr: 3.46e-03, grad_scale: 8.0 2023-05-17 04:07:55,965 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.0734, 4.7421, 4.7977, 4.9999, 4.8563, 4.8669, 4.9510, 2.6577], device='cuda:0'), covar=tensor([0.0091, 0.0077, 0.0095, 0.0065, 0.0047, 0.0120, 0.0076, 0.0851], device='cuda:0'), in_proj_covar=tensor([0.0072, 0.0083, 0.0086, 0.0077, 0.0063, 0.0097, 0.0086, 0.0101], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-17 04:07:56,012 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.5364, 3.6416, 3.1976, 3.0926, 2.8677, 2.7553, 3.6485, 2.4064], device='cuda:0'), covar=tensor([0.0431, 0.0154, 0.0255, 0.0246, 0.0479, 0.0375, 0.0164, 0.0523], device='cuda:0'), in_proj_covar=tensor([0.0195, 0.0167, 0.0170, 0.0191, 0.0205, 0.0203, 0.0176, 0.0206], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-17 04:08:00,062 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=280970.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 04:08:05,058 INFO [finetune.py:992] (0/2) Epoch 15, batch 9700, loss[loss=0.176, simple_loss=0.2575, pruned_loss=0.04727, over 12124.00 frames. ], tot_loss[loss=0.1647, simple_loss=0.2542, pruned_loss=0.03758, over 2381704.18 frames. ], batch size: 30, lr: 3.46e-03, grad_scale: 8.0 2023-05-17 04:08:09,517 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=280983.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 04:08:12,216 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.848e+02 2.705e+02 3.092e+02 3.654e+02 5.516e+02, threshold=6.184e+02, percent-clipped=0.0 2023-05-17 04:08:41,606 INFO [finetune.py:992] (0/2) Epoch 15, batch 9750, loss[loss=0.175, simple_loss=0.2672, pruned_loss=0.04141, over 12124.00 frames. ], tot_loss[loss=0.1652, simple_loss=0.255, pruned_loss=0.0377, over 2378443.56 frames. ], batch size: 38, lr: 3.46e-03, grad_scale: 8.0 2023-05-17 04:08:44,391 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=281031.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 04:08:57,467 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=281049.0, num_to_drop=1, layers_to_drop={1} 2023-05-17 04:09:00,340 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.7437, 4.3923, 4.4099, 4.6944, 4.5149, 4.6439, 4.6422, 2.4853], device='cuda:0'), covar=tensor([0.0126, 0.0097, 0.0121, 0.0065, 0.0061, 0.0113, 0.0091, 0.0904], device='cuda:0'), in_proj_covar=tensor([0.0073, 0.0083, 0.0087, 0.0077, 0.0064, 0.0098, 0.0086, 0.0102], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-17 04:09:03,866 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=281058.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 04:09:16,621 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.2107, 2.6582, 3.8111, 3.2453, 3.6467, 3.3195, 2.7333, 3.6832], device='cuda:0'), covar=tensor([0.0152, 0.0368, 0.0164, 0.0255, 0.0140, 0.0194, 0.0349, 0.0165], device='cuda:0'), in_proj_covar=tensor([0.0185, 0.0210, 0.0198, 0.0191, 0.0222, 0.0172, 0.0201, 0.0194], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-17 04:09:17,867 INFO [finetune.py:992] (0/2) Epoch 15, batch 9800, loss[loss=0.1603, simple_loss=0.2395, pruned_loss=0.04059, over 11757.00 frames. ], tot_loss[loss=0.166, simple_loss=0.2555, pruned_loss=0.03821, over 2366156.19 frames. ], batch size: 26, lr: 3.46e-03, grad_scale: 8.0 2023-05-17 04:09:18,082 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.6508, 2.8174, 3.8522, 4.6681, 4.1723, 4.6361, 4.1413, 3.5394], device='cuda:0'), covar=tensor([0.0042, 0.0370, 0.0148, 0.0040, 0.0125, 0.0080, 0.0121, 0.0326], device='cuda:0'), in_proj_covar=tensor([0.0091, 0.0125, 0.0108, 0.0081, 0.0107, 0.0119, 0.0102, 0.0142], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-17 04:09:25,045 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.010e+02 2.638e+02 3.044e+02 3.630e+02 1.190e+03, threshold=6.087e+02, percent-clipped=4.0 2023-05-17 04:09:41,493 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=281110.0, num_to_drop=1, layers_to_drop={0} 2023-05-17 04:09:53,391 INFO [finetune.py:992] (0/2) Epoch 15, batch 9850, loss[loss=0.2327, simple_loss=0.3002, pruned_loss=0.08264, over 8198.00 frames. ], tot_loss[loss=0.1658, simple_loss=0.2554, pruned_loss=0.03815, over 2368483.53 frames. ], batch size: 98, lr: 3.46e-03, grad_scale: 8.0 2023-05-17 04:10:21,788 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.4513, 5.2017, 5.3217, 5.3802, 4.9930, 5.0745, 4.7970, 5.3011], device='cuda:0'), covar=tensor([0.0625, 0.0574, 0.0835, 0.0566, 0.2022, 0.1276, 0.0533, 0.1061], device='cuda:0'), in_proj_covar=tensor([0.0566, 0.0727, 0.0641, 0.0661, 0.0887, 0.0767, 0.0593, 0.0500], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0005, 0.0004, 0.0004, 0.0005, 0.0005, 0.0004, 0.0003], device='cuda:0') 2023-05-17 04:10:23,915 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=281169.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 04:10:27,316 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=281174.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 04:10:29,384 INFO [finetune.py:992] (0/2) Epoch 15, batch 9900, loss[loss=0.1745, simple_loss=0.2665, pruned_loss=0.04123, over 11790.00 frames. ], tot_loss[loss=0.166, simple_loss=0.2553, pruned_loss=0.03835, over 2362714.55 frames. ], batch size: 44, lr: 3.46e-03, grad_scale: 8.0 2023-05-17 04:10:36,016 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.627e+02 2.700e+02 3.215e+02 3.866e+02 9.804e+02, threshold=6.430e+02, percent-clipped=3.0 2023-05-17 04:10:37,586 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=281189.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 04:10:43,343 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.9149, 4.2321, 3.6278, 4.4917, 3.9982, 2.7408, 3.8972, 2.8566], device='cuda:0'), covar=tensor([0.0905, 0.0878, 0.1743, 0.0608, 0.1334, 0.1744, 0.1130, 0.3393], device='cuda:0'), in_proj_covar=tensor([0.0308, 0.0384, 0.0363, 0.0327, 0.0372, 0.0276, 0.0348, 0.0368], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-17 04:10:47,832 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=281203.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 04:11:01,818 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=281222.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 04:11:05,367 INFO [finetune.py:992] (0/2) Epoch 15, batch 9950, loss[loss=0.1659, simple_loss=0.2634, pruned_loss=0.03424, over 12364.00 frames. ], tot_loss[loss=0.1669, simple_loss=0.2566, pruned_loss=0.03859, over 2365364.04 frames. ], batch size: 38, lr: 3.46e-03, grad_scale: 16.0 2023-05-17 04:11:07,614 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=281230.0, num_to_drop=1, layers_to_drop={3} 2023-05-17 04:11:12,382 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=281237.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 04:11:19,273 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=3.57 vs. limit=5.0 2023-05-17 04:11:22,404 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=281251.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 04:11:35,317 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.5969, 2.9353, 3.8538, 4.6011, 4.1060, 4.6365, 4.0061, 3.5393], device='cuda:0'), covar=tensor([0.0048, 0.0379, 0.0149, 0.0054, 0.0096, 0.0075, 0.0144, 0.0333], device='cuda:0'), in_proj_covar=tensor([0.0092, 0.0125, 0.0108, 0.0081, 0.0108, 0.0119, 0.0102, 0.0142], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-17 04:11:35,988 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=281270.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 04:11:40,905 INFO [finetune.py:992] (0/2) Epoch 15, batch 10000, loss[loss=0.1454, simple_loss=0.2319, pruned_loss=0.02945, over 12035.00 frames. ], tot_loss[loss=0.1663, simple_loss=0.2563, pruned_loss=0.03814, over 2372873.29 frames. ], batch size: 31, lr: 3.46e-03, grad_scale: 16.0 2023-05-17 04:11:47,836 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.784e+02 2.734e+02 3.185e+02 3.967e+02 7.095e+02, threshold=6.371e+02, percent-clipped=3.0 2023-05-17 04:11:49,420 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([6.2077, 6.1454, 5.9430, 5.4376, 5.2893, 6.1171, 5.7164, 5.4851], device='cuda:0'), covar=tensor([0.0750, 0.1088, 0.0731, 0.1692, 0.0721, 0.0741, 0.1615, 0.1063], device='cuda:0'), in_proj_covar=tensor([0.0643, 0.0573, 0.0525, 0.0648, 0.0424, 0.0736, 0.0794, 0.0580], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-05-17 04:12:09,395 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.7540, 3.8411, 3.2660, 3.2849, 3.1226, 2.9089, 3.8343, 2.5551], device='cuda:0'), covar=tensor([0.0389, 0.0136, 0.0251, 0.0267, 0.0401, 0.0379, 0.0141, 0.0498], device='cuda:0'), in_proj_covar=tensor([0.0197, 0.0167, 0.0172, 0.0193, 0.0207, 0.0204, 0.0177, 0.0208], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-17 04:12:10,688 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=281318.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 04:12:16,991 INFO [finetune.py:992] (0/2) Epoch 15, batch 10050, loss[loss=0.1572, simple_loss=0.247, pruned_loss=0.03368, over 12362.00 frames. ], tot_loss[loss=0.1658, simple_loss=0.2562, pruned_loss=0.03774, over 2374548.01 frames. ], batch size: 38, lr: 3.46e-03, grad_scale: 16.0 2023-05-17 04:12:27,719 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.1436, 2.5834, 3.8167, 3.1361, 3.5644, 3.2294, 2.7517, 3.6155], device='cuda:0'), covar=tensor([0.0166, 0.0358, 0.0133, 0.0250, 0.0147, 0.0203, 0.0360, 0.0133], device='cuda:0'), in_proj_covar=tensor([0.0185, 0.0210, 0.0198, 0.0191, 0.0222, 0.0172, 0.0201, 0.0194], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-17 04:12:39,012 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=281358.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 04:12:48,963 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.4285, 3.4702, 3.0605, 3.1308, 2.7772, 2.6683, 3.5100, 2.3621], device='cuda:0'), covar=tensor([0.0433, 0.0154, 0.0282, 0.0213, 0.0460, 0.0367, 0.0125, 0.0511], device='cuda:0'), in_proj_covar=tensor([0.0197, 0.0167, 0.0172, 0.0193, 0.0207, 0.0204, 0.0177, 0.0208], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-17 04:12:53,037 INFO [finetune.py:992] (0/2) Epoch 15, batch 10100, loss[loss=0.2121, simple_loss=0.3023, pruned_loss=0.06101, over 12147.00 frames. ], tot_loss[loss=0.1654, simple_loss=0.2556, pruned_loss=0.03763, over 2377570.50 frames. ], batch size: 39, lr: 3.46e-03, grad_scale: 16.0 2023-05-17 04:13:00,142 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.043e+02 2.617e+02 3.084e+02 3.596e+02 6.290e+02, threshold=6.168e+02, percent-clipped=0.0 2023-05-17 04:13:12,714 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.8729, 4.3725, 3.7138, 4.6079, 3.9536, 2.6914, 3.8645, 2.8260], device='cuda:0'), covar=tensor([0.0982, 0.0783, 0.1638, 0.0541, 0.1545, 0.1894, 0.1270, 0.3648], device='cuda:0'), in_proj_covar=tensor([0.0311, 0.0386, 0.0365, 0.0329, 0.0375, 0.0279, 0.0351, 0.0370], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-17 04:13:13,224 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=281405.0, num_to_drop=1, layers_to_drop={1} 2023-05-17 04:13:13,936 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=281406.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 04:13:28,379 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=281426.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 04:13:28,947 INFO [finetune.py:992] (0/2) Epoch 15, batch 10150, loss[loss=0.1475, simple_loss=0.231, pruned_loss=0.03202, over 11869.00 frames. ], tot_loss[loss=0.166, simple_loss=0.2559, pruned_loss=0.03806, over 2367260.35 frames. ], batch size: 26, lr: 3.46e-03, grad_scale: 16.0 2023-05-17 04:13:49,557 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=281456.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 04:14:04,824 INFO [finetune.py:992] (0/2) Epoch 15, batch 10200, loss[loss=0.1552, simple_loss=0.2452, pruned_loss=0.0326, over 12076.00 frames. ], tot_loss[loss=0.1664, simple_loss=0.2563, pruned_loss=0.0383, over 2371310.94 frames. ], batch size: 32, lr: 3.46e-03, grad_scale: 16.0 2023-05-17 04:14:11,943 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.850e+02 2.679e+02 3.131e+02 3.626e+02 7.226e+02, threshold=6.261e+02, percent-clipped=3.0 2023-05-17 04:14:12,185 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=281487.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 04:14:23,703 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.1832, 4.8393, 4.8938, 5.0919, 4.9817, 5.0049, 4.9662, 2.7818], device='cuda:0'), covar=tensor([0.0093, 0.0069, 0.0090, 0.0057, 0.0042, 0.0104, 0.0072, 0.0789], device='cuda:0'), in_proj_covar=tensor([0.0072, 0.0083, 0.0086, 0.0077, 0.0063, 0.0097, 0.0085, 0.0101], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-17 04:14:34,151 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=281517.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 04:14:39,713 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=281525.0, num_to_drop=1, layers_to_drop={3} 2023-05-17 04:14:40,859 INFO [finetune.py:992] (0/2) Epoch 15, batch 10250, loss[loss=0.1965, simple_loss=0.2859, pruned_loss=0.05355, over 12069.00 frames. ], tot_loss[loss=0.166, simple_loss=0.2555, pruned_loss=0.03822, over 2369661.70 frames. ], batch size: 42, lr: 3.46e-03, grad_scale: 16.0 2023-05-17 04:15:16,370 INFO [finetune.py:992] (0/2) Epoch 15, batch 10300, loss[loss=0.153, simple_loss=0.2383, pruned_loss=0.03384, over 12242.00 frames. ], tot_loss[loss=0.1658, simple_loss=0.2555, pruned_loss=0.03799, over 2374077.13 frames. ], batch size: 32, lr: 3.46e-03, grad_scale: 16.0 2023-05-17 04:15:23,593 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.931e+02 2.676e+02 3.123e+02 3.776e+02 1.681e+03, threshold=6.246e+02, percent-clipped=5.0 2023-05-17 04:15:48,397 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 2023-05-17 04:15:53,066 INFO [finetune.py:992] (0/2) Epoch 15, batch 10350, loss[loss=0.1751, simple_loss=0.2612, pruned_loss=0.04447, over 12101.00 frames. ], tot_loss[loss=0.1662, simple_loss=0.2561, pruned_loss=0.0382, over 2367412.47 frames. ], batch size: 38, lr: 3.46e-03, grad_scale: 8.0 2023-05-17 04:16:04,504 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.4999, 2.5644, 3.6832, 4.5203, 3.9891, 4.4927, 3.8361, 3.1109], device='cuda:0'), covar=tensor([0.0043, 0.0413, 0.0159, 0.0041, 0.0113, 0.0072, 0.0141, 0.0397], device='cuda:0'), in_proj_covar=tensor([0.0092, 0.0126, 0.0108, 0.0081, 0.0108, 0.0119, 0.0102, 0.0143], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-17 04:16:29,394 INFO [finetune.py:992] (0/2) Epoch 15, batch 10400, loss[loss=0.1706, simple_loss=0.2691, pruned_loss=0.03601, over 12159.00 frames. ], tot_loss[loss=0.1655, simple_loss=0.2555, pruned_loss=0.03772, over 2369075.11 frames. ], batch size: 34, lr: 3.46e-03, grad_scale: 8.0 2023-05-17 04:16:30,932 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.3729, 6.2210, 5.7606, 5.8208, 6.2759, 5.5522, 5.7313, 5.8042], device='cuda:0'), covar=tensor([0.1410, 0.1010, 0.1115, 0.1844, 0.0896, 0.1989, 0.1825, 0.1063], device='cuda:0'), in_proj_covar=tensor([0.0367, 0.0514, 0.0409, 0.0464, 0.0475, 0.0444, 0.0410, 0.0403], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-17 04:16:36,285 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.4706, 3.5961, 3.6023, 4.1851, 3.0539, 3.5018, 2.3083, 3.8050], device='cuda:0'), covar=tensor([0.1332, 0.0853, 0.1203, 0.0774, 0.1039, 0.0723, 0.1807, 0.1097], device='cuda:0'), in_proj_covar=tensor([0.0228, 0.0266, 0.0300, 0.0359, 0.0242, 0.0245, 0.0261, 0.0375], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-17 04:16:37,413 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.674e+02 2.484e+02 2.960e+02 3.585e+02 5.106e+02, threshold=5.920e+02, percent-clipped=0.0 2023-05-17 04:16:41,782 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.2092, 4.1189, 4.1194, 4.5471, 2.9511, 3.8458, 2.4549, 4.0442], device='cuda:0'), covar=tensor([0.1765, 0.0742, 0.0897, 0.0526, 0.1295, 0.0704, 0.2076, 0.1181], device='cuda:0'), in_proj_covar=tensor([0.0227, 0.0266, 0.0300, 0.0358, 0.0242, 0.0245, 0.0261, 0.0374], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-17 04:16:49,653 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=281705.0, num_to_drop=1, layers_to_drop={0} 2023-05-17 04:16:51,917 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.9375, 3.5353, 5.3582, 2.9507, 2.9945, 3.8839, 3.2831, 3.8517], device='cuda:0'), covar=tensor([0.0354, 0.1089, 0.0284, 0.1014, 0.1831, 0.1447, 0.1320, 0.1146], device='cuda:0'), in_proj_covar=tensor([0.0240, 0.0241, 0.0259, 0.0184, 0.0238, 0.0298, 0.0227, 0.0269], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-17 04:17:05,130 INFO [finetune.py:992] (0/2) Epoch 15, batch 10450, loss[loss=0.1621, simple_loss=0.2615, pruned_loss=0.03137, over 12155.00 frames. ], tot_loss[loss=0.1645, simple_loss=0.2547, pruned_loss=0.03717, over 2369985.88 frames. ], batch size: 34, lr: 3.45e-03, grad_scale: 8.0 2023-05-17 04:17:23,907 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=281753.0, num_to_drop=1, layers_to_drop={0} 2023-05-17 04:17:26,651 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.9089, 5.5837, 5.2146, 5.1876, 5.7056, 5.0937, 5.1896, 5.2471], device='cuda:0'), covar=tensor([0.1468, 0.1030, 0.1144, 0.1798, 0.0960, 0.2117, 0.1737, 0.1005], device='cuda:0'), in_proj_covar=tensor([0.0363, 0.0512, 0.0406, 0.0459, 0.0471, 0.0441, 0.0406, 0.0399], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-17 04:17:40,764 INFO [finetune.py:992] (0/2) Epoch 15, batch 10500, loss[loss=0.1603, simple_loss=0.2431, pruned_loss=0.03872, over 12344.00 frames. ], tot_loss[loss=0.1648, simple_loss=0.2551, pruned_loss=0.03726, over 2381968.85 frames. ], batch size: 30, lr: 3.45e-03, grad_scale: 8.0 2023-05-17 04:17:44,495 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=281782.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 04:17:48,814 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.917e+02 2.691e+02 3.266e+02 3.969e+02 1.128e+03, threshold=6.532e+02, percent-clipped=4.0 2023-05-17 04:18:07,095 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=281812.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 04:18:16,330 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=281825.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 04:18:17,623 INFO [finetune.py:992] (0/2) Epoch 15, batch 10550, loss[loss=0.1551, simple_loss=0.2485, pruned_loss=0.03089, over 12106.00 frames. ], tot_loss[loss=0.1641, simple_loss=0.2543, pruned_loss=0.03693, over 2381573.68 frames. ], batch size: 33, lr: 3.45e-03, grad_scale: 8.0 2023-05-17 04:18:32,198 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.3184, 2.4397, 3.6068, 4.3430, 3.8835, 4.2785, 3.7648, 3.0410], device='cuda:0'), covar=tensor([0.0051, 0.0453, 0.0162, 0.0037, 0.0123, 0.0087, 0.0133, 0.0413], device='cuda:0'), in_proj_covar=tensor([0.0091, 0.0125, 0.0107, 0.0080, 0.0107, 0.0118, 0.0102, 0.0141], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-17 04:18:35,684 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=281852.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 04:18:37,254 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.7505, 2.8430, 4.6937, 4.9171, 2.7604, 2.5912, 3.0079, 2.2288], device='cuda:0'), covar=tensor([0.1694, 0.3089, 0.0446, 0.0379, 0.1503, 0.2631, 0.2826, 0.4144], device='cuda:0'), in_proj_covar=tensor([0.0304, 0.0394, 0.0284, 0.0308, 0.0278, 0.0317, 0.0399, 0.0381], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-17 04:18:50,600 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=281873.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 04:18:53,535 INFO [finetune.py:992] (0/2) Epoch 15, batch 10600, loss[loss=0.1733, simple_loss=0.265, pruned_loss=0.04086, over 12352.00 frames. ], tot_loss[loss=0.1654, simple_loss=0.2555, pruned_loss=0.03766, over 2365372.76 frames. ], batch size: 38, lr: 3.45e-03, grad_scale: 8.0 2023-05-17 04:19:01,391 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.816e+02 2.591e+02 2.933e+02 3.502e+02 6.423e+02, threshold=5.866e+02, percent-clipped=0.0 2023-05-17 04:19:18,068 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=281910.0, num_to_drop=1, layers_to_drop={1} 2023-05-17 04:19:20,428 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=281913.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 04:19:29,924 INFO [finetune.py:992] (0/2) Epoch 15, batch 10650, loss[loss=0.1526, simple_loss=0.2472, pruned_loss=0.02899, over 12263.00 frames. ], tot_loss[loss=0.1656, simple_loss=0.2556, pruned_loss=0.03779, over 2365586.34 frames. ], batch size: 37, lr: 3.45e-03, grad_scale: 8.0 2023-05-17 04:19:43,563 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=281945.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 04:20:02,188 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=281971.0, num_to_drop=1, layers_to_drop={0} 2023-05-17 04:20:06,165 INFO [finetune.py:992] (0/2) Epoch 15, batch 10700, loss[loss=0.171, simple_loss=0.2619, pruned_loss=0.04007, over 12038.00 frames. ], tot_loss[loss=0.1655, simple_loss=0.2553, pruned_loss=0.03784, over 2368479.22 frames. ], batch size: 40, lr: 3.45e-03, grad_scale: 8.0 2023-05-17 04:20:14,070 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.806e+02 2.695e+02 3.139e+02 3.686e+02 1.665e+03, threshold=6.278e+02, percent-clipped=5.0 2023-05-17 04:20:22,980 INFO [checkpoint.py:75] (0/2) Saving checkpoint to pruned_transducer_stateless7/exp_giga_finetune/checkpoint-182000.pt 2023-05-17 04:20:30,560 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=282006.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 04:20:34,290 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.3384, 4.8776, 5.3354, 4.6157, 4.9115, 4.6969, 5.3772, 4.9762], device='cuda:0'), covar=tensor([0.0298, 0.0424, 0.0296, 0.0290, 0.0479, 0.0418, 0.0225, 0.0298], device='cuda:0'), in_proj_covar=tensor([0.0274, 0.0281, 0.0302, 0.0273, 0.0277, 0.0277, 0.0249, 0.0224], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-17 04:20:45,553 INFO [finetune.py:992] (0/2) Epoch 15, batch 10750, loss[loss=0.1666, simple_loss=0.2629, pruned_loss=0.03516, over 12143.00 frames. ], tot_loss[loss=0.1648, simple_loss=0.2547, pruned_loss=0.03745, over 2372198.36 frames. ], batch size: 34, lr: 3.45e-03, grad_scale: 8.0 2023-05-17 04:21:21,462 INFO [finetune.py:992] (0/2) Epoch 15, batch 10800, loss[loss=0.1754, simple_loss=0.2779, pruned_loss=0.03639, over 12149.00 frames. ], tot_loss[loss=0.1656, simple_loss=0.2553, pruned_loss=0.03792, over 2354338.13 frames. ], batch size: 36, lr: 3.45e-03, grad_scale: 8.0 2023-05-17 04:21:25,267 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=282082.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 04:21:30,150 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.997e+02 2.673e+02 3.156e+02 3.969e+02 9.157e+02, threshold=6.313e+02, percent-clipped=4.0 2023-05-17 04:21:35,931 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=282096.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 04:21:47,205 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=282112.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 04:21:57,547 INFO [finetune.py:992] (0/2) Epoch 15, batch 10850, loss[loss=0.1498, simple_loss=0.2465, pruned_loss=0.02654, over 12116.00 frames. ], tot_loss[loss=0.1653, simple_loss=0.2549, pruned_loss=0.03785, over 2369656.69 frames. ], batch size: 39, lr: 3.45e-03, grad_scale: 8.0 2023-05-17 04:21:59,793 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=282130.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 04:22:19,620 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=282157.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 04:22:21,653 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=282160.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 04:22:34,239 INFO [finetune.py:992] (0/2) Epoch 15, batch 10900, loss[loss=0.1681, simple_loss=0.2625, pruned_loss=0.03684, over 12299.00 frames. ], tot_loss[loss=0.1651, simple_loss=0.2547, pruned_loss=0.03772, over 2375241.42 frames. ], batch size: 33, lr: 3.45e-03, grad_scale: 8.0 2023-05-17 04:22:42,559 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.986e+02 2.563e+02 2.975e+02 3.620e+02 1.018e+03, threshold=5.950e+02, percent-clipped=3.0 2023-05-17 04:22:57,236 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=282208.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 04:23:10,308 INFO [finetune.py:992] (0/2) Epoch 15, batch 10950, loss[loss=0.1951, simple_loss=0.2829, pruned_loss=0.05362, over 12371.00 frames. ], tot_loss[loss=0.166, simple_loss=0.2557, pruned_loss=0.03815, over 2373419.14 frames. ], batch size: 38, lr: 3.45e-03, grad_scale: 8.0 2023-05-17 04:23:12,034 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.1849, 4.2026, 4.2036, 4.6108, 3.3338, 4.0257, 2.8210, 4.3714], device='cuda:0'), covar=tensor([0.1689, 0.0661, 0.0868, 0.0610, 0.0988, 0.0578, 0.1676, 0.1013], device='cuda:0'), in_proj_covar=tensor([0.0229, 0.0268, 0.0301, 0.0361, 0.0243, 0.0245, 0.0263, 0.0377], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-17 04:23:14,146 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.2567, 4.8497, 5.2545, 4.5868, 4.8390, 4.6413, 5.2855, 4.9336], device='cuda:0'), covar=tensor([0.0298, 0.0383, 0.0278, 0.0282, 0.0477, 0.0393, 0.0208, 0.0351], device='cuda:0'), in_proj_covar=tensor([0.0272, 0.0279, 0.0299, 0.0271, 0.0276, 0.0274, 0.0248, 0.0222], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-17 04:23:26,664 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.0435, 2.3169, 3.4379, 4.1399, 3.7537, 4.0631, 3.6340, 2.7615], device='cuda:0'), covar=tensor([0.0059, 0.0457, 0.0163, 0.0051, 0.0116, 0.0093, 0.0141, 0.0438], device='cuda:0'), in_proj_covar=tensor([0.0090, 0.0124, 0.0106, 0.0080, 0.0106, 0.0118, 0.0101, 0.0140], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-17 04:23:29,601 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.8269, 2.4220, 3.4026, 3.9584, 3.6661, 3.9846, 3.5610, 2.7926], device='cuda:0'), covar=tensor([0.0071, 0.0413, 0.0168, 0.0062, 0.0122, 0.0080, 0.0164, 0.0417], device='cuda:0'), in_proj_covar=tensor([0.0090, 0.0124, 0.0106, 0.0080, 0.0106, 0.0118, 0.0101, 0.0140], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-17 04:23:38,818 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=282266.0, num_to_drop=1, layers_to_drop={0} 2023-05-17 04:23:46,254 INFO [finetune.py:992] (0/2) Epoch 15, batch 11000, loss[loss=0.181, simple_loss=0.2676, pruned_loss=0.04726, over 12130.00 frames. ], tot_loss[loss=0.1687, simple_loss=0.258, pruned_loss=0.03969, over 2344648.86 frames. ], batch size: 38, lr: 3.45e-03, grad_scale: 8.0 2023-05-17 04:23:54,116 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.807e+02 2.897e+02 3.408e+02 4.046e+02 8.329e+02, threshold=6.815e+02, percent-clipped=5.0 2023-05-17 04:24:03,442 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=282301.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 04:24:22,501 INFO [finetune.py:992] (0/2) Epoch 15, batch 11050, loss[loss=0.1533, simple_loss=0.243, pruned_loss=0.03183, over 12029.00 frames. ], tot_loss[loss=0.1713, simple_loss=0.2609, pruned_loss=0.04085, over 2322980.70 frames. ], batch size: 31, lr: 3.45e-03, grad_scale: 8.0 2023-05-17 04:24:50,562 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.0722, 3.6845, 3.7605, 4.2352, 2.8110, 3.6944, 2.5635, 3.6684], device='cuda:0'), covar=tensor([0.1766, 0.0895, 0.0840, 0.0549, 0.1264, 0.0736, 0.1824, 0.1011], device='cuda:0'), in_proj_covar=tensor([0.0229, 0.0267, 0.0301, 0.0360, 0.0243, 0.0244, 0.0261, 0.0375], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-17 04:24:58,050 INFO [finetune.py:992] (0/2) Epoch 15, batch 11100, loss[loss=0.2555, simple_loss=0.3447, pruned_loss=0.08315, over 10275.00 frames. ], tot_loss[loss=0.1765, simple_loss=0.2655, pruned_loss=0.04378, over 2259317.02 frames. ], batch size: 68, lr: 3.45e-03, grad_scale: 8.0 2023-05-17 04:25:05,856 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.138e+02 2.930e+02 3.609e+02 4.369e+02 7.372e+02, threshold=7.218e+02, percent-clipped=3.0 2023-05-17 04:25:08,412 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.40 vs. limit=2.0 2023-05-17 04:25:33,894 INFO [finetune.py:992] (0/2) Epoch 15, batch 11150, loss[loss=0.2286, simple_loss=0.3237, pruned_loss=0.0667, over 10182.00 frames. ], tot_loss[loss=0.1832, simple_loss=0.2717, pruned_loss=0.04733, over 2203104.35 frames. ], batch size: 68, lr: 3.45e-03, grad_scale: 8.0 2023-05-17 04:25:51,896 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=282452.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 04:26:00,347 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.2333, 3.1127, 3.1581, 3.4573, 2.5988, 3.1155, 2.6524, 2.9038], device='cuda:0'), covar=tensor([0.1508, 0.0952, 0.0770, 0.0552, 0.1038, 0.0818, 0.1493, 0.0601], device='cuda:0'), in_proj_covar=tensor([0.0226, 0.0265, 0.0297, 0.0356, 0.0240, 0.0242, 0.0259, 0.0372], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-17 04:26:08,905 INFO [finetune.py:992] (0/2) Epoch 15, batch 11200, loss[loss=0.2414, simple_loss=0.3234, pruned_loss=0.07969, over 12056.00 frames. ], tot_loss[loss=0.1905, simple_loss=0.2779, pruned_loss=0.05153, over 2140348.68 frames. ], batch size: 42, lr: 3.45e-03, grad_scale: 8.0 2023-05-17 04:26:09,997 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.99 vs. limit=2.0 2023-05-17 04:26:17,126 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.224e+02 3.308e+02 3.948e+02 4.776e+02 9.685e+02, threshold=7.896e+02, percent-clipped=3.0 2023-05-17 04:26:18,027 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=282489.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 04:26:23,739 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 2023-05-17 04:26:31,732 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=282508.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 04:26:44,875 INFO [finetune.py:992] (0/2) Epoch 15, batch 11250, loss[loss=0.2627, simple_loss=0.3303, pruned_loss=0.09759, over 6838.00 frames. ], tot_loss[loss=0.1965, simple_loss=0.2831, pruned_loss=0.05493, over 2089985.84 frames. ], batch size: 99, lr: 3.45e-03, grad_scale: 8.0 2023-05-17 04:27:01,258 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=282550.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 04:27:04,180 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 2023-05-17 04:27:05,972 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=282556.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 04:27:12,882 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=282566.0, num_to_drop=1, layers_to_drop={2} 2023-05-17 04:27:17,711 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.6698, 2.2351, 3.0142, 2.6521, 2.8596, 2.8396, 2.2371, 2.9790], device='cuda:0'), covar=tensor([0.0111, 0.0335, 0.0099, 0.0197, 0.0161, 0.0166, 0.0327, 0.0123], device='cuda:0'), in_proj_covar=tensor([0.0181, 0.0207, 0.0195, 0.0188, 0.0219, 0.0169, 0.0199, 0.0191], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-17 04:27:20,973 INFO [finetune.py:992] (0/2) Epoch 15, batch 11300, loss[loss=0.2257, simple_loss=0.3199, pruned_loss=0.06573, over 10217.00 frames. ], tot_loss[loss=0.2049, simple_loss=0.2902, pruned_loss=0.05981, over 2016746.78 frames. ], batch size: 68, lr: 3.45e-03, grad_scale: 8.0 2023-05-17 04:27:21,113 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.4898, 4.4466, 4.3306, 3.9713, 4.1175, 4.4314, 4.1524, 4.0269], device='cuda:0'), covar=tensor([0.0841, 0.0993, 0.0695, 0.1448, 0.1922, 0.0807, 0.1498, 0.0988], device='cuda:0'), in_proj_covar=tensor([0.0629, 0.0558, 0.0510, 0.0631, 0.0414, 0.0718, 0.0772, 0.0567], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002], device='cuda:0') 2023-05-17 04:27:28,506 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.112e+02 3.356e+02 3.951e+02 4.972e+02 8.345e+02, threshold=7.901e+02, percent-clipped=2.0 2023-05-17 04:27:37,975 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=282601.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 04:27:46,830 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=282614.0, num_to_drop=1, layers_to_drop={1} 2023-05-17 04:27:50,208 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.6853, 3.2230, 3.5395, 3.5901, 3.5763, 3.6957, 3.4995, 2.8493], device='cuda:0'), covar=tensor([0.0119, 0.0144, 0.0139, 0.0090, 0.0079, 0.0115, 0.0094, 0.0652], device='cuda:0'), in_proj_covar=tensor([0.0070, 0.0080, 0.0084, 0.0074, 0.0061, 0.0095, 0.0083, 0.0098], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-17 04:27:55,934 INFO [finetune.py:992] (0/2) Epoch 15, batch 11350, loss[loss=0.2112, simple_loss=0.2993, pruned_loss=0.06156, over 10522.00 frames. ], tot_loss[loss=0.2106, simple_loss=0.2951, pruned_loss=0.06301, over 1961533.39 frames. ], batch size: 68, lr: 3.45e-03, grad_scale: 8.0 2023-05-17 04:27:56,856 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.9055, 2.3234, 3.2288, 2.8884, 3.0578, 3.0561, 2.3733, 3.2112], device='cuda:0'), covar=tensor([0.0117, 0.0341, 0.0084, 0.0220, 0.0152, 0.0150, 0.0329, 0.0110], device='cuda:0'), in_proj_covar=tensor([0.0180, 0.0205, 0.0193, 0.0187, 0.0217, 0.0167, 0.0197, 0.0189], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-17 04:28:11,638 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=282649.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 04:28:17,881 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=282658.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 04:28:27,967 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.8805, 4.4730, 4.0909, 4.1518, 4.5098, 3.8906, 4.1078, 3.9985], device='cuda:0'), covar=tensor([0.1526, 0.1029, 0.1276, 0.1880, 0.0961, 0.2153, 0.1664, 0.1339], device='cuda:0'), in_proj_covar=tensor([0.0353, 0.0495, 0.0397, 0.0445, 0.0459, 0.0429, 0.0394, 0.0387], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-17 04:28:30,526 INFO [finetune.py:992] (0/2) Epoch 15, batch 11400, loss[loss=0.2505, simple_loss=0.3227, pruned_loss=0.08918, over 7513.00 frames. ], tot_loss[loss=0.2145, simple_loss=0.2985, pruned_loss=0.06529, over 1919887.21 frames. ], batch size: 98, lr: 3.45e-03, grad_scale: 8.0 2023-05-17 04:28:30,742 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=282677.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 04:28:37,840 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.754e+02 3.597e+02 4.130e+02 4.941e+02 1.183e+03, threshold=8.259e+02, percent-clipped=2.0 2023-05-17 04:28:45,493 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=282698.0, num_to_drop=1, layers_to_drop={1} 2023-05-17 04:29:00,674 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=282719.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 04:29:05,962 INFO [finetune.py:992] (0/2) Epoch 15, batch 11450, loss[loss=0.2051, simple_loss=0.2944, pruned_loss=0.0579, over 10284.00 frames. ], tot_loss[loss=0.2185, simple_loss=0.3015, pruned_loss=0.06777, over 1892295.82 frames. ], batch size: 68, lr: 3.45e-03, grad_scale: 8.0 2023-05-17 04:29:11,764 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=4.44 vs. limit=5.0 2023-05-17 04:29:13,465 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=282738.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 04:29:22,631 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=282752.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 04:29:27,894 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=282759.0, num_to_drop=1, layers_to_drop={2} 2023-05-17 04:29:40,754 INFO [finetune.py:992] (0/2) Epoch 15, batch 11500, loss[loss=0.2071, simple_loss=0.3001, pruned_loss=0.05698, over 11208.00 frames. ], tot_loss[loss=0.2236, simple_loss=0.3052, pruned_loss=0.07098, over 1836140.27 frames. ], batch size: 55, lr: 3.45e-03, grad_scale: 8.0 2023-05-17 04:29:48,009 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.245e+02 3.457e+02 4.104e+02 4.955e+02 2.185e+03, threshold=8.208e+02, percent-clipped=6.0 2023-05-17 04:29:56,379 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=282800.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 04:30:00,017 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=282805.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 04:30:15,608 INFO [finetune.py:992] (0/2) Epoch 15, batch 11550, loss[loss=0.2657, simple_loss=0.3254, pruned_loss=0.103, over 6958.00 frames. ], tot_loss[loss=0.2263, simple_loss=0.307, pruned_loss=0.07276, over 1797098.17 frames. ], batch size: 98, lr: 3.45e-03, grad_scale: 8.0 2023-05-17 04:30:23,784 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.5843, 5.5619, 5.3735, 4.9438, 4.8637, 5.5006, 5.1939, 5.0314], device='cuda:0'), covar=tensor([0.0712, 0.0793, 0.0615, 0.1601, 0.0951, 0.0642, 0.1435, 0.0794], device='cuda:0'), in_proj_covar=tensor([0.0612, 0.0546, 0.0498, 0.0613, 0.0403, 0.0700, 0.0750, 0.0551], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002], device='cuda:0') 2023-05-17 04:30:28,559 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=282845.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 04:30:42,826 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=282866.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 04:30:42,987 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-05-17 04:30:49,824 INFO [finetune.py:992] (0/2) Epoch 15, batch 11600, loss[loss=0.1943, simple_loss=0.2842, pruned_loss=0.0522, over 9977.00 frames. ], tot_loss[loss=0.2288, simple_loss=0.3088, pruned_loss=0.07443, over 1776401.95 frames. ], batch size: 68, lr: 3.45e-03, grad_scale: 8.0 2023-05-17 04:30:58,120 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.453e+02 3.489e+02 4.044e+02 4.459e+02 7.552e+02, threshold=8.088e+02, percent-clipped=0.0 2023-05-17 04:31:12,531 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 2023-05-17 04:31:19,373 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.38 vs. limit=2.0 2023-05-17 04:31:27,094 INFO [finetune.py:992] (0/2) Epoch 15, batch 11650, loss[loss=0.2391, simple_loss=0.3079, pruned_loss=0.08512, over 6606.00 frames. ], tot_loss[loss=0.2274, simple_loss=0.3074, pruned_loss=0.07371, over 1774443.42 frames. ], batch size: 98, lr: 3.45e-03, grad_scale: 8.0 2023-05-17 04:31:30,508 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.36 vs. limit=2.0 2023-05-17 04:32:02,106 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.53 vs. limit=2.0 2023-05-17 04:32:02,323 INFO [finetune.py:992] (0/2) Epoch 15, batch 11700, loss[loss=0.1931, simple_loss=0.2828, pruned_loss=0.05172, over 10333.00 frames. ], tot_loss[loss=0.2262, simple_loss=0.3059, pruned_loss=0.07327, over 1765621.97 frames. ], batch size: 68, lr: 3.45e-03, grad_scale: 8.0 2023-05-17 04:32:09,581 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.623e+02 3.290e+02 3.801e+02 4.370e+02 9.229e+02, threshold=7.602e+02, percent-clipped=4.0 2023-05-17 04:32:20,454 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.61 vs. limit=2.0 2023-05-17 04:32:28,144 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=283014.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 04:32:37,596 INFO [finetune.py:992] (0/2) Epoch 15, batch 11750, loss[loss=0.2096, simple_loss=0.2904, pruned_loss=0.06441, over 6947.00 frames. ], tot_loss[loss=0.2274, simple_loss=0.3065, pruned_loss=0.07413, over 1751205.29 frames. ], batch size: 97, lr: 3.45e-03, grad_scale: 8.0 2023-05-17 04:32:41,907 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=283033.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 04:32:55,931 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=283054.0, num_to_drop=1, layers_to_drop={0} 2023-05-17 04:32:59,500 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 2023-05-17 04:33:12,090 INFO [finetune.py:992] (0/2) Epoch 15, batch 11800, loss[loss=0.2678, simple_loss=0.3336, pruned_loss=0.101, over 6259.00 frames. ], tot_loss[loss=0.2305, simple_loss=0.3088, pruned_loss=0.07609, over 1719360.85 frames. ], batch size: 99, lr: 3.45e-03, grad_scale: 8.0 2023-05-17 04:33:20,190 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.485e+02 3.503e+02 4.210e+02 5.100e+02 7.921e+02, threshold=8.420e+02, percent-clipped=1.0 2023-05-17 04:33:38,618 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=283115.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 04:33:47,132 INFO [finetune.py:992] (0/2) Epoch 15, batch 11850, loss[loss=0.1917, simple_loss=0.291, pruned_loss=0.04622, over 10132.00 frames. ], tot_loss[loss=0.2326, simple_loss=0.3109, pruned_loss=0.07716, over 1687764.67 frames. ], batch size: 68, lr: 3.44e-03, grad_scale: 8.0 2023-05-17 04:33:58,950 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=283144.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 04:34:00,248 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=283145.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 04:34:10,722 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.36 vs. limit=2.0 2023-05-17 04:34:11,027 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=283161.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 04:34:21,549 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=283176.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 04:34:22,044 INFO [finetune.py:992] (0/2) Epoch 15, batch 11900, loss[loss=0.2333, simple_loss=0.3068, pruned_loss=0.07988, over 6767.00 frames. ], tot_loss[loss=0.2303, simple_loss=0.3095, pruned_loss=0.07553, over 1678646.00 frames. ], batch size: 98, lr: 3.44e-03, grad_scale: 8.0 2023-05-17 04:34:23,588 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.8414, 3.8200, 3.8106, 3.9114, 3.7124, 3.7067, 3.6662, 3.8008], device='cuda:0'), covar=tensor([0.1409, 0.0769, 0.1335, 0.0761, 0.1651, 0.1394, 0.0592, 0.1162], device='cuda:0'), in_proj_covar=tensor([0.0512, 0.0660, 0.0580, 0.0601, 0.0790, 0.0700, 0.0539, 0.0458], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0003], device='cuda:0') 2023-05-17 04:34:30,115 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.255e+02 3.323e+02 3.837e+02 4.368e+02 2.807e+03, threshold=7.674e+02, percent-clipped=2.0 2023-05-17 04:34:33,649 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=283193.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 04:34:42,751 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=283205.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 04:34:49,160 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.93 vs. limit=2.0 2023-05-17 04:34:57,386 INFO [finetune.py:992] (0/2) Epoch 15, batch 11950, loss[loss=0.1726, simple_loss=0.266, pruned_loss=0.03954, over 10400.00 frames. ], tot_loss[loss=0.2258, simple_loss=0.3061, pruned_loss=0.07274, over 1673174.53 frames. ], batch size: 69, lr: 3.44e-03, grad_scale: 8.0 2023-05-17 04:35:28,139 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.1034, 4.9601, 5.0774, 5.0929, 4.7847, 4.8590, 4.7077, 4.9298], device='cuda:0'), covar=tensor([0.0684, 0.0578, 0.0793, 0.0574, 0.1769, 0.1199, 0.0480, 0.1303], device='cuda:0'), in_proj_covar=tensor([0.0509, 0.0658, 0.0578, 0.0600, 0.0788, 0.0698, 0.0537, 0.0456], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0003], device='cuda:0') 2023-05-17 04:35:32,770 INFO [finetune.py:992] (0/2) Epoch 15, batch 12000, loss[loss=0.2019, simple_loss=0.2847, pruned_loss=0.0595, over 7123.00 frames. ], tot_loss[loss=0.2199, simple_loss=0.3017, pruned_loss=0.06905, over 1679134.30 frames. ], batch size: 99, lr: 3.44e-03, grad_scale: 8.0 2023-05-17 04:35:32,770 INFO [finetune.py:1017] (0/2) Computing validation loss 2023-05-17 04:35:50,350 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.7817, 2.4062, 3.5341, 3.9738, 3.7642, 3.7598, 3.7816, 2.3927], device='cuda:0'), covar=tensor([0.0060, 0.0437, 0.0143, 0.0046, 0.0100, 0.0129, 0.0099, 0.0572], device='cuda:0'), in_proj_covar=tensor([0.0087, 0.0120, 0.0102, 0.0076, 0.0101, 0.0114, 0.0097, 0.0135], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-17 04:35:50,737 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.4619, 3.2292, 3.3887, 3.4562, 3.1560, 3.5103, 3.4495, 3.5233], device='cuda:0'), covar=tensor([0.0365, 0.0273, 0.0226, 0.0422, 0.0618, 0.0422, 0.0244, 0.0418], device='cuda:0'), in_proj_covar=tensor([0.0180, 0.0178, 0.0175, 0.0225, 0.0218, 0.0202, 0.0161, 0.0214], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:0') 2023-05-17 04:35:51,766 INFO [finetune.py:1026] (0/2) Epoch 15, validation: loss=0.2887, simple_loss=0.3627, pruned_loss=0.1073, over 1020973.00 frames. 2023-05-17 04:35:51,767 INFO [finetune.py:1027] (0/2) Maximum memory allocated so far is 12827MB 2023-05-17 04:35:56,564 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=283283.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 04:35:59,709 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.902e+02 2.961e+02 3.359e+02 3.925e+02 1.066e+03, threshold=6.717e+02, percent-clipped=1.0 2023-05-17 04:36:06,103 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.85 vs. limit=2.0 2023-05-17 04:36:11,256 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.8781, 2.2467, 2.7547, 2.9082, 2.9387, 3.0068, 2.8387, 2.4866], device='cuda:0'), covar=tensor([0.0098, 0.0384, 0.0209, 0.0090, 0.0123, 0.0107, 0.0151, 0.0359], device='cuda:0'), in_proj_covar=tensor([0.0086, 0.0120, 0.0101, 0.0076, 0.0101, 0.0113, 0.0097, 0.0134], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-17 04:36:18,011 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=283314.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 04:36:26,863 INFO [finetune.py:992] (0/2) Epoch 15, batch 12050, loss[loss=0.1853, simple_loss=0.2771, pruned_loss=0.04675, over 11154.00 frames. ], tot_loss[loss=0.2149, simple_loss=0.2975, pruned_loss=0.06616, over 1687119.39 frames. ], batch size: 55, lr: 3.44e-03, grad_scale: 8.0 2023-05-17 04:36:31,033 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=283333.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 04:36:38,306 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=283344.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 04:36:45,175 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=283354.0, num_to_drop=1, layers_to_drop={0} 2023-05-17 04:36:50,273 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=283362.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 04:36:59,659 INFO [finetune.py:992] (0/2) Epoch 15, batch 12100, loss[loss=0.233, simple_loss=0.3081, pruned_loss=0.07896, over 6731.00 frames. ], tot_loss[loss=0.2137, simple_loss=0.2968, pruned_loss=0.0653, over 1685350.23 frames. ], batch size: 98, lr: 3.44e-03, grad_scale: 8.0 2023-05-17 04:36:59,983 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 2023-05-17 04:37:02,296 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=283381.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 04:37:06,551 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.150e+02 3.015e+02 3.368e+02 3.745e+02 9.475e+02, threshold=6.736e+02, percent-clipped=5.0 2023-05-17 04:37:15,727 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=283402.0, num_to_drop=1, layers_to_drop={0} 2023-05-17 04:37:31,952 INFO [finetune.py:992] (0/2) Epoch 15, batch 12150, loss[loss=0.226, simple_loss=0.2952, pruned_loss=0.07838, over 6803.00 frames. ], tot_loss[loss=0.2144, simple_loss=0.2973, pruned_loss=0.06575, over 1683542.10 frames. ], batch size: 98, lr: 3.44e-03, grad_scale: 8.0 2023-05-17 04:37:42,994 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 2023-05-17 04:37:53,434 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=283461.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 04:37:53,652 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=5.00 vs. limit=5.0 2023-05-17 04:37:59,571 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=283471.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 04:38:03,196 INFO [finetune.py:992] (0/2) Epoch 15, batch 12200, loss[loss=0.2196, simple_loss=0.2941, pruned_loss=0.07254, over 7097.00 frames. ], tot_loss[loss=0.2158, simple_loss=0.2984, pruned_loss=0.06662, over 1683604.41 frames. ], batch size: 99, lr: 3.44e-03, grad_scale: 8.0 2023-05-17 04:38:10,053 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.356e+02 3.265e+02 3.898e+02 4.626e+02 7.846e+02, threshold=7.795e+02, percent-clipped=3.0 2023-05-17 04:38:17,590 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=283500.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 04:38:18,982 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.0383, 2.3304, 2.6958, 3.0049, 2.2557, 3.0958, 3.0187, 3.1153], device='cuda:0'), covar=tensor([0.0193, 0.1085, 0.0467, 0.0216, 0.1127, 0.0373, 0.0327, 0.0165], device='cuda:0'), in_proj_covar=tensor([0.0116, 0.0197, 0.0176, 0.0117, 0.0184, 0.0172, 0.0166, 0.0119], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-17 04:38:23,172 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=283509.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 04:38:25,301 INFO [checkpoint.py:75] (0/2) Saving checkpoint to pruned_transducer_stateless7/exp_giga_finetune/epoch-15.pt 2023-05-17 04:38:48,074 INFO [finetune.py:992] (0/2) Epoch 16, batch 0, loss[loss=0.1938, simple_loss=0.2852, pruned_loss=0.05115, over 12036.00 frames. ], tot_loss[loss=0.1938, simple_loss=0.2852, pruned_loss=0.05115, over 12036.00 frames. ], batch size: 40, lr: 3.44e-03, grad_scale: 8.0 2023-05-17 04:38:48,075 INFO [finetune.py:1017] (0/2) Computing validation loss 2023-05-17 04:39:03,580 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([1.9549, 1.7260, 1.9499, 1.8621, 2.0340, 2.0602, 1.6305, 1.9417], device='cuda:0'), covar=tensor([0.0116, 0.0257, 0.0097, 0.0150, 0.0136, 0.0136, 0.0258, 0.0113], device='cuda:0'), in_proj_covar=tensor([0.0170, 0.0198, 0.0180, 0.0177, 0.0205, 0.0159, 0.0189, 0.0179], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-17 04:39:05,208 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([1.9918, 1.6529, 2.0321, 1.9090, 2.0615, 2.1224, 1.6683, 2.0428], device='cuda:0'), covar=tensor([0.0130, 0.0342, 0.0118, 0.0183, 0.0154, 0.0136, 0.0332, 0.0120], device='cuda:0'), in_proj_covar=tensor([0.0170, 0.0198, 0.0180, 0.0177, 0.0205, 0.0159, 0.0189, 0.0179], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-17 04:39:06,183 INFO [finetune.py:1026] (0/2) Epoch 16, validation: loss=0.2859, simple_loss=0.3611, pruned_loss=0.1054, over 1020973.00 frames. 2023-05-17 04:39:06,183 INFO [finetune.py:1027] (0/2) Maximum memory allocated so far is 12827MB 2023-05-17 04:39:29,632 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=283544.0, num_to_drop=1, layers_to_drop={0} 2023-05-17 04:39:41,582 INFO [finetune.py:992] (0/2) Epoch 16, batch 50, loss[loss=0.1745, simple_loss=0.2616, pruned_loss=0.04374, over 12111.00 frames. ], tot_loss[loss=0.1744, simple_loss=0.2641, pruned_loss=0.04232, over 534046.86 frames. ], batch size: 33, lr: 3.44e-03, grad_scale: 8.0 2023-05-17 04:39:43,475 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.82 vs. limit=2.0 2023-05-17 04:39:46,103 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=4.79 vs. limit=5.0 2023-05-17 04:40:01,189 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.077e+02 2.943e+02 3.465e+02 4.160e+02 7.410e+02, threshold=6.930e+02, percent-clipped=0.0 2023-05-17 04:40:14,102 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=283605.0, num_to_drop=1, layers_to_drop={1} 2023-05-17 04:40:18,288 INFO [finetune.py:992] (0/2) Epoch 16, batch 100, loss[loss=0.1701, simple_loss=0.272, pruned_loss=0.0341, over 12145.00 frames. ], tot_loss[loss=0.1707, simple_loss=0.2612, pruned_loss=0.04009, over 941446.59 frames. ], batch size: 36, lr: 3.44e-03, grad_scale: 16.0 2023-05-17 04:40:22,131 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.0649, 4.9559, 4.8121, 4.8596, 4.5926, 4.9495, 4.8678, 5.1620], device='cuda:0'), covar=tensor([0.0219, 0.0179, 0.0212, 0.0379, 0.0763, 0.0413, 0.0182, 0.0203], device='cuda:0'), in_proj_covar=tensor([0.0180, 0.0179, 0.0176, 0.0227, 0.0220, 0.0203, 0.0162, 0.0215], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:0') 2023-05-17 04:40:38,315 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=283639.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 04:40:53,906 INFO [finetune.py:992] (0/2) Epoch 16, batch 150, loss[loss=0.155, simple_loss=0.2446, pruned_loss=0.03274, over 12291.00 frames. ], tot_loss[loss=0.1699, simple_loss=0.2601, pruned_loss=0.03982, over 1264993.70 frames. ], batch size: 33, lr: 3.44e-03, grad_scale: 16.0 2023-05-17 04:41:13,868 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.118e+02 2.555e+02 2.911e+02 3.555e+02 5.516e+02, threshold=5.823e+02, percent-clipped=0.0 2023-05-17 04:41:22,860 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.1747, 2.4685, 3.7235, 3.1508, 3.5899, 3.2524, 2.5852, 3.6349], device='cuda:0'), covar=tensor([0.0162, 0.0435, 0.0132, 0.0253, 0.0174, 0.0199, 0.0411, 0.0156], device='cuda:0'), in_proj_covar=tensor([0.0175, 0.0203, 0.0186, 0.0183, 0.0211, 0.0163, 0.0195, 0.0184], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-17 04:41:30,406 INFO [finetune.py:992] (0/2) Epoch 16, batch 200, loss[loss=0.1974, simple_loss=0.2844, pruned_loss=0.0552, over 12055.00 frames. ], tot_loss[loss=0.1687, simple_loss=0.2585, pruned_loss=0.0394, over 1517828.80 frames. ], batch size: 42, lr: 3.44e-03, grad_scale: 16.0 2023-05-17 04:41:54,206 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.36 vs. limit=2.0 2023-05-17 04:42:03,860 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.7084, 2.7312, 4.7479, 4.9818, 2.7859, 2.5588, 2.9849, 2.1626], device='cuda:0'), covar=tensor([0.1868, 0.3434, 0.0455, 0.0367, 0.1564, 0.2933, 0.3096, 0.4711], device='cuda:0'), in_proj_covar=tensor([0.0299, 0.0386, 0.0274, 0.0297, 0.0273, 0.0312, 0.0391, 0.0377], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-17 04:42:05,148 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.3814, 4.9765, 5.3497, 4.6207, 4.9928, 4.7547, 5.3980, 5.0730], device='cuda:0'), covar=tensor([0.0350, 0.0488, 0.0406, 0.0329, 0.0512, 0.0425, 0.0314, 0.0297], device='cuda:0'), in_proj_covar=tensor([0.0260, 0.0265, 0.0285, 0.0258, 0.0263, 0.0261, 0.0238, 0.0209], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-17 04:42:06,344 INFO [finetune.py:992] (0/2) Epoch 16, batch 250, loss[loss=0.1372, simple_loss=0.2319, pruned_loss=0.02127, over 12123.00 frames. ], tot_loss[loss=0.1678, simple_loss=0.2576, pruned_loss=0.03897, over 1719990.85 frames. ], batch size: 30, lr: 3.44e-03, grad_scale: 16.0 2023-05-17 04:42:13,673 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=283771.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 04:42:25,827 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.730e+02 2.578e+02 3.087e+02 3.629e+02 5.392e+02, threshold=6.174e+02, percent-clipped=0.0 2023-05-17 04:42:31,151 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.3512, 4.8063, 2.9595, 2.6774, 4.1719, 2.4608, 4.0211, 3.2872], device='cuda:0'), covar=tensor([0.0703, 0.0480, 0.1142, 0.1529, 0.0264, 0.1494, 0.0405, 0.0793], device='cuda:0'), in_proj_covar=tensor([0.0181, 0.0245, 0.0171, 0.0195, 0.0137, 0.0177, 0.0188, 0.0167], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-17 04:42:34,749 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=283800.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 04:42:42,627 INFO [finetune.py:992] (0/2) Epoch 16, batch 300, loss[loss=0.155, simple_loss=0.2528, pruned_loss=0.02858, over 12316.00 frames. ], tot_loss[loss=0.1666, simple_loss=0.2564, pruned_loss=0.03835, over 1871459.30 frames. ], batch size: 34, lr: 3.44e-03, grad_scale: 16.0 2023-05-17 04:42:48,460 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=283819.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 04:42:53,677 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-05-17 04:43:05,282 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.43 vs. limit=2.0 2023-05-17 04:43:05,623 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=283842.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 04:43:09,855 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=283848.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 04:43:19,236 INFO [finetune.py:992] (0/2) Epoch 16, batch 350, loss[loss=0.1208, simple_loss=0.2082, pruned_loss=0.01664, over 12184.00 frames. ], tot_loss[loss=0.1657, simple_loss=0.2556, pruned_loss=0.03794, over 1989014.58 frames. ], batch size: 29, lr: 3.44e-03, grad_scale: 16.0 2023-05-17 04:43:39,049 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.908e+02 2.698e+02 3.166e+02 3.611e+02 6.102e+02, threshold=6.331e+02, percent-clipped=0.0 2023-05-17 04:43:47,829 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=283900.0, num_to_drop=1, layers_to_drop={0} 2023-05-17 04:43:49,962 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=283903.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 04:43:55,564 INFO [finetune.py:992] (0/2) Epoch 16, batch 400, loss[loss=0.159, simple_loss=0.2476, pruned_loss=0.0352, over 12353.00 frames. ], tot_loss[loss=0.1658, simple_loss=0.2557, pruned_loss=0.03792, over 2073689.84 frames. ], batch size: 30, lr: 3.44e-03, grad_scale: 8.0 2023-05-17 04:44:15,408 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=283939.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 04:44:31,094 INFO [finetune.py:992] (0/2) Epoch 16, batch 450, loss[loss=0.1652, simple_loss=0.2525, pruned_loss=0.03893, over 12022.00 frames. ], tot_loss[loss=0.1662, simple_loss=0.2559, pruned_loss=0.03827, over 2138822.02 frames. ], batch size: 40, lr: 3.44e-03, grad_scale: 8.0 2023-05-17 04:44:34,836 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=283966.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 04:44:50,463 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=283987.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 04:44:51,822 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.752e+02 2.511e+02 3.056e+02 3.598e+02 7.738e+02, threshold=6.112e+02, percent-clipped=2.0 2023-05-17 04:45:00,154 INFO [checkpoint.py:75] (0/2) Saving checkpoint to pruned_transducer_stateless7/exp_giga_finetune/checkpoint-184000.pt 2023-05-17 04:45:10,895 INFO [finetune.py:992] (0/2) Epoch 16, batch 500, loss[loss=0.1294, simple_loss=0.2164, pruned_loss=0.02115, over 12266.00 frames. ], tot_loss[loss=0.1656, simple_loss=0.2557, pruned_loss=0.03775, over 2192919.18 frames. ], batch size: 28, lr: 3.44e-03, grad_scale: 8.0 2023-05-17 04:45:16,670 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.9638, 5.7658, 5.3530, 5.2965, 5.8193, 5.1263, 5.2707, 5.3302], device='cuda:0'), covar=tensor([0.1664, 0.0956, 0.1176, 0.2176, 0.0962, 0.2321, 0.1976, 0.1237], device='cuda:0'), in_proj_covar=tensor([0.0360, 0.0499, 0.0404, 0.0453, 0.0463, 0.0435, 0.0397, 0.0390], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-17 04:45:22,499 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=284027.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 04:45:47,230 INFO [finetune.py:992] (0/2) Epoch 16, batch 550, loss[loss=0.1785, simple_loss=0.2707, pruned_loss=0.04318, over 12021.00 frames. ], tot_loss[loss=0.1645, simple_loss=0.2544, pruned_loss=0.03733, over 2245977.61 frames. ], batch size: 42, lr: 3.44e-03, grad_scale: 8.0 2023-05-17 04:45:52,514 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.8896, 4.4846, 4.1976, 4.5535, 3.2629, 4.0402, 2.6147, 4.1435], device='cuda:0'), covar=tensor([0.1485, 0.0584, 0.1062, 0.0708, 0.1329, 0.0706, 0.2094, 0.1389], device='cuda:0'), in_proj_covar=tensor([0.0234, 0.0272, 0.0304, 0.0361, 0.0247, 0.0248, 0.0267, 0.0379], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-17 04:46:02,354 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.2182, 5.1127, 4.9971, 5.0269, 4.8204, 5.1921, 5.1114, 5.4130], device='cuda:0'), covar=tensor([0.0238, 0.0154, 0.0183, 0.0360, 0.0710, 0.0360, 0.0175, 0.0189], device='cuda:0'), in_proj_covar=tensor([0.0192, 0.0191, 0.0188, 0.0241, 0.0234, 0.0216, 0.0174, 0.0229], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:0') 2023-05-17 04:46:02,395 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.7449, 3.1258, 3.3763, 4.5007, 2.5606, 4.5061, 4.6457, 4.7627], device='cuda:0'), covar=tensor([0.0093, 0.0974, 0.0440, 0.0140, 0.1259, 0.0234, 0.0121, 0.0081], device='cuda:0'), in_proj_covar=tensor([0.0119, 0.0201, 0.0180, 0.0119, 0.0188, 0.0175, 0.0171, 0.0122], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-17 04:46:04,580 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.4124, 4.3345, 4.1838, 4.4884, 3.0358, 4.1887, 2.8146, 4.1589], device='cuda:0'), covar=tensor([0.1625, 0.0614, 0.0974, 0.0711, 0.1305, 0.0586, 0.1901, 0.1207], device='cuda:0'), in_proj_covar=tensor([0.0234, 0.0272, 0.0304, 0.0361, 0.0247, 0.0248, 0.0267, 0.0379], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-17 04:46:07,262 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.018e+02 2.524e+02 2.956e+02 3.674e+02 6.026e+02, threshold=5.913e+02, percent-clipped=0.0 2023-05-17 04:46:22,577 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.7790, 2.8155, 4.7940, 4.8958, 2.7900, 2.6788, 2.9783, 2.2865], device='cuda:0'), covar=tensor([0.1763, 0.3347, 0.0451, 0.0413, 0.1477, 0.2569, 0.3109, 0.4495], device='cuda:0'), in_proj_covar=tensor([0.0301, 0.0389, 0.0277, 0.0301, 0.0275, 0.0315, 0.0394, 0.0380], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-17 04:46:23,026 INFO [finetune.py:992] (0/2) Epoch 16, batch 600, loss[loss=0.1384, simple_loss=0.2212, pruned_loss=0.02782, over 12357.00 frames. ], tot_loss[loss=0.1651, simple_loss=0.255, pruned_loss=0.03765, over 2267756.18 frames. ], batch size: 30, lr: 3.44e-03, grad_scale: 8.0 2023-05-17 04:46:46,263 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.54 vs. limit=2.0 2023-05-17 04:46:52,744 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.58 vs. limit=2.0 2023-05-17 04:46:59,389 INFO [finetune.py:992] (0/2) Epoch 16, batch 650, loss[loss=0.1898, simple_loss=0.2781, pruned_loss=0.05073, over 12096.00 frames. ], tot_loss[loss=0.1651, simple_loss=0.2551, pruned_loss=0.03757, over 2295100.26 frames. ], batch size: 40, lr: 3.43e-03, grad_scale: 8.0 2023-05-17 04:47:19,453 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.649e+02 2.652e+02 3.123e+02 3.791e+02 1.719e+03, threshold=6.246e+02, percent-clipped=4.0 2023-05-17 04:47:25,964 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=284198.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 04:47:27,847 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=284200.0, num_to_drop=1, layers_to_drop={0} 2023-05-17 04:47:33,752 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.5204, 2.9449, 3.6282, 4.5627, 3.9856, 4.5961, 3.7587, 3.2871], device='cuda:0'), covar=tensor([0.0038, 0.0336, 0.0155, 0.0035, 0.0117, 0.0073, 0.0157, 0.0330], device='cuda:0'), in_proj_covar=tensor([0.0088, 0.0122, 0.0103, 0.0077, 0.0102, 0.0115, 0.0099, 0.0137], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-17 04:47:35,676 INFO [finetune.py:992] (0/2) Epoch 16, batch 700, loss[loss=0.155, simple_loss=0.2453, pruned_loss=0.0323, over 12292.00 frames. ], tot_loss[loss=0.1663, simple_loss=0.2566, pruned_loss=0.03799, over 2315638.23 frames. ], batch size: 33, lr: 3.43e-03, grad_scale: 8.0 2023-05-17 04:47:55,635 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.1365, 4.9913, 4.8804, 4.9536, 4.6145, 5.1521, 5.0270, 5.3061], device='cuda:0'), covar=tensor([0.0196, 0.0158, 0.0192, 0.0366, 0.0777, 0.0293, 0.0162, 0.0201], device='cuda:0'), in_proj_covar=tensor([0.0192, 0.0191, 0.0188, 0.0242, 0.0234, 0.0217, 0.0174, 0.0229], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:0') 2023-05-17 04:48:02,000 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=284248.0, num_to_drop=1, layers_to_drop={0} 2023-05-17 04:48:02,896 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-05-17 04:48:11,147 INFO [finetune.py:992] (0/2) Epoch 16, batch 750, loss[loss=0.1626, simple_loss=0.2457, pruned_loss=0.03974, over 12007.00 frames. ], tot_loss[loss=0.1656, simple_loss=0.2557, pruned_loss=0.03782, over 2328531.33 frames. ], batch size: 28, lr: 3.43e-03, grad_scale: 8.0 2023-05-17 04:48:31,966 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.844e+02 2.420e+02 3.067e+02 4.052e+02 9.430e+02, threshold=6.135e+02, percent-clipped=7.0 2023-05-17 04:48:39,412 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.64 vs. limit=2.0 2023-05-17 04:48:47,811 INFO [finetune.py:992] (0/2) Epoch 16, batch 800, loss[loss=0.1768, simple_loss=0.2611, pruned_loss=0.04624, over 10529.00 frames. ], tot_loss[loss=0.1655, simple_loss=0.2556, pruned_loss=0.03769, over 2346934.01 frames. ], batch size: 68, lr: 3.43e-03, grad_scale: 8.0 2023-05-17 04:48:55,912 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=284322.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 04:49:17,252 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.3789, 5.1305, 5.2952, 5.3275, 4.9681, 5.0115, 4.7653, 5.2387], device='cuda:0'), covar=tensor([0.0597, 0.0636, 0.0826, 0.0584, 0.1970, 0.1319, 0.0567, 0.1079], device='cuda:0'), in_proj_covar=tensor([0.0538, 0.0696, 0.0610, 0.0629, 0.0838, 0.0740, 0.0567, 0.0480], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0005, 0.0004, 0.0003, 0.0003], device='cuda:0') 2023-05-17 04:49:24,320 INFO [finetune.py:992] (0/2) Epoch 16, batch 850, loss[loss=0.1368, simple_loss=0.2199, pruned_loss=0.02686, over 12180.00 frames. ], tot_loss[loss=0.1652, simple_loss=0.2553, pruned_loss=0.03758, over 2353766.86 frames. ], batch size: 31, lr: 3.43e-03, grad_scale: 8.0 2023-05-17 04:49:44,522 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.845e+02 2.671e+02 2.987e+02 3.565e+02 5.859e+02, threshold=5.974e+02, percent-clipped=0.0 2023-05-17 04:50:00,395 INFO [finetune.py:992] (0/2) Epoch 16, batch 900, loss[loss=0.1808, simple_loss=0.2713, pruned_loss=0.04519, over 12045.00 frames. ], tot_loss[loss=0.1648, simple_loss=0.2548, pruned_loss=0.03733, over 2365449.81 frames. ], batch size: 37, lr: 3.43e-03, grad_scale: 8.0 2023-05-17 04:50:06,707 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=4.72 vs. limit=5.0 2023-05-17 04:50:11,986 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.2769, 4.8389, 5.2609, 4.5828, 4.9214, 4.6615, 5.3137, 5.0588], device='cuda:0'), covar=tensor([0.0314, 0.0449, 0.0332, 0.0297, 0.0476, 0.0380, 0.0236, 0.0284], device='cuda:0'), in_proj_covar=tensor([0.0268, 0.0274, 0.0294, 0.0267, 0.0271, 0.0271, 0.0246, 0.0217], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-17 04:50:22,840 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.9612, 3.1902, 4.3944, 2.3730, 2.5390, 3.3630, 3.0077, 3.4036], device='cuda:0'), covar=tensor([0.0554, 0.1164, 0.0411, 0.1513, 0.2211, 0.1704, 0.1507, 0.1294], device='cuda:0'), in_proj_covar=tensor([0.0237, 0.0240, 0.0255, 0.0184, 0.0239, 0.0293, 0.0226, 0.0263], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-17 04:50:24,151 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=284443.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 04:50:36,944 INFO [finetune.py:992] (0/2) Epoch 16, batch 950, loss[loss=0.1568, simple_loss=0.251, pruned_loss=0.03133, over 12153.00 frames. ], tot_loss[loss=0.1645, simple_loss=0.2551, pruned_loss=0.03696, over 2364862.51 frames. ], batch size: 34, lr: 3.43e-03, grad_scale: 8.0 2023-05-17 04:50:40,575 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.3056, 4.9267, 5.3334, 4.6815, 4.9862, 4.7303, 5.3574, 5.0774], device='cuda:0'), covar=tensor([0.0310, 0.0425, 0.0317, 0.0282, 0.0464, 0.0370, 0.0257, 0.0280], device='cuda:0'), in_proj_covar=tensor([0.0269, 0.0274, 0.0294, 0.0267, 0.0272, 0.0271, 0.0247, 0.0218], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-17 04:50:57,332 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.745e+02 2.691e+02 3.092e+02 3.404e+02 8.207e+02, threshold=6.184e+02, percent-clipped=1.0 2023-05-17 04:51:03,877 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=284498.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 04:51:08,148 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=284504.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 04:51:12,822 INFO [finetune.py:992] (0/2) Epoch 16, batch 1000, loss[loss=0.1913, simple_loss=0.2694, pruned_loss=0.05661, over 8347.00 frames. ], tot_loss[loss=0.1651, simple_loss=0.2557, pruned_loss=0.03726, over 2367105.78 frames. ], batch size: 97, lr: 3.43e-03, grad_scale: 8.0 2023-05-17 04:51:37,809 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=284546.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 04:51:48,471 INFO [finetune.py:992] (0/2) Epoch 16, batch 1050, loss[loss=0.1705, simple_loss=0.265, pruned_loss=0.03793, over 12061.00 frames. ], tot_loss[loss=0.165, simple_loss=0.2553, pruned_loss=0.03735, over 2369501.89 frames. ], batch size: 40, lr: 3.43e-03, grad_scale: 8.0 2023-05-17 04:52:09,085 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.958e+02 2.650e+02 3.060e+02 3.658e+02 8.576e+02, threshold=6.119e+02, percent-clipped=2.0 2023-05-17 04:52:25,103 INFO [finetune.py:992] (0/2) Epoch 16, batch 1100, loss[loss=0.1481, simple_loss=0.2316, pruned_loss=0.03233, over 11803.00 frames. ], tot_loss[loss=0.165, simple_loss=0.2551, pruned_loss=0.03752, over 2371522.11 frames. ], batch size: 26, lr: 3.43e-03, grad_scale: 8.0 2023-05-17 04:52:28,924 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.8440, 3.3255, 5.1821, 2.8075, 3.0612, 3.9053, 3.3677, 3.9004], device='cuda:0'), covar=tensor([0.0471, 0.1253, 0.0330, 0.1274, 0.1941, 0.1527, 0.1369, 0.1274], device='cuda:0'), in_proj_covar=tensor([0.0236, 0.0240, 0.0255, 0.0184, 0.0239, 0.0292, 0.0226, 0.0263], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-17 04:52:33,779 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=284622.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 04:53:01,507 INFO [finetune.py:992] (0/2) Epoch 16, batch 1150, loss[loss=0.1605, simple_loss=0.2499, pruned_loss=0.03556, over 12045.00 frames. ], tot_loss[loss=0.1651, simple_loss=0.2552, pruned_loss=0.0375, over 2378301.41 frames. ], batch size: 40, lr: 3.43e-03, grad_scale: 8.0 2023-05-17 04:53:08,041 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=284670.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 04:53:21,777 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.856e+02 2.636e+02 3.102e+02 3.453e+02 5.251e+02, threshold=6.203e+02, percent-clipped=0.0 2023-05-17 04:53:37,878 INFO [finetune.py:992] (0/2) Epoch 16, batch 1200, loss[loss=0.1661, simple_loss=0.2587, pruned_loss=0.0367, over 12032.00 frames. ], tot_loss[loss=0.1639, simple_loss=0.2537, pruned_loss=0.03708, over 2385843.93 frames. ], batch size: 40, lr: 3.43e-03, grad_scale: 8.0 2023-05-17 04:54:14,005 INFO [finetune.py:992] (0/2) Epoch 16, batch 1250, loss[loss=0.1595, simple_loss=0.2388, pruned_loss=0.04011, over 12254.00 frames. ], tot_loss[loss=0.1641, simple_loss=0.2537, pruned_loss=0.03729, over 2385655.41 frames. ], batch size: 28, lr: 3.43e-03, grad_scale: 8.0 2023-05-17 04:54:34,828 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.966e+02 2.693e+02 3.233e+02 3.746e+02 7.896e+02, threshold=6.466e+02, percent-clipped=1.0 2023-05-17 04:54:42,100 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=284799.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 04:54:48,569 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.37 vs. limit=2.0 2023-05-17 04:54:50,941 INFO [finetune.py:992] (0/2) Epoch 16, batch 1300, loss[loss=0.1582, simple_loss=0.2494, pruned_loss=0.0335, over 12185.00 frames. ], tot_loss[loss=0.1648, simple_loss=0.2545, pruned_loss=0.03761, over 2381799.91 frames. ], batch size: 31, lr: 3.43e-03, grad_scale: 8.0 2023-05-17 04:54:54,732 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=284816.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 04:55:26,495 INFO [finetune.py:992] (0/2) Epoch 16, batch 1350, loss[loss=0.1847, simple_loss=0.2811, pruned_loss=0.04417, over 12157.00 frames. ], tot_loss[loss=0.1655, simple_loss=0.2554, pruned_loss=0.03775, over 2391544.08 frames. ], batch size: 34, lr: 3.43e-03, grad_scale: 8.0 2023-05-17 04:55:30,330 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.3544, 2.2023, 3.0085, 4.1293, 2.1852, 4.2810, 4.3379, 4.4595], device='cuda:0'), covar=tensor([0.0137, 0.1484, 0.0583, 0.0178, 0.1469, 0.0213, 0.0169, 0.0099], device='cuda:0'), in_proj_covar=tensor([0.0122, 0.0206, 0.0184, 0.0123, 0.0192, 0.0180, 0.0176, 0.0125], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-17 04:55:38,561 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=284877.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 04:55:43,429 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.0520, 5.7636, 5.5167, 5.2339, 5.8888, 5.2099, 5.2514, 5.2999], device='cuda:0'), covar=tensor([0.1538, 0.1078, 0.1131, 0.2082, 0.1072, 0.2143, 0.2349, 0.1344], device='cuda:0'), in_proj_covar=tensor([0.0367, 0.0508, 0.0410, 0.0458, 0.0473, 0.0441, 0.0404, 0.0397], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-17 04:55:46,809 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.028e+02 2.652e+02 3.047e+02 3.454e+02 8.344e+02, threshold=6.094e+02, percent-clipped=2.0 2023-05-17 04:55:58,163 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.94 vs. limit=2.0 2023-05-17 04:56:02,640 INFO [finetune.py:992] (0/2) Epoch 16, batch 1400, loss[loss=0.1647, simple_loss=0.2465, pruned_loss=0.04144, over 12112.00 frames. ], tot_loss[loss=0.1648, simple_loss=0.2551, pruned_loss=0.03728, over 2394077.23 frames. ], batch size: 30, lr: 3.43e-03, grad_scale: 8.0 2023-05-17 04:56:17,272 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.92 vs. limit=2.0 2023-05-17 04:56:23,492 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=284939.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 04:56:39,286 INFO [finetune.py:992] (0/2) Epoch 16, batch 1450, loss[loss=0.1695, simple_loss=0.2609, pruned_loss=0.03903, over 12050.00 frames. ], tot_loss[loss=0.1647, simple_loss=0.2552, pruned_loss=0.03713, over 2390174.92 frames. ], batch size: 37, lr: 3.43e-03, grad_scale: 8.0 2023-05-17 04:56:59,195 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.046e+02 2.503e+02 2.956e+02 3.648e+02 7.736e+02, threshold=5.913e+02, percent-clipped=3.0 2023-05-17 04:57:04,424 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=284996.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 04:57:07,774 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=285000.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 04:57:15,268 INFO [finetune.py:992] (0/2) Epoch 16, batch 1500, loss[loss=0.1629, simple_loss=0.2468, pruned_loss=0.03945, over 12022.00 frames. ], tot_loss[loss=0.1649, simple_loss=0.2553, pruned_loss=0.03725, over 2386657.48 frames. ], batch size: 31, lr: 3.43e-03, grad_scale: 8.0 2023-05-17 04:57:20,504 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.3201, 4.3989, 2.7507, 2.1466, 3.8839, 2.1862, 3.8779, 2.9445], device='cuda:0'), covar=tensor([0.0666, 0.0843, 0.1272, 0.2285, 0.0450, 0.1995, 0.0502, 0.1135], device='cuda:0'), in_proj_covar=tensor([0.0189, 0.0256, 0.0176, 0.0203, 0.0143, 0.0184, 0.0197, 0.0174], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-17 04:57:26,172 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.1811, 4.8037, 4.2101, 4.9108, 4.6845, 2.8864, 4.2206, 2.9441], device='cuda:0'), covar=tensor([0.0876, 0.0622, 0.1359, 0.0568, 0.1116, 0.1669, 0.1067, 0.3290], device='cuda:0'), in_proj_covar=tensor([0.0308, 0.0383, 0.0363, 0.0326, 0.0373, 0.0277, 0.0350, 0.0369], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-17 04:57:36,224 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.3015, 4.6443, 2.8735, 2.6646, 3.8915, 2.6246, 3.9375, 3.1948], device='cuda:0'), covar=tensor([0.0757, 0.0644, 0.1225, 0.1610, 0.0398, 0.1388, 0.0517, 0.0895], device='cuda:0'), in_proj_covar=tensor([0.0189, 0.0257, 0.0177, 0.0203, 0.0144, 0.0184, 0.0197, 0.0174], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-17 04:57:38,268 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=285042.0, num_to_drop=1, layers_to_drop={1} 2023-05-17 04:57:48,957 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=285057.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 04:57:51,531 INFO [finetune.py:992] (0/2) Epoch 16, batch 1550, loss[loss=0.1756, simple_loss=0.2668, pruned_loss=0.04217, over 12151.00 frames. ], tot_loss[loss=0.165, simple_loss=0.2552, pruned_loss=0.03742, over 2393121.44 frames. ], batch size: 36, lr: 3.43e-03, grad_scale: 8.0 2023-05-17 04:58:11,929 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.936e+02 2.594e+02 3.142e+02 3.617e+02 6.812e+02, threshold=6.284e+02, percent-clipped=2.0 2023-05-17 04:58:19,266 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=285099.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 04:58:22,222 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=285103.0, num_to_drop=1, layers_to_drop={2} 2023-05-17 04:58:24,674 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=4.82 vs. limit=5.0 2023-05-17 04:58:27,694 INFO [finetune.py:992] (0/2) Epoch 16, batch 1600, loss[loss=0.1519, simple_loss=0.2368, pruned_loss=0.03356, over 12029.00 frames. ], tot_loss[loss=0.1643, simple_loss=0.2545, pruned_loss=0.03703, over 2397805.50 frames. ], batch size: 31, lr: 3.43e-03, grad_scale: 8.0 2023-05-17 04:58:29,900 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.1603, 5.9687, 5.6740, 5.5671, 6.0409, 5.3501, 5.3885, 5.5254], device='cuda:0'), covar=tensor([0.1512, 0.0986, 0.0907, 0.1841, 0.1002, 0.1999, 0.2266, 0.1314], device='cuda:0'), in_proj_covar=tensor([0.0366, 0.0508, 0.0409, 0.0458, 0.0474, 0.0441, 0.0403, 0.0397], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-17 04:58:53,374 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=285147.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 04:59:02,899 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.5680, 2.5053, 3.2897, 4.3989, 2.5641, 4.4798, 4.5821, 4.6872], device='cuda:0'), covar=tensor([0.0147, 0.1345, 0.0498, 0.0171, 0.1310, 0.0249, 0.0153, 0.0108], device='cuda:0'), in_proj_covar=tensor([0.0123, 0.0206, 0.0184, 0.0123, 0.0192, 0.0181, 0.0177, 0.0126], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-17 04:59:03,415 INFO [finetune.py:992] (0/2) Epoch 16, batch 1650, loss[loss=0.1534, simple_loss=0.2379, pruned_loss=0.03448, over 12184.00 frames. ], tot_loss[loss=0.1641, simple_loss=0.2544, pruned_loss=0.03684, over 2397231.99 frames. ], batch size: 31, lr: 3.43e-03, grad_scale: 8.0 2023-05-17 04:59:11,343 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=285172.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 04:59:23,743 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.874e+02 2.631e+02 3.056e+02 3.621e+02 1.098e+03, threshold=6.112e+02, percent-clipped=1.0 2023-05-17 04:59:40,071 INFO [finetune.py:992] (0/2) Epoch 16, batch 1700, loss[loss=0.135, simple_loss=0.2207, pruned_loss=0.02465, over 12295.00 frames. ], tot_loss[loss=0.1635, simple_loss=0.2535, pruned_loss=0.03681, over 2387874.21 frames. ], batch size: 28, lr: 3.43e-03, grad_scale: 8.0 2023-05-17 04:59:52,892 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.35 vs. limit=2.0 2023-05-17 05:00:02,670 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.4254, 2.3335, 3.1293, 4.2709, 2.4855, 4.3405, 4.4038, 4.5615], device='cuda:0'), covar=tensor([0.0170, 0.1487, 0.0561, 0.0174, 0.1322, 0.0267, 0.0202, 0.0110], device='cuda:0'), in_proj_covar=tensor([0.0124, 0.0207, 0.0186, 0.0124, 0.0194, 0.0182, 0.0178, 0.0127], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-17 05:00:05,544 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.0285, 4.6625, 4.8582, 4.8915, 4.5886, 4.8564, 4.7842, 2.9413], device='cuda:0'), covar=tensor([0.0092, 0.0067, 0.0082, 0.0063, 0.0062, 0.0100, 0.0080, 0.0706], device='cuda:0'), in_proj_covar=tensor([0.0070, 0.0080, 0.0083, 0.0074, 0.0062, 0.0094, 0.0083, 0.0100], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-17 05:00:16,845 INFO [finetune.py:992] (0/2) Epoch 16, batch 1750, loss[loss=0.1681, simple_loss=0.2477, pruned_loss=0.04422, over 12282.00 frames. ], tot_loss[loss=0.1639, simple_loss=0.2538, pruned_loss=0.03698, over 2388931.30 frames. ], batch size: 33, lr: 3.43e-03, grad_scale: 8.0 2023-05-17 05:00:17,119 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.7844, 2.6100, 3.5307, 3.6320, 2.8495, 2.5887, 2.6292, 2.3593], device='cuda:0'), covar=tensor([0.1431, 0.2577, 0.0702, 0.0548, 0.1111, 0.2332, 0.2696, 0.3870], device='cuda:0'), in_proj_covar=tensor([0.0306, 0.0397, 0.0284, 0.0305, 0.0280, 0.0322, 0.0401, 0.0385], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-17 05:00:36,545 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.022e+02 2.639e+02 3.105e+02 3.707e+02 1.479e+03, threshold=6.210e+02, percent-clipped=4.0 2023-05-17 05:00:40,973 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=285295.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 05:00:52,193 INFO [finetune.py:992] (0/2) Epoch 16, batch 1800, loss[loss=0.1737, simple_loss=0.266, pruned_loss=0.04071, over 12151.00 frames. ], tot_loss[loss=0.1639, simple_loss=0.2536, pruned_loss=0.03711, over 2381182.80 frames. ], batch size: 34, lr: 3.43e-03, grad_scale: 8.0 2023-05-17 05:01:22,441 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=285352.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 05:01:29,101 INFO [finetune.py:992] (0/2) Epoch 16, batch 1850, loss[loss=0.2081, simple_loss=0.3035, pruned_loss=0.05635, over 12265.00 frames. ], tot_loss[loss=0.1634, simple_loss=0.2531, pruned_loss=0.03687, over 2388490.71 frames. ], batch size: 37, lr: 3.43e-03, grad_scale: 8.0 2023-05-17 05:01:49,236 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.881e+02 2.598e+02 2.984e+02 3.373e+02 7.020e+02, threshold=5.967e+02, percent-clipped=3.0 2023-05-17 05:01:55,498 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=285398.0, num_to_drop=1, layers_to_drop={0} 2023-05-17 05:02:04,951 INFO [finetune.py:992] (0/2) Epoch 16, batch 1900, loss[loss=0.1517, simple_loss=0.2342, pruned_loss=0.03458, over 12177.00 frames. ], tot_loss[loss=0.1635, simple_loss=0.2532, pruned_loss=0.03688, over 2384597.94 frames. ], batch size: 29, lr: 3.43e-03, grad_scale: 8.0 2023-05-17 05:02:22,207 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.3533, 3.5190, 3.2147, 3.1851, 2.7653, 2.6423, 3.5334, 2.1635], device='cuda:0'), covar=tensor([0.0439, 0.0149, 0.0199, 0.0194, 0.0435, 0.0434, 0.0156, 0.0592], device='cuda:0'), in_proj_covar=tensor([0.0196, 0.0165, 0.0170, 0.0191, 0.0205, 0.0202, 0.0177, 0.0206], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-17 05:02:40,371 INFO [finetune.py:992] (0/2) Epoch 16, batch 1950, loss[loss=0.1397, simple_loss=0.2293, pruned_loss=0.02509, over 12287.00 frames. ], tot_loss[loss=0.1634, simple_loss=0.2531, pruned_loss=0.03681, over 2390775.11 frames. ], batch size: 33, lr: 3.43e-03, grad_scale: 8.0 2023-05-17 05:02:49,036 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=285472.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 05:02:55,022 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 2023-05-17 05:03:00,831 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.926e+02 2.602e+02 3.218e+02 3.856e+02 6.883e+02, threshold=6.436e+02, percent-clipped=4.0 2023-05-17 05:03:17,180 INFO [finetune.py:992] (0/2) Epoch 16, batch 2000, loss[loss=0.1675, simple_loss=0.2653, pruned_loss=0.0348, over 12138.00 frames. ], tot_loss[loss=0.1638, simple_loss=0.2542, pruned_loss=0.0367, over 2384928.58 frames. ], batch size: 34, lr: 3.43e-03, grad_scale: 8.0 2023-05-17 05:03:23,759 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=285520.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 05:03:53,026 INFO [finetune.py:992] (0/2) Epoch 16, batch 2050, loss[loss=0.1682, simple_loss=0.2676, pruned_loss=0.03436, over 11625.00 frames. ], tot_loss[loss=0.1629, simple_loss=0.2532, pruned_loss=0.03634, over 2386426.38 frames. ], batch size: 48, lr: 3.42e-03, grad_scale: 8.0 2023-05-17 05:04:13,121 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.783e+02 2.449e+02 2.928e+02 3.605e+02 5.021e+02, threshold=5.856e+02, percent-clipped=0.0 2023-05-17 05:04:17,458 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=285595.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 05:04:21,596 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.5260, 4.0317, 4.0468, 4.5227, 3.0400, 3.7680, 2.4549, 3.9389], device='cuda:0'), covar=tensor([0.1727, 0.0886, 0.1099, 0.0591, 0.1393, 0.0799, 0.2340, 0.1227], device='cuda:0'), in_proj_covar=tensor([0.0231, 0.0270, 0.0300, 0.0359, 0.0243, 0.0245, 0.0264, 0.0371], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-17 05:04:29,153 INFO [finetune.py:992] (0/2) Epoch 16, batch 2100, loss[loss=0.1756, simple_loss=0.2688, pruned_loss=0.04119, over 12113.00 frames. ], tot_loss[loss=0.1633, simple_loss=0.2534, pruned_loss=0.03665, over 2382192.86 frames. ], batch size: 45, lr: 3.42e-03, grad_scale: 8.0 2023-05-17 05:04:36,238 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.9013, 5.8910, 5.5963, 5.1351, 5.1145, 5.7463, 5.3317, 5.2013], device='cuda:0'), covar=tensor([0.0929, 0.1044, 0.0799, 0.1737, 0.0821, 0.0838, 0.1871, 0.1183], device='cuda:0'), in_proj_covar=tensor([0.0641, 0.0578, 0.0528, 0.0653, 0.0423, 0.0739, 0.0796, 0.0583], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-05-17 05:04:37,689 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=285622.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 05:04:42,785 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.7945, 3.4320, 5.2069, 2.7895, 2.9651, 3.8685, 3.3620, 3.8437], device='cuda:0'), covar=tensor([0.0418, 0.1171, 0.0302, 0.1194, 0.1884, 0.1488, 0.1299, 0.1168], device='cuda:0'), in_proj_covar=tensor([0.0238, 0.0242, 0.0259, 0.0186, 0.0242, 0.0296, 0.0228, 0.0266], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-17 05:04:49,768 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.4151, 3.7447, 3.3391, 3.3141, 3.1395, 2.8600, 3.7521, 2.3710], device='cuda:0'), covar=tensor([0.0468, 0.0152, 0.0192, 0.0201, 0.0350, 0.0372, 0.0123, 0.0563], device='cuda:0'), in_proj_covar=tensor([0.0195, 0.0164, 0.0169, 0.0189, 0.0203, 0.0199, 0.0175, 0.0204], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-17 05:04:52,437 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=285643.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 05:04:58,825 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=285652.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 05:05:05,866 INFO [finetune.py:992] (0/2) Epoch 16, batch 2150, loss[loss=0.1537, simple_loss=0.2392, pruned_loss=0.03409, over 12249.00 frames. ], tot_loss[loss=0.1629, simple_loss=0.2527, pruned_loss=0.03653, over 2380504.40 frames. ], batch size: 32, lr: 3.42e-03, grad_scale: 8.0 2023-05-17 05:05:21,776 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=285683.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 05:05:25,729 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.986e+02 2.738e+02 3.209e+02 3.836e+02 6.080e+02, threshold=6.417e+02, percent-clipped=1.0 2023-05-17 05:05:32,427 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=285698.0, num_to_drop=1, layers_to_drop={0} 2023-05-17 05:05:33,807 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=285700.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 05:05:41,647 INFO [finetune.py:992] (0/2) Epoch 16, batch 2200, loss[loss=0.1709, simple_loss=0.2584, pruned_loss=0.04174, over 12158.00 frames. ], tot_loss[loss=0.163, simple_loss=0.2526, pruned_loss=0.0367, over 2375696.01 frames. ], batch size: 34, lr: 3.42e-03, grad_scale: 8.0 2023-05-17 05:05:50,684 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.39 vs. limit=2.0 2023-05-17 05:05:56,177 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.86 vs. limit=2.0 2023-05-17 05:06:06,589 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=285746.0, num_to_drop=1, layers_to_drop={0} 2023-05-17 05:06:17,891 INFO [finetune.py:992] (0/2) Epoch 16, batch 2250, loss[loss=0.1603, simple_loss=0.2524, pruned_loss=0.03411, over 12013.00 frames. ], tot_loss[loss=0.1628, simple_loss=0.2524, pruned_loss=0.03659, over 2371778.01 frames. ], batch size: 40, lr: 3.42e-03, grad_scale: 8.0 2023-05-17 05:06:35,941 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.0133, 4.5611, 4.7841, 4.8495, 4.5834, 4.8167, 4.7130, 2.7893], device='cuda:0'), covar=tensor([0.0106, 0.0075, 0.0097, 0.0064, 0.0064, 0.0113, 0.0098, 0.0728], device='cuda:0'), in_proj_covar=tensor([0.0070, 0.0080, 0.0083, 0.0074, 0.0061, 0.0093, 0.0083, 0.0099], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-17 05:06:37,892 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.593e+02 2.712e+02 3.166e+02 3.755e+02 7.301e+02, threshold=6.332e+02, percent-clipped=2.0 2023-05-17 05:06:54,481 INFO [finetune.py:992] (0/2) Epoch 16, batch 2300, loss[loss=0.1682, simple_loss=0.2677, pruned_loss=0.03434, over 11784.00 frames. ], tot_loss[loss=0.1625, simple_loss=0.2518, pruned_loss=0.03662, over 2376396.77 frames. ], batch size: 44, lr: 3.42e-03, grad_scale: 8.0 2023-05-17 05:07:04,206 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.3567, 4.0667, 4.2246, 4.2545, 4.1153, 4.3028, 4.2145, 2.6547], device='cuda:0'), covar=tensor([0.0108, 0.0102, 0.0119, 0.0074, 0.0074, 0.0114, 0.0136, 0.0796], device='cuda:0'), in_proj_covar=tensor([0.0070, 0.0080, 0.0083, 0.0074, 0.0061, 0.0093, 0.0083, 0.0099], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-17 05:07:30,540 INFO [finetune.py:992] (0/2) Epoch 16, batch 2350, loss[loss=0.1553, simple_loss=0.2559, pruned_loss=0.02731, over 12355.00 frames. ], tot_loss[loss=0.1619, simple_loss=0.2512, pruned_loss=0.03625, over 2381691.24 frames. ], batch size: 35, lr: 3.42e-03, grad_scale: 8.0 2023-05-17 05:07:42,832 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.8457, 3.2569, 2.4269, 2.2231, 3.0294, 2.3300, 3.1027, 2.6435], device='cuda:0'), covar=tensor([0.0667, 0.0742, 0.1090, 0.1429, 0.0319, 0.1159, 0.0552, 0.0819], device='cuda:0'), in_proj_covar=tensor([0.0192, 0.0262, 0.0178, 0.0205, 0.0146, 0.0186, 0.0200, 0.0176], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-17 05:07:50,191 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.2538, 2.5954, 3.8363, 3.2168, 3.5870, 3.3264, 2.7116, 3.6486], device='cuda:0'), covar=tensor([0.0158, 0.0425, 0.0154, 0.0275, 0.0178, 0.0206, 0.0374, 0.0155], device='cuda:0'), in_proj_covar=tensor([0.0187, 0.0212, 0.0200, 0.0194, 0.0225, 0.0173, 0.0204, 0.0195], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-17 05:07:50,641 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.717e+02 2.624e+02 3.039e+02 3.508e+02 6.519e+02, threshold=6.078e+02, percent-clipped=2.0 2023-05-17 05:08:07,064 INFO [finetune.py:992] (0/2) Epoch 16, batch 2400, loss[loss=0.174, simple_loss=0.2633, pruned_loss=0.04235, over 12394.00 frames. ], tot_loss[loss=0.1622, simple_loss=0.2517, pruned_loss=0.03631, over 2383990.12 frames. ], batch size: 38, lr: 3.42e-03, grad_scale: 16.0 2023-05-17 05:08:43,443 INFO [finetune.py:992] (0/2) Epoch 16, batch 2450, loss[loss=0.1427, simple_loss=0.2267, pruned_loss=0.02929, over 12315.00 frames. ], tot_loss[loss=0.1637, simple_loss=0.2531, pruned_loss=0.03708, over 2376750.59 frames. ], batch size: 30, lr: 3.42e-03, grad_scale: 16.0 2023-05-17 05:08:55,797 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=285978.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 05:08:56,069 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.47 vs. limit=2.0 2023-05-17 05:09:03,567 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.876e+02 2.739e+02 3.151e+02 3.666e+02 5.852e+02, threshold=6.301e+02, percent-clipped=0.0 2023-05-17 05:09:11,790 INFO [checkpoint.py:75] (0/2) Saving checkpoint to pruned_transducer_stateless7/exp_giga_finetune/checkpoint-186000.pt 2023-05-17 05:09:22,625 INFO [finetune.py:992] (0/2) Epoch 16, batch 2500, loss[loss=0.1732, simple_loss=0.2682, pruned_loss=0.03914, over 12353.00 frames. ], tot_loss[loss=0.1635, simple_loss=0.2531, pruned_loss=0.03694, over 2376783.64 frames. ], batch size: 35, lr: 3.42e-03, grad_scale: 16.0 2023-05-17 05:09:29,273 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.4535, 4.7839, 3.0511, 2.8574, 4.1052, 2.9066, 4.0021, 3.3573], device='cuda:0'), covar=tensor([0.0741, 0.0584, 0.1069, 0.1457, 0.0316, 0.1201, 0.0482, 0.0799], device='cuda:0'), in_proj_covar=tensor([0.0190, 0.0258, 0.0176, 0.0203, 0.0144, 0.0184, 0.0197, 0.0174], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-17 05:09:37,122 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.2363, 4.8882, 5.0212, 5.0966, 4.9257, 5.0970, 4.9603, 2.8319], device='cuda:0'), covar=tensor([0.0076, 0.0066, 0.0087, 0.0059, 0.0054, 0.0101, 0.0073, 0.0768], device='cuda:0'), in_proj_covar=tensor([0.0071, 0.0081, 0.0084, 0.0075, 0.0062, 0.0094, 0.0083, 0.0099], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-17 05:09:42,067 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.5678, 2.7649, 3.6870, 4.5806, 4.1350, 4.6224, 3.8824, 3.2619], device='cuda:0'), covar=tensor([0.0042, 0.0398, 0.0159, 0.0049, 0.0099, 0.0083, 0.0135, 0.0381], device='cuda:0'), in_proj_covar=tensor([0.0088, 0.0122, 0.0104, 0.0079, 0.0104, 0.0116, 0.0099, 0.0138], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-17 05:09:44,842 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.3036, 4.8848, 5.3097, 4.6751, 4.8955, 4.7616, 5.2960, 4.9961], device='cuda:0'), covar=tensor([0.0249, 0.0440, 0.0256, 0.0275, 0.0474, 0.0302, 0.0238, 0.0263], device='cuda:0'), in_proj_covar=tensor([0.0272, 0.0280, 0.0297, 0.0272, 0.0275, 0.0274, 0.0249, 0.0222], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-17 05:09:58,908 INFO [finetune.py:992] (0/2) Epoch 16, batch 2550, loss[loss=0.1665, simple_loss=0.265, pruned_loss=0.03402, over 12089.00 frames. ], tot_loss[loss=0.1639, simple_loss=0.2534, pruned_loss=0.03723, over 2376456.40 frames. ], batch size: 32, lr: 3.42e-03, grad_scale: 16.0 2023-05-17 05:10:19,542 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.966e+02 2.621e+02 2.966e+02 3.355e+02 6.018e+02, threshold=5.932e+02, percent-clipped=0.0 2023-05-17 05:10:26,258 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.9553, 2.3037, 3.4986, 2.8661, 3.2939, 3.0839, 2.4197, 3.4478], device='cuda:0'), covar=tensor([0.0150, 0.0446, 0.0186, 0.0337, 0.0180, 0.0206, 0.0414, 0.0163], device='cuda:0'), in_proj_covar=tensor([0.0186, 0.0211, 0.0198, 0.0193, 0.0224, 0.0172, 0.0202, 0.0193], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-17 05:10:35,296 INFO [finetune.py:992] (0/2) Epoch 16, batch 2600, loss[loss=0.1317, simple_loss=0.2107, pruned_loss=0.02637, over 12335.00 frames. ], tot_loss[loss=0.1642, simple_loss=0.2537, pruned_loss=0.0374, over 2378821.54 frames. ], batch size: 30, lr: 3.42e-03, grad_scale: 16.0 2023-05-17 05:10:44,029 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=286123.0, num_to_drop=1, layers_to_drop={0} 2023-05-17 05:10:48,050 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=286129.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 05:11:10,669 INFO [finetune.py:992] (0/2) Epoch 16, batch 2650, loss[loss=0.1592, simple_loss=0.2554, pruned_loss=0.03149, over 12106.00 frames. ], tot_loss[loss=0.1633, simple_loss=0.2524, pruned_loss=0.03705, over 2384453.48 frames. ], batch size: 33, lr: 3.42e-03, grad_scale: 16.0 2023-05-17 05:11:27,769 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=286184.0, num_to_drop=1, layers_to_drop={0} 2023-05-17 05:11:31,135 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.779e+02 2.530e+02 2.932e+02 3.499e+02 1.091e+03, threshold=5.863e+02, percent-clipped=1.0 2023-05-17 05:11:32,137 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=286190.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 05:11:47,726 INFO [finetune.py:992] (0/2) Epoch 16, batch 2700, loss[loss=0.1545, simple_loss=0.2363, pruned_loss=0.0364, over 12123.00 frames. ], tot_loss[loss=0.1631, simple_loss=0.2523, pruned_loss=0.03695, over 2384588.00 frames. ], batch size: 30, lr: 3.42e-03, grad_scale: 16.0 2023-05-17 05:12:04,978 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.2747, 5.0937, 4.9733, 5.0975, 4.7220, 5.1713, 5.1564, 5.4314], device='cuda:0'), covar=tensor([0.0293, 0.0166, 0.0213, 0.0370, 0.0809, 0.0313, 0.0183, 0.0182], device='cuda:0'), in_proj_covar=tensor([0.0203, 0.0202, 0.0197, 0.0255, 0.0247, 0.0229, 0.0183, 0.0240], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0005, 0.0004, 0.0004, 0.0003, 0.0005], device='cuda:0') 2023-05-17 05:12:24,243 INFO [finetune.py:992] (0/2) Epoch 16, batch 2750, loss[loss=0.2104, simple_loss=0.2938, pruned_loss=0.06351, over 8065.00 frames. ], tot_loss[loss=0.1628, simple_loss=0.2523, pruned_loss=0.0366, over 2379133.01 frames. ], batch size: 98, lr: 3.42e-03, grad_scale: 16.0 2023-05-17 05:12:25,907 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.3553, 4.9394, 5.3717, 4.6733, 4.9865, 4.8007, 5.3643, 4.9824], device='cuda:0'), covar=tensor([0.0272, 0.0431, 0.0284, 0.0286, 0.0452, 0.0293, 0.0242, 0.0317], device='cuda:0'), in_proj_covar=tensor([0.0275, 0.0282, 0.0299, 0.0273, 0.0275, 0.0275, 0.0249, 0.0223], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-17 05:12:36,682 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=286278.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 05:12:41,919 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=4.79 vs. limit=5.0 2023-05-17 05:12:44,316 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.624e+02 2.654e+02 3.141e+02 3.640e+02 5.193e+02, threshold=6.282e+02, percent-clipped=0.0 2023-05-17 05:12:48,933 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=286295.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 05:13:00,145 INFO [finetune.py:992] (0/2) Epoch 16, batch 2800, loss[loss=0.1436, simple_loss=0.2275, pruned_loss=0.02986, over 12359.00 frames. ], tot_loss[loss=0.1621, simple_loss=0.2516, pruned_loss=0.03625, over 2383970.72 frames. ], batch size: 30, lr: 3.42e-03, grad_scale: 16.0 2023-05-17 05:13:05,358 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.3589, 4.7819, 4.1853, 5.0038, 4.6111, 2.8999, 4.3506, 3.2419], device='cuda:0'), covar=tensor([0.0800, 0.0711, 0.1356, 0.0514, 0.1057, 0.1819, 0.1002, 0.3135], device='cuda:0'), in_proj_covar=tensor([0.0309, 0.0383, 0.0362, 0.0329, 0.0374, 0.0277, 0.0349, 0.0368], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-17 05:13:10,768 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=286326.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 05:13:22,937 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.1832, 6.0300, 5.6873, 5.5968, 6.1086, 5.3616, 5.5515, 5.5479], device='cuda:0'), covar=tensor([0.1743, 0.0964, 0.0931, 0.2017, 0.1033, 0.2412, 0.1906, 0.1340], device='cuda:0'), in_proj_covar=tensor([0.0364, 0.0506, 0.0406, 0.0452, 0.0473, 0.0437, 0.0400, 0.0392], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-17 05:13:24,452 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=286344.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 05:13:31,849 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.2805, 4.3906, 4.1893, 4.5599, 3.2495, 3.9953, 2.9071, 4.2839], device='cuda:0'), covar=tensor([0.1544, 0.0577, 0.0900, 0.0597, 0.1112, 0.0592, 0.1644, 0.1032], device='cuda:0'), in_proj_covar=tensor([0.0230, 0.0267, 0.0297, 0.0359, 0.0242, 0.0243, 0.0263, 0.0369], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-17 05:13:33,161 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=286356.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 05:13:36,298 INFO [finetune.py:992] (0/2) Epoch 16, batch 2850, loss[loss=0.1653, simple_loss=0.2532, pruned_loss=0.03866, over 12099.00 frames. ], tot_loss[loss=0.1621, simple_loss=0.2517, pruned_loss=0.03626, over 2384351.28 frames. ], batch size: 32, lr: 3.42e-03, grad_scale: 16.0 2023-05-17 05:13:52,077 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.2578, 5.2095, 4.9496, 4.5992, 4.6345, 5.1465, 4.8220, 4.6008], device='cuda:0'), covar=tensor([0.0768, 0.0926, 0.0783, 0.1793, 0.1565, 0.0869, 0.1629, 0.1217], device='cuda:0'), in_proj_covar=tensor([0.0635, 0.0576, 0.0529, 0.0653, 0.0426, 0.0741, 0.0792, 0.0581], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-05-17 05:13:52,935 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.8090, 3.1697, 2.4000, 2.2417, 2.9083, 2.3909, 3.0025, 2.6269], device='cuda:0'), covar=tensor([0.0583, 0.0526, 0.0903, 0.1194, 0.0275, 0.1015, 0.0494, 0.0663], device='cuda:0'), in_proj_covar=tensor([0.0189, 0.0258, 0.0176, 0.0203, 0.0143, 0.0183, 0.0198, 0.0173], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-17 05:13:56,933 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.885e+02 2.626e+02 3.103e+02 3.601e+02 5.239e+02, threshold=6.207e+02, percent-clipped=0.0 2023-05-17 05:14:08,992 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.88 vs. limit=2.0 2023-05-17 05:14:09,251 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=286405.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 05:14:13,107 INFO [finetune.py:992] (0/2) Epoch 16, batch 2900, loss[loss=0.1772, simple_loss=0.266, pruned_loss=0.04415, over 8029.00 frames. ], tot_loss[loss=0.1618, simple_loss=0.2514, pruned_loss=0.03616, over 2380498.63 frames. ], batch size: 97, lr: 3.42e-03, grad_scale: 16.0 2023-05-17 05:14:31,903 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.93 vs. limit=2.0 2023-05-17 05:14:48,801 INFO [finetune.py:992] (0/2) Epoch 16, batch 2950, loss[loss=0.1779, simple_loss=0.2619, pruned_loss=0.04692, over 12182.00 frames. ], tot_loss[loss=0.163, simple_loss=0.2522, pruned_loss=0.03688, over 2371816.04 frames. ], batch size: 35, lr: 3.42e-03, grad_scale: 16.0 2023-05-17 05:15:01,566 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=286479.0, num_to_drop=1, layers_to_drop={0} 2023-05-17 05:15:06,578 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=286485.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 05:15:09,210 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.775e+02 2.711e+02 3.100e+02 3.818e+02 6.057e+02, threshold=6.200e+02, percent-clipped=0.0 2023-05-17 05:15:24,606 INFO [finetune.py:992] (0/2) Epoch 16, batch 3000, loss[loss=0.1596, simple_loss=0.2446, pruned_loss=0.03726, over 12293.00 frames. ], tot_loss[loss=0.1627, simple_loss=0.2522, pruned_loss=0.03663, over 2372285.03 frames. ], batch size: 33, lr: 3.42e-03, grad_scale: 16.0 2023-05-17 05:15:24,607 INFO [finetune.py:1017] (0/2) Computing validation loss 2023-05-17 05:15:32,368 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.4392, 5.1101, 5.4344, 4.9620, 5.0894, 5.0469, 5.4410, 5.1785], device='cuda:0'), covar=tensor([0.0209, 0.0288, 0.0208, 0.0219, 0.0358, 0.0251, 0.0196, 0.0164], device='cuda:0'), in_proj_covar=tensor([0.0275, 0.0282, 0.0299, 0.0274, 0.0275, 0.0275, 0.0250, 0.0224], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-17 05:15:43,070 INFO [finetune.py:1026] (0/2) Epoch 16, validation: loss=0.3133, simple_loss=0.39, pruned_loss=0.1183, over 1020973.00 frames. 2023-05-17 05:15:43,071 INFO [finetune.py:1027] (0/2) Maximum memory allocated so far is 12827MB 2023-05-17 05:15:59,032 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.9189, 3.6740, 5.3116, 2.8246, 3.0525, 3.8939, 3.3771, 3.9337], device='cuda:0'), covar=tensor([0.0369, 0.1072, 0.0303, 0.1204, 0.1909, 0.1562, 0.1321, 0.1219], device='cuda:0'), in_proj_covar=tensor([0.0242, 0.0245, 0.0261, 0.0188, 0.0243, 0.0299, 0.0230, 0.0269], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-17 05:16:18,966 INFO [finetune.py:992] (0/2) Epoch 16, batch 3050, loss[loss=0.1985, simple_loss=0.2764, pruned_loss=0.0603, over 8111.00 frames. ], tot_loss[loss=0.1634, simple_loss=0.2529, pruned_loss=0.037, over 2363248.32 frames. ], batch size: 98, lr: 3.42e-03, grad_scale: 16.0 2023-05-17 05:16:39,045 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=286588.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 05:16:39,594 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.891e+02 2.577e+02 3.122e+02 3.634e+02 7.326e+02, threshold=6.244e+02, percent-clipped=2.0 2023-05-17 05:16:55,652 INFO [finetune.py:992] (0/2) Epoch 16, batch 3100, loss[loss=0.1578, simple_loss=0.2453, pruned_loss=0.03519, over 11813.00 frames. ], tot_loss[loss=0.1635, simple_loss=0.2529, pruned_loss=0.03703, over 2363376.35 frames. ], batch size: 44, lr: 3.42e-03, grad_scale: 16.0 2023-05-17 05:17:01,053 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.58 vs. limit=2.0 2023-05-17 05:17:24,266 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=286649.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 05:17:25,547 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=286651.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 05:17:32,333 INFO [finetune.py:992] (0/2) Epoch 16, batch 3150, loss[loss=0.1385, simple_loss=0.2146, pruned_loss=0.03121, over 12272.00 frames. ], tot_loss[loss=0.1636, simple_loss=0.2529, pruned_loss=0.03714, over 2362633.63 frames. ], batch size: 28, lr: 3.42e-03, grad_scale: 16.0 2023-05-17 05:17:39,753 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=286671.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 05:17:52,330 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.049e+02 2.642e+02 3.020e+02 3.724e+02 5.646e+02, threshold=6.040e+02, percent-clipped=0.0 2023-05-17 05:18:00,291 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=286700.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 05:18:07,792 INFO [finetune.py:992] (0/2) Epoch 16, batch 3200, loss[loss=0.1714, simple_loss=0.2582, pruned_loss=0.04234, over 10589.00 frames. ], tot_loss[loss=0.1639, simple_loss=0.2535, pruned_loss=0.03715, over 2367627.48 frames. ], batch size: 68, lr: 3.42e-03, grad_scale: 16.0 2023-05-17 05:18:23,420 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=286732.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 05:18:43,781 INFO [finetune.py:992] (0/2) Epoch 16, batch 3250, loss[loss=0.1673, simple_loss=0.2608, pruned_loss=0.03691, over 12054.00 frames. ], tot_loss[loss=0.1634, simple_loss=0.2529, pruned_loss=0.037, over 2373481.70 frames. ], batch size: 40, lr: 3.42e-03, grad_scale: 16.0 2023-05-17 05:18:51,047 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=286770.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 05:18:57,228 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=286779.0, num_to_drop=1, layers_to_drop={0} 2023-05-17 05:19:01,457 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=286785.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 05:19:04,244 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.783e+02 2.645e+02 3.195e+02 3.737e+02 6.329e+02, threshold=6.389e+02, percent-clipped=3.0 2023-05-17 05:19:14,851 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.2317, 3.6686, 3.7888, 4.2714, 2.7100, 3.4741, 2.2307, 3.5174], device='cuda:0'), covar=tensor([0.1811, 0.0961, 0.1040, 0.0510, 0.1507, 0.0890, 0.2440, 0.1180], device='cuda:0'), in_proj_covar=tensor([0.0230, 0.0268, 0.0299, 0.0358, 0.0242, 0.0244, 0.0263, 0.0369], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-17 05:19:20,208 INFO [finetune.py:992] (0/2) Epoch 16, batch 3300, loss[loss=0.1648, simple_loss=0.2523, pruned_loss=0.03866, over 12321.00 frames. ], tot_loss[loss=0.1635, simple_loss=0.2532, pruned_loss=0.03693, over 2381064.95 frames. ], batch size: 34, lr: 3.42e-03, grad_scale: 16.0 2023-05-17 05:19:31,433 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=286827.0, num_to_drop=1, layers_to_drop={1} 2023-05-17 05:19:34,311 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=286831.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 05:19:35,606 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=286833.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 05:19:55,549 INFO [finetune.py:992] (0/2) Epoch 16, batch 3350, loss[loss=0.1454, simple_loss=0.2317, pruned_loss=0.02948, over 12094.00 frames. ], tot_loss[loss=0.1635, simple_loss=0.2532, pruned_loss=0.03693, over 2384516.85 frames. ], batch size: 32, lr: 3.42e-03, grad_scale: 16.0 2023-05-17 05:20:04,143 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.4427, 5.0479, 5.4523, 4.7755, 5.1326, 4.8139, 5.4731, 5.0943], device='cuda:0'), covar=tensor([0.0343, 0.0442, 0.0295, 0.0286, 0.0409, 0.0418, 0.0247, 0.0279], device='cuda:0'), in_proj_covar=tensor([0.0279, 0.0285, 0.0302, 0.0277, 0.0278, 0.0279, 0.0252, 0.0227], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-17 05:20:15,939 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.971e+02 2.653e+02 2.997e+02 3.607e+02 6.484e+02, threshold=5.995e+02, percent-clipped=1.0 2023-05-17 05:20:31,628 INFO [finetune.py:992] (0/2) Epoch 16, batch 3400, loss[loss=0.1712, simple_loss=0.2645, pruned_loss=0.03895, over 12064.00 frames. ], tot_loss[loss=0.1638, simple_loss=0.2533, pruned_loss=0.03711, over 2388620.36 frames. ], batch size: 42, lr: 3.42e-03, grad_scale: 16.0 2023-05-17 05:20:55,601 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=286944.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 05:21:00,672 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=286951.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 05:21:07,602 INFO [finetune.py:992] (0/2) Epoch 16, batch 3450, loss[loss=0.1694, simple_loss=0.2581, pruned_loss=0.04037, over 11274.00 frames. ], tot_loss[loss=0.1638, simple_loss=0.2534, pruned_loss=0.03708, over 2389269.04 frames. ], batch size: 55, lr: 3.41e-03, grad_scale: 16.0 2023-05-17 05:21:17,432 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=2.25 vs. limit=5.0 2023-05-17 05:21:27,923 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.987e+02 2.606e+02 3.061e+02 3.584e+02 6.458e+02, threshold=6.122e+02, percent-clipped=1.0 2023-05-17 05:21:35,161 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=286999.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 05:21:36,248 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=287000.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 05:21:43,897 INFO [finetune.py:992] (0/2) Epoch 16, batch 3500, loss[loss=0.1709, simple_loss=0.253, pruned_loss=0.04442, over 12081.00 frames. ], tot_loss[loss=0.163, simple_loss=0.2524, pruned_loss=0.03676, over 2387261.35 frames. ], batch size: 32, lr: 3.41e-03, grad_scale: 16.0 2023-05-17 05:21:55,456 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=287027.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 05:22:11,118 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=287048.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 05:22:20,978 INFO [finetune.py:992] (0/2) Epoch 16, batch 3550, loss[loss=0.1411, simple_loss=0.2312, pruned_loss=0.02551, over 12085.00 frames. ], tot_loss[loss=0.1631, simple_loss=0.2526, pruned_loss=0.03679, over 2389224.11 frames. ], batch size: 32, lr: 3.41e-03, grad_scale: 16.0 2023-05-17 05:22:41,036 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.779e+02 2.536e+02 2.993e+02 3.580e+02 5.373e+02, threshold=5.986e+02, percent-clipped=0.0 2023-05-17 05:22:44,107 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.6075, 2.5568, 3.2303, 4.3903, 2.2886, 4.4467, 4.6108, 4.6665], device='cuda:0'), covar=tensor([0.0128, 0.1401, 0.0571, 0.0177, 0.1460, 0.0269, 0.0153, 0.0110], device='cuda:0'), in_proj_covar=tensor([0.0124, 0.0205, 0.0184, 0.0123, 0.0191, 0.0182, 0.0178, 0.0126], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-17 05:22:56,518 INFO [finetune.py:992] (0/2) Epoch 16, batch 3600, loss[loss=0.1611, simple_loss=0.2514, pruned_loss=0.03538, over 12135.00 frames. ], tot_loss[loss=0.1627, simple_loss=0.2521, pruned_loss=0.03664, over 2387335.60 frames. ], batch size: 30, lr: 3.41e-03, grad_scale: 16.0 2023-05-17 05:23:07,164 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=287126.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 05:23:21,674 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.7145, 2.9089, 4.4134, 4.5071, 2.8915, 2.4830, 2.9156, 2.1875], device='cuda:0'), covar=tensor([0.1609, 0.2904, 0.0481, 0.0453, 0.1326, 0.2602, 0.2735, 0.3985], device='cuda:0'), in_proj_covar=tensor([0.0309, 0.0399, 0.0285, 0.0308, 0.0283, 0.0324, 0.0404, 0.0386], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-17 05:23:23,703 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.0898, 4.8886, 4.8196, 4.8255, 4.5805, 4.9931, 4.9645, 5.0731], device='cuda:0'), covar=tensor([0.0204, 0.0169, 0.0203, 0.0407, 0.0740, 0.0360, 0.0156, 0.0187], device='cuda:0'), in_proj_covar=tensor([0.0205, 0.0202, 0.0198, 0.0256, 0.0248, 0.0230, 0.0183, 0.0240], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0005, 0.0004, 0.0004, 0.0003, 0.0005], device='cuda:0') 2023-05-17 05:23:32,141 INFO [finetune.py:992] (0/2) Epoch 16, batch 3650, loss[loss=0.1383, simple_loss=0.2277, pruned_loss=0.0244, over 12155.00 frames. ], tot_loss[loss=0.1625, simple_loss=0.2517, pruned_loss=0.03667, over 2392461.27 frames. ], batch size: 34, lr: 3.41e-03, grad_scale: 16.0 2023-05-17 05:23:52,647 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.588e+02 2.548e+02 3.085e+02 3.772e+02 8.104e+02, threshold=6.169e+02, percent-clipped=4.0 2023-05-17 05:24:09,415 INFO [finetune.py:992] (0/2) Epoch 16, batch 3700, loss[loss=0.1604, simple_loss=0.2452, pruned_loss=0.03782, over 12101.00 frames. ], tot_loss[loss=0.1627, simple_loss=0.2522, pruned_loss=0.03655, over 2391412.73 frames. ], batch size: 33, lr: 3.41e-03, grad_scale: 16.0 2023-05-17 05:24:33,328 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=287244.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 05:24:45,394 INFO [finetune.py:992] (0/2) Epoch 16, batch 3750, loss[loss=0.1818, simple_loss=0.2682, pruned_loss=0.04769, over 12147.00 frames. ], tot_loss[loss=0.1627, simple_loss=0.2519, pruned_loss=0.03672, over 2382868.20 frames. ], batch size: 39, lr: 3.41e-03, grad_scale: 16.0 2023-05-17 05:25:05,615 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.986e+02 2.523e+02 2.915e+02 3.511e+02 5.231e+02, threshold=5.830e+02, percent-clipped=0.0 2023-05-17 05:25:06,019 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.99 vs. limit=2.0 2023-05-17 05:25:07,708 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=287292.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 05:25:21,161 INFO [finetune.py:992] (0/2) Epoch 16, batch 3800, loss[loss=0.1947, simple_loss=0.2648, pruned_loss=0.06227, over 8112.00 frames. ], tot_loss[loss=0.1629, simple_loss=0.2521, pruned_loss=0.03683, over 2375027.16 frames. ], batch size: 98, lr: 3.41e-03, grad_scale: 16.0 2023-05-17 05:25:33,318 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=287327.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 05:25:58,055 INFO [finetune.py:992] (0/2) Epoch 16, batch 3850, loss[loss=0.1936, simple_loss=0.2721, pruned_loss=0.0576, over 8185.00 frames. ], tot_loss[loss=0.1635, simple_loss=0.2526, pruned_loss=0.03721, over 2360875.77 frames. ], batch size: 98, lr: 3.41e-03, grad_scale: 16.0 2023-05-17 05:26:03,277 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.4729, 5.0746, 5.4093, 4.7512, 5.1449, 4.7738, 5.3707, 5.1734], device='cuda:0'), covar=tensor([0.0384, 0.0532, 0.0432, 0.0345, 0.0455, 0.0399, 0.0450, 0.0284], device='cuda:0'), in_proj_covar=tensor([0.0277, 0.0283, 0.0299, 0.0273, 0.0276, 0.0275, 0.0250, 0.0225], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-17 05:26:08,176 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=287375.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 05:26:18,140 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.852e+02 2.775e+02 3.174e+02 3.684e+02 7.449e+02, threshold=6.349e+02, percent-clipped=3.0 2023-05-17 05:26:33,965 INFO [finetune.py:992] (0/2) Epoch 16, batch 3900, loss[loss=0.1362, simple_loss=0.2154, pruned_loss=0.02851, over 12282.00 frames. ], tot_loss[loss=0.1637, simple_loss=0.2528, pruned_loss=0.03727, over 2363182.64 frames. ], batch size: 28, lr: 3.41e-03, grad_scale: 16.0 2023-05-17 05:26:34,191 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=287411.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 05:26:44,674 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=287426.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 05:27:09,359 INFO [finetune.py:992] (0/2) Epoch 16, batch 3950, loss[loss=0.1479, simple_loss=0.2383, pruned_loss=0.02877, over 12285.00 frames. ], tot_loss[loss=0.1632, simple_loss=0.2523, pruned_loss=0.037, over 2363249.95 frames. ], batch size: 33, lr: 3.41e-03, grad_scale: 16.0 2023-05-17 05:27:17,675 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=4.62 vs. limit=5.0 2023-05-17 05:27:18,068 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=287472.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 05:27:19,374 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=287474.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 05:27:21,970 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.38 vs. limit=2.0 2023-05-17 05:27:30,785 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.581e+02 2.806e+02 3.106e+02 3.714e+02 5.718e+02, threshold=6.213e+02, percent-clipped=0.0 2023-05-17 05:27:46,593 INFO [finetune.py:992] (0/2) Epoch 16, batch 4000, loss[loss=0.1786, simple_loss=0.267, pruned_loss=0.04504, over 11172.00 frames. ], tot_loss[loss=0.1637, simple_loss=0.2528, pruned_loss=0.03735, over 2366855.93 frames. ], batch size: 55, lr: 3.41e-03, grad_scale: 16.0 2023-05-17 05:28:04,676 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=287536.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 05:28:18,330 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.94 vs. limit=2.0 2023-05-17 05:28:21,773 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.39 vs. limit=2.0 2023-05-17 05:28:21,989 INFO [finetune.py:992] (0/2) Epoch 16, batch 4050, loss[loss=0.166, simple_loss=0.2613, pruned_loss=0.0354, over 12117.00 frames. ], tot_loss[loss=0.1647, simple_loss=0.2536, pruned_loss=0.03793, over 2363933.29 frames. ], batch size: 33, lr: 3.41e-03, grad_scale: 16.0 2023-05-17 05:28:22,763 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=287562.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 05:28:26,194 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.2906, 6.2488, 5.7670, 5.8113, 6.2646, 5.5442, 5.7161, 5.7603], device='cuda:0'), covar=tensor([0.1546, 0.0917, 0.1363, 0.1872, 0.0981, 0.2341, 0.1944, 0.1149], device='cuda:0'), in_proj_covar=tensor([0.0362, 0.0498, 0.0403, 0.0449, 0.0470, 0.0433, 0.0399, 0.0391], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-17 05:28:36,440 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.6462, 3.3147, 5.1369, 2.6533, 2.8820, 3.8337, 3.1839, 3.8812], device='cuda:0'), covar=tensor([0.0443, 0.1233, 0.0278, 0.1271, 0.1935, 0.1412, 0.1472, 0.1189], device='cuda:0'), in_proj_covar=tensor([0.0243, 0.0245, 0.0263, 0.0189, 0.0244, 0.0299, 0.0231, 0.0272], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-17 05:28:37,783 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=287583.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 05:28:41,930 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.034e+02 2.640e+02 3.108e+02 3.631e+02 7.122e+02, threshold=6.216e+02, percent-clipped=2.0 2023-05-17 05:28:47,761 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=287597.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 05:28:49,261 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.1052, 2.6139, 3.6539, 3.0802, 3.4389, 3.2116, 2.6911, 3.5419], device='cuda:0'), covar=tensor([0.0147, 0.0345, 0.0159, 0.0246, 0.0187, 0.0191, 0.0337, 0.0149], device='cuda:0'), in_proj_covar=tensor([0.0188, 0.0212, 0.0202, 0.0196, 0.0227, 0.0174, 0.0205, 0.0196], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-17 05:28:58,562 INFO [finetune.py:992] (0/2) Epoch 16, batch 4100, loss[loss=0.1746, simple_loss=0.2648, pruned_loss=0.04213, over 12297.00 frames. ], tot_loss[loss=0.1651, simple_loss=0.2542, pruned_loss=0.03803, over 2368521.27 frames. ], batch size: 34, lr: 3.41e-03, grad_scale: 16.0 2023-05-17 05:29:01,014 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.5367, 3.3062, 5.0390, 2.6271, 2.6420, 3.6547, 3.0525, 3.6264], device='cuda:0'), covar=tensor([0.0524, 0.1214, 0.0372, 0.1269, 0.2143, 0.1616, 0.1502, 0.1365], device='cuda:0'), in_proj_covar=tensor([0.0243, 0.0245, 0.0263, 0.0189, 0.0243, 0.0299, 0.0231, 0.0271], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-17 05:29:07,405 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=287623.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 05:29:23,106 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=287644.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 05:29:29,521 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.3821, 4.7554, 3.0060, 2.9015, 4.1052, 2.7870, 3.9516, 3.3955], device='cuda:0'), covar=tensor([0.0719, 0.0558, 0.1122, 0.1403, 0.0346, 0.1293, 0.0511, 0.0770], device='cuda:0'), in_proj_covar=tensor([0.0193, 0.0264, 0.0180, 0.0206, 0.0146, 0.0188, 0.0201, 0.0176], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-17 05:29:29,725 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.98 vs. limit=2.0 2023-05-17 05:29:35,003 INFO [finetune.py:992] (0/2) Epoch 16, batch 4150, loss[loss=0.1652, simple_loss=0.2588, pruned_loss=0.03582, over 12044.00 frames. ], tot_loss[loss=0.1647, simple_loss=0.2542, pruned_loss=0.03766, over 2375452.86 frames. ], batch size: 37, lr: 3.41e-03, grad_scale: 16.0 2023-05-17 05:29:55,480 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.709e+02 2.445e+02 3.091e+02 3.464e+02 5.875e+02, threshold=6.183e+02, percent-clipped=0.0 2023-05-17 05:30:08,297 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.3735, 3.4705, 3.2192, 3.0518, 2.7896, 2.7178, 3.5345, 2.3422], device='cuda:0'), covar=tensor([0.0465, 0.0156, 0.0220, 0.0258, 0.0482, 0.0374, 0.0143, 0.0533], device='cuda:0'), in_proj_covar=tensor([0.0199, 0.0167, 0.0171, 0.0193, 0.0205, 0.0204, 0.0176, 0.0209], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-17 05:30:08,947 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.1667, 4.7365, 4.8754, 4.9449, 4.7394, 5.0108, 4.9517, 2.8658], device='cuda:0'), covar=tensor([0.0105, 0.0066, 0.0087, 0.0068, 0.0052, 0.0089, 0.0068, 0.0751], device='cuda:0'), in_proj_covar=tensor([0.0070, 0.0080, 0.0083, 0.0074, 0.0061, 0.0094, 0.0082, 0.0098], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-17 05:30:10,886 INFO [finetune.py:992] (0/2) Epoch 16, batch 4200, loss[loss=0.16, simple_loss=0.2541, pruned_loss=0.03291, over 11618.00 frames. ], tot_loss[loss=0.1648, simple_loss=0.2543, pruned_loss=0.03761, over 2377245.51 frames. ], batch size: 48, lr: 3.41e-03, grad_scale: 8.0 2023-05-17 05:30:17,323 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=287720.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 05:30:18,806 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.1334, 2.5075, 3.7035, 3.0779, 3.5353, 3.1730, 2.6656, 3.6179], device='cuda:0'), covar=tensor([0.0142, 0.0366, 0.0145, 0.0268, 0.0141, 0.0212, 0.0359, 0.0123], device='cuda:0'), in_proj_covar=tensor([0.0186, 0.0210, 0.0200, 0.0194, 0.0225, 0.0173, 0.0203, 0.0195], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-17 05:30:42,428 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=4.48 vs. limit=5.0 2023-05-17 05:30:46,910 INFO [finetune.py:992] (0/2) Epoch 16, batch 4250, loss[loss=0.1661, simple_loss=0.2533, pruned_loss=0.03947, over 12361.00 frames. ], tot_loss[loss=0.1652, simple_loss=0.2547, pruned_loss=0.03784, over 2381645.85 frames. ], batch size: 35, lr: 3.41e-03, grad_scale: 8.0 2023-05-17 05:30:51,459 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=287767.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 05:30:55,961 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=287773.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 05:31:01,703 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=287781.0, num_to_drop=1, layers_to_drop={2} 2023-05-17 05:31:08,557 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.968e+02 2.769e+02 3.131e+02 3.644e+02 2.115e+03, threshold=6.263e+02, percent-clipped=1.0 2023-05-17 05:31:23,789 INFO [finetune.py:992] (0/2) Epoch 16, batch 4300, loss[loss=0.1678, simple_loss=0.2588, pruned_loss=0.03845, over 12192.00 frames. ], tot_loss[loss=0.1651, simple_loss=0.2547, pruned_loss=0.03774, over 2381654.82 frames. ], batch size: 35, lr: 3.41e-03, grad_scale: 8.0 2023-05-17 05:31:40,629 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=287834.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 05:31:59,318 INFO [finetune.py:992] (0/2) Epoch 16, batch 4350, loss[loss=0.16, simple_loss=0.2597, pruned_loss=0.03013, over 12110.00 frames. ], tot_loss[loss=0.1654, simple_loss=0.2548, pruned_loss=0.038, over 2377198.47 frames. ], batch size: 39, lr: 3.41e-03, grad_scale: 8.0 2023-05-17 05:32:02,295 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.2004, 2.6772, 3.7785, 3.2160, 3.4919, 3.2844, 2.8954, 3.6210], device='cuda:0'), covar=tensor([0.0165, 0.0374, 0.0168, 0.0274, 0.0183, 0.0205, 0.0327, 0.0139], device='cuda:0'), in_proj_covar=tensor([0.0188, 0.0212, 0.0201, 0.0195, 0.0227, 0.0175, 0.0205, 0.0197], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-17 05:32:03,704 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.5882, 2.8974, 3.2649, 4.4456, 2.3496, 4.4107, 4.5992, 4.6858], device='cuda:0'), covar=tensor([0.0142, 0.1165, 0.0512, 0.0191, 0.1387, 0.0293, 0.0156, 0.0119], device='cuda:0'), in_proj_covar=tensor([0.0124, 0.0204, 0.0183, 0.0123, 0.0191, 0.0182, 0.0177, 0.0126], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-17 05:32:06,508 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.3028, 6.1794, 5.7895, 5.7230, 6.2432, 5.6361, 5.6072, 5.7513], device='cuda:0'), covar=tensor([0.1634, 0.0917, 0.1029, 0.1752, 0.0845, 0.2102, 0.2079, 0.1136], device='cuda:0'), in_proj_covar=tensor([0.0369, 0.0506, 0.0408, 0.0458, 0.0477, 0.0440, 0.0404, 0.0397], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-17 05:32:19,829 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.668e+02 2.910e+02 3.370e+02 3.789e+02 2.398e+03, threshold=6.741e+02, percent-clipped=1.0 2023-05-17 05:32:21,378 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=287892.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 05:32:35,329 INFO [finetune.py:992] (0/2) Epoch 16, batch 4400, loss[loss=0.1538, simple_loss=0.2431, pruned_loss=0.03222, over 12344.00 frames. ], tot_loss[loss=0.1651, simple_loss=0.2545, pruned_loss=0.0378, over 2377714.31 frames. ], batch size: 31, lr: 3.41e-03, grad_scale: 8.0 2023-05-17 05:32:40,443 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=287918.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 05:32:56,142 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=287939.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 05:32:59,734 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.1023, 2.2316, 2.6826, 3.1138, 2.1940, 3.1839, 3.1493, 3.2204], device='cuda:0'), covar=tensor([0.0217, 0.1028, 0.0478, 0.0200, 0.1060, 0.0328, 0.0292, 0.0168], device='cuda:0'), in_proj_covar=tensor([0.0123, 0.0204, 0.0183, 0.0123, 0.0191, 0.0182, 0.0177, 0.0126], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-17 05:33:11,643 INFO [finetune.py:992] (0/2) Epoch 16, batch 4450, loss[loss=0.1432, simple_loss=0.2349, pruned_loss=0.02574, over 12148.00 frames. ], tot_loss[loss=0.1651, simple_loss=0.2547, pruned_loss=0.03774, over 2369844.45 frames. ], batch size: 34, lr: 3.41e-03, grad_scale: 8.0 2023-05-17 05:33:23,207 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.0326, 3.8550, 4.0170, 3.7177, 3.8654, 3.6950, 3.9944, 3.6370], device='cuda:0'), covar=tensor([0.0368, 0.0391, 0.0330, 0.0269, 0.0391, 0.0361, 0.0298, 0.1373], device='cuda:0'), in_proj_covar=tensor([0.0280, 0.0286, 0.0304, 0.0277, 0.0279, 0.0279, 0.0254, 0.0228], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-17 05:33:32,224 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.706e+02 2.609e+02 3.072e+02 3.662e+02 6.835e+02, threshold=6.144e+02, percent-clipped=1.0 2023-05-17 05:33:39,766 INFO [checkpoint.py:75] (0/2) Saving checkpoint to pruned_transducer_stateless7/exp_giga_finetune/checkpoint-188000.pt 2023-05-17 05:33:45,247 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=288003.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 05:33:50,918 INFO [finetune.py:992] (0/2) Epoch 16, batch 4500, loss[loss=0.1498, simple_loss=0.2467, pruned_loss=0.02645, over 12282.00 frames. ], tot_loss[loss=0.164, simple_loss=0.2536, pruned_loss=0.03716, over 2377922.24 frames. ], batch size: 33, lr: 3.41e-03, grad_scale: 8.0 2023-05-17 05:34:10,279 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.6143, 3.6335, 3.2998, 3.1879, 2.9660, 2.8265, 3.6496, 2.4095], device='cuda:0'), covar=tensor([0.0434, 0.0164, 0.0228, 0.0229, 0.0412, 0.0387, 0.0176, 0.0520], device='cuda:0'), in_proj_covar=tensor([0.0199, 0.0167, 0.0171, 0.0194, 0.0206, 0.0203, 0.0177, 0.0209], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-17 05:34:27,055 INFO [finetune.py:992] (0/2) Epoch 16, batch 4550, loss[loss=0.1851, simple_loss=0.2745, pruned_loss=0.04788, over 11795.00 frames. ], tot_loss[loss=0.165, simple_loss=0.2543, pruned_loss=0.03785, over 2363770.00 frames. ], batch size: 44, lr: 3.41e-03, grad_scale: 8.0 2023-05-17 05:34:29,369 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=288064.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 05:34:31,433 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=288067.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 05:34:38,573 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=288076.0, num_to_drop=1, layers_to_drop={1} 2023-05-17 05:34:48,455 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.884e+02 2.623e+02 3.177e+02 3.615e+02 5.730e+02, threshold=6.355e+02, percent-clipped=0.0 2023-05-17 05:34:51,660 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=2.24 vs. limit=5.0 2023-05-17 05:34:59,865 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=288106.0, num_to_drop=1, layers_to_drop={0} 2023-05-17 05:35:03,221 INFO [finetune.py:992] (0/2) Epoch 16, batch 4600, loss[loss=0.1647, simple_loss=0.2677, pruned_loss=0.03083, over 12340.00 frames. ], tot_loss[loss=0.1647, simple_loss=0.2544, pruned_loss=0.03747, over 2372489.39 frames. ], batch size: 36, lr: 3.41e-03, grad_scale: 8.0 2023-05-17 05:35:06,214 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=288115.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 05:35:16,260 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=288129.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 05:35:23,109 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.39 vs. limit=2.0 2023-05-17 05:35:32,571 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=288152.0, num_to_drop=1, layers_to_drop={0} 2023-05-17 05:35:38,561 INFO [finetune.py:992] (0/2) Epoch 16, batch 4650, loss[loss=0.2143, simple_loss=0.2882, pruned_loss=0.07018, over 8011.00 frames. ], tot_loss[loss=0.1652, simple_loss=0.2548, pruned_loss=0.03786, over 2363152.88 frames. ], batch size: 101, lr: 3.41e-03, grad_scale: 8.0 2023-05-17 05:35:43,054 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=288167.0, num_to_drop=1, layers_to_drop={3} 2023-05-17 05:36:00,069 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.899e+02 2.786e+02 3.203e+02 3.745e+02 6.992e+02, threshold=6.407e+02, percent-clipped=1.0 2023-05-17 05:36:01,470 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=288192.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 05:36:15,076 INFO [finetune.py:992] (0/2) Epoch 16, batch 4700, loss[loss=0.127, simple_loss=0.2148, pruned_loss=0.01965, over 12270.00 frames. ], tot_loss[loss=0.165, simple_loss=0.2544, pruned_loss=0.03783, over 2370064.73 frames. ], batch size: 28, lr: 3.41e-03, grad_scale: 8.0 2023-05-17 05:36:16,678 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=288213.0, num_to_drop=1, layers_to_drop={2} 2023-05-17 05:36:20,193 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=288218.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 05:36:32,638 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.35 vs. limit=2.0 2023-05-17 05:36:35,753 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=288239.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 05:36:36,396 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=288240.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 05:36:47,793 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.1535, 5.0233, 4.9201, 5.0064, 4.6774, 5.1783, 5.0927, 5.2789], device='cuda:0'), covar=tensor([0.0306, 0.0161, 0.0207, 0.0302, 0.0746, 0.0262, 0.0142, 0.0163], device='cuda:0'), in_proj_covar=tensor([0.0204, 0.0202, 0.0197, 0.0255, 0.0246, 0.0229, 0.0182, 0.0240], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0005, 0.0004, 0.0004, 0.0003, 0.0005], device='cuda:0') 2023-05-17 05:36:51,211 INFO [finetune.py:992] (0/2) Epoch 16, batch 4750, loss[loss=0.1779, simple_loss=0.2723, pruned_loss=0.04178, over 11214.00 frames. ], tot_loss[loss=0.1658, simple_loss=0.2551, pruned_loss=0.0382, over 2360013.12 frames. ], batch size: 55, lr: 3.41e-03, grad_scale: 8.0 2023-05-17 05:36:54,739 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=288266.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 05:37:00,646 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.4540, 4.8375, 4.3089, 5.1311, 4.5788, 3.1348, 4.2807, 3.1260], device='cuda:0'), covar=tensor([0.0831, 0.0712, 0.1286, 0.0539, 0.1173, 0.1640, 0.1140, 0.3491], device='cuda:0'), in_proj_covar=tensor([0.0312, 0.0384, 0.0365, 0.0333, 0.0377, 0.0278, 0.0353, 0.0372], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-17 05:37:09,778 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=288287.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 05:37:11,885 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.140e+02 2.896e+02 3.408e+02 4.008e+02 6.839e+02, threshold=6.815e+02, percent-clipped=1.0 2023-05-17 05:37:24,729 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.0633, 5.9205, 5.5804, 5.4039, 6.0379, 5.4465, 5.3665, 5.5430], device='cuda:0'), covar=tensor([0.1491, 0.0953, 0.0956, 0.2119, 0.0887, 0.1773, 0.2256, 0.1063], device='cuda:0'), in_proj_covar=tensor([0.0368, 0.0508, 0.0408, 0.0462, 0.0479, 0.0441, 0.0404, 0.0397], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-17 05:37:26,692 INFO [finetune.py:992] (0/2) Epoch 16, batch 4800, loss[loss=0.1676, simple_loss=0.2654, pruned_loss=0.03495, over 12285.00 frames. ], tot_loss[loss=0.1649, simple_loss=0.2544, pruned_loss=0.03765, over 2369609.12 frames. ], batch size: 37, lr: 3.41e-03, grad_scale: 8.0 2023-05-17 05:37:38,274 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=288327.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 05:37:51,742 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=288345.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 05:38:01,587 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=288359.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 05:38:02,953 INFO [finetune.py:992] (0/2) Epoch 16, batch 4850, loss[loss=0.1536, simple_loss=0.2501, pruned_loss=0.02853, over 12312.00 frames. ], tot_loss[loss=0.1648, simple_loss=0.2544, pruned_loss=0.03756, over 2374221.45 frames. ], batch size: 34, lr: 3.40e-03, grad_scale: 8.0 2023-05-17 05:38:14,343 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=288376.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 05:38:20,161 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.45 vs. limit=2.0 2023-05-17 05:38:22,843 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=288388.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 05:38:24,079 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.011e+02 2.698e+02 3.119e+02 3.725e+02 7.730e+02, threshold=6.239e+02, percent-clipped=1.0 2023-05-17 05:38:28,612 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=288396.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 05:38:35,862 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=288406.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 05:38:39,180 INFO [finetune.py:992] (0/2) Epoch 16, batch 4900, loss[loss=0.1585, simple_loss=0.2457, pruned_loss=0.03567, over 12154.00 frames. ], tot_loss[loss=0.1648, simple_loss=0.2546, pruned_loss=0.0375, over 2361764.03 frames. ], batch size: 36, lr: 3.40e-03, grad_scale: 8.0 2023-05-17 05:38:43,665 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.2402, 4.4584, 2.8668, 2.8550, 3.9131, 2.5099, 3.8242, 3.1257], device='cuda:0'), covar=tensor([0.0735, 0.0568, 0.1162, 0.1362, 0.0306, 0.1429, 0.0526, 0.0874], device='cuda:0'), in_proj_covar=tensor([0.0192, 0.0264, 0.0179, 0.0206, 0.0146, 0.0187, 0.0201, 0.0177], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-17 05:38:48,469 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=288424.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 05:38:52,144 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=288429.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 05:39:12,012 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=288457.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 05:39:14,049 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.2152, 3.8914, 4.1172, 4.3188, 2.7414, 4.0000, 2.6114, 4.0149], device='cuda:0'), covar=tensor([0.1633, 0.0785, 0.0869, 0.0604, 0.1321, 0.0580, 0.1822, 0.1104], device='cuda:0'), in_proj_covar=tensor([0.0230, 0.0269, 0.0301, 0.0360, 0.0244, 0.0245, 0.0263, 0.0372], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-17 05:39:14,526 INFO [finetune.py:992] (0/2) Epoch 16, batch 4950, loss[loss=0.1584, simple_loss=0.2548, pruned_loss=0.03094, over 12277.00 frames. ], tot_loss[loss=0.1648, simple_loss=0.2543, pruned_loss=0.03763, over 2360006.12 frames. ], batch size: 37, lr: 3.40e-03, grad_scale: 8.0 2023-05-17 05:39:15,328 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=288462.0, num_to_drop=1, layers_to_drop={0} 2023-05-17 05:39:26,321 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=288477.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 05:39:35,409 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.842e+02 2.739e+02 3.110e+02 3.721e+02 3.368e+03, threshold=6.220e+02, percent-clipped=4.0 2023-05-17 05:39:43,878 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.58 vs. limit=2.0 2023-05-17 05:39:46,532 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.6257, 2.7970, 4.2406, 4.4474, 2.8912, 2.5244, 2.8347, 2.1644], device='cuda:0'), covar=tensor([0.1706, 0.2978, 0.0564, 0.0443, 0.1317, 0.2600, 0.2938, 0.4238], device='cuda:0'), in_proj_covar=tensor([0.0306, 0.0396, 0.0283, 0.0306, 0.0281, 0.0322, 0.0400, 0.0382], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-17 05:39:49,186 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=288508.0, num_to_drop=1, layers_to_drop={1} 2023-05-17 05:39:51,247 INFO [finetune.py:992] (0/2) Epoch 16, batch 5000, loss[loss=0.1456, simple_loss=0.234, pruned_loss=0.02859, over 12115.00 frames. ], tot_loss[loss=0.1644, simple_loss=0.254, pruned_loss=0.03742, over 2352857.39 frames. ], batch size: 33, lr: 3.40e-03, grad_scale: 8.0 2023-05-17 05:40:12,105 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.7263, 2.6788, 3.9683, 4.1682, 2.9150, 2.5910, 2.7648, 2.2651], device='cuda:0'), covar=tensor([0.1689, 0.2861, 0.0618, 0.0471, 0.1230, 0.2644, 0.2796, 0.4058], device='cuda:0'), in_proj_covar=tensor([0.0305, 0.0395, 0.0282, 0.0305, 0.0280, 0.0321, 0.0399, 0.0381], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-17 05:40:26,549 INFO [finetune.py:992] (0/2) Epoch 16, batch 5050, loss[loss=0.1875, simple_loss=0.2828, pruned_loss=0.04608, over 11342.00 frames. ], tot_loss[loss=0.165, simple_loss=0.2549, pruned_loss=0.03756, over 2353701.69 frames. ], batch size: 55, lr: 3.40e-03, grad_scale: 8.0 2023-05-17 05:40:47,124 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.419e+02 2.542e+02 3.078e+02 3.605e+02 6.274e+02, threshold=6.155e+02, percent-clipped=1.0 2023-05-17 05:40:55,134 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.5472, 5.1592, 5.5200, 4.9470, 5.1362, 4.9985, 5.5785, 5.1148], device='cuda:0'), covar=tensor([0.0249, 0.0365, 0.0261, 0.0228, 0.0398, 0.0313, 0.0176, 0.0281], device='cuda:0'), in_proj_covar=tensor([0.0276, 0.0282, 0.0301, 0.0274, 0.0277, 0.0276, 0.0249, 0.0225], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-17 05:41:00,907 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=288609.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 05:41:02,187 INFO [finetune.py:992] (0/2) Epoch 16, batch 5100, loss[loss=0.1573, simple_loss=0.2503, pruned_loss=0.03215, over 12138.00 frames. ], tot_loss[loss=0.1638, simple_loss=0.2536, pruned_loss=0.03697, over 2362427.04 frames. ], batch size: 34, lr: 3.40e-03, grad_scale: 8.0 2023-05-17 05:41:22,378 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.82 vs. limit=2.0 2023-05-17 05:41:32,677 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.1776, 5.0198, 4.8817, 4.9257, 4.6948, 5.0912, 5.0418, 5.2866], device='cuda:0'), covar=tensor([0.0251, 0.0155, 0.0205, 0.0358, 0.0720, 0.0309, 0.0170, 0.0173], device='cuda:0'), in_proj_covar=tensor([0.0203, 0.0201, 0.0197, 0.0256, 0.0246, 0.0228, 0.0183, 0.0239], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0005, 0.0004, 0.0004, 0.0003, 0.0005], device='cuda:0') 2023-05-17 05:41:37,772 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=288659.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 05:41:39,080 INFO [finetune.py:992] (0/2) Epoch 16, batch 5150, loss[loss=0.1735, simple_loss=0.2681, pruned_loss=0.03941, over 10492.00 frames. ], tot_loss[loss=0.1644, simple_loss=0.2538, pruned_loss=0.03748, over 2357998.19 frames. ], batch size: 68, lr: 3.40e-03, grad_scale: 8.0 2023-05-17 05:41:45,622 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=288670.0, num_to_drop=1, layers_to_drop={0} 2023-05-17 05:41:54,732 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=288683.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 05:41:59,641 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.783e+02 2.642e+02 3.043e+02 3.741e+02 6.453e+02, threshold=6.085e+02, percent-clipped=1.0 2023-05-17 05:42:07,207 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=288701.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 05:42:08,161 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.48 vs. limit=2.0 2023-05-17 05:42:11,328 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=288707.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 05:42:14,090 INFO [finetune.py:992] (0/2) Epoch 16, batch 5200, loss[loss=0.1554, simple_loss=0.247, pruned_loss=0.0319, over 12038.00 frames. ], tot_loss[loss=0.1647, simple_loss=0.2539, pruned_loss=0.03771, over 2351599.87 frames. ], batch size: 40, lr: 3.40e-03, grad_scale: 8.0 2023-05-17 05:42:43,593 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=288752.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 05:42:50,562 INFO [finetune.py:992] (0/2) Epoch 16, batch 5250, loss[loss=0.1866, simple_loss=0.2793, pruned_loss=0.04689, over 12351.00 frames. ], tot_loss[loss=0.1645, simple_loss=0.2537, pruned_loss=0.03767, over 2350915.03 frames. ], batch size: 38, lr: 3.40e-03, grad_scale: 8.0 2023-05-17 05:42:51,473 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=288762.0, num_to_drop=1, layers_to_drop={2} 2023-05-17 05:42:55,572 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.9223, 5.6860, 5.3410, 5.2424, 5.7798, 5.0377, 5.1930, 5.2289], device='cuda:0'), covar=tensor([0.1610, 0.0910, 0.0965, 0.1841, 0.0828, 0.2241, 0.1975, 0.1113], device='cuda:0'), in_proj_covar=tensor([0.0370, 0.0510, 0.0411, 0.0460, 0.0479, 0.0442, 0.0402, 0.0399], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-17 05:43:02,153 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.2271, 4.0488, 4.2220, 4.4141, 3.0778, 3.9927, 2.7948, 4.1682], device='cuda:0'), covar=tensor([0.1612, 0.0710, 0.0766, 0.0621, 0.1139, 0.0607, 0.1698, 0.1027], device='cuda:0'), in_proj_covar=tensor([0.0229, 0.0269, 0.0300, 0.0360, 0.0244, 0.0245, 0.0263, 0.0371], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-17 05:43:11,481 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.2790, 3.6206, 3.3696, 3.1687, 3.0029, 2.8915, 3.6223, 2.2561], device='cuda:0'), covar=tensor([0.0556, 0.0168, 0.0220, 0.0231, 0.0377, 0.0349, 0.0146, 0.0582], device='cuda:0'), in_proj_covar=tensor([0.0198, 0.0168, 0.0172, 0.0194, 0.0205, 0.0205, 0.0178, 0.0210], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-17 05:43:11,940 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.852e+02 2.670e+02 3.124e+02 3.667e+02 1.117e+03, threshold=6.248e+02, percent-clipped=2.0 2023-05-17 05:43:17,135 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=288797.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 05:43:25,231 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=288808.0, num_to_drop=1, layers_to_drop={0} 2023-05-17 05:43:25,955 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.9557, 5.8906, 5.6479, 5.1058, 5.1529, 5.7876, 5.4024, 5.2011], device='cuda:0'), covar=tensor([0.0704, 0.0941, 0.0724, 0.1662, 0.0736, 0.0702, 0.1609, 0.1095], device='cuda:0'), in_proj_covar=tensor([0.0647, 0.0581, 0.0538, 0.0656, 0.0426, 0.0755, 0.0801, 0.0585], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0004, 0.0004, 0.0003], device='cuda:0') 2023-05-17 05:43:26,593 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=288810.0, num_to_drop=1, layers_to_drop={0} 2023-05-17 05:43:26,884 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2023-05-17 05:43:27,197 INFO [finetune.py:992] (0/2) Epoch 16, batch 5300, loss[loss=0.1715, simple_loss=0.2486, pruned_loss=0.04725, over 12207.00 frames. ], tot_loss[loss=0.1638, simple_loss=0.2527, pruned_loss=0.03749, over 2355398.17 frames. ], batch size: 29, lr: 3.40e-03, grad_scale: 8.0 2023-05-17 05:43:59,316 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=288856.0, num_to_drop=1, layers_to_drop={1} 2023-05-17 05:44:00,831 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=288858.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 05:44:02,777 INFO [finetune.py:992] (0/2) Epoch 16, batch 5350, loss[loss=0.1688, simple_loss=0.2629, pruned_loss=0.03734, over 11213.00 frames. ], tot_loss[loss=0.1634, simple_loss=0.2525, pruned_loss=0.03717, over 2361609.33 frames. ], batch size: 55, lr: 3.40e-03, grad_scale: 8.0 2023-05-17 05:44:23,248 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.829e+02 2.696e+02 3.125e+02 3.859e+02 8.981e+02, threshold=6.250e+02, percent-clipped=2.0 2023-05-17 05:44:38,734 INFO [finetune.py:992] (0/2) Epoch 16, batch 5400, loss[loss=0.1661, simple_loss=0.2568, pruned_loss=0.03769, over 12311.00 frames. ], tot_loss[loss=0.1633, simple_loss=0.2525, pruned_loss=0.0371, over 2361325.55 frames. ], batch size: 34, lr: 3.40e-03, grad_scale: 8.0 2023-05-17 05:45:14,481 INFO [finetune.py:992] (0/2) Epoch 16, batch 5450, loss[loss=0.1769, simple_loss=0.2656, pruned_loss=0.04412, over 12307.00 frames. ], tot_loss[loss=0.1638, simple_loss=0.2532, pruned_loss=0.03723, over 2358311.12 frames. ], batch size: 34, lr: 3.40e-03, grad_scale: 8.0 2023-05-17 05:45:17,479 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=288965.0, num_to_drop=1, layers_to_drop={1} 2023-05-17 05:45:30,276 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=288983.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 05:45:35,056 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.974e+02 2.608e+02 3.066e+02 3.634e+02 6.198e+02, threshold=6.131e+02, percent-clipped=0.0 2023-05-17 05:45:43,344 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=289001.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 05:45:50,245 INFO [finetune.py:992] (0/2) Epoch 16, batch 5500, loss[loss=0.1719, simple_loss=0.2711, pruned_loss=0.03635, over 10321.00 frames. ], tot_loss[loss=0.1643, simple_loss=0.254, pruned_loss=0.03731, over 2363625.38 frames. ], batch size: 68, lr: 3.40e-03, grad_scale: 8.0 2023-05-17 05:45:52,492 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.4692, 5.2951, 5.4596, 5.4443, 5.0535, 5.1331, 4.8750, 5.3594], device='cuda:0'), covar=tensor([0.0681, 0.0559, 0.0772, 0.0558, 0.2036, 0.1359, 0.0582, 0.1060], device='cuda:0'), in_proj_covar=tensor([0.0564, 0.0728, 0.0639, 0.0659, 0.0882, 0.0776, 0.0589, 0.0504], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0005, 0.0004, 0.0004, 0.0005, 0.0005, 0.0004, 0.0003], device='cuda:0') 2023-05-17 05:45:53,897 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.8465, 3.8645, 3.3863, 3.2490, 3.0380, 2.9491, 3.8202, 2.5493], device='cuda:0'), covar=tensor([0.0367, 0.0130, 0.0244, 0.0195, 0.0389, 0.0385, 0.0155, 0.0443], device='cuda:0'), in_proj_covar=tensor([0.0201, 0.0169, 0.0173, 0.0196, 0.0207, 0.0207, 0.0180, 0.0211], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-17 05:46:04,640 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=289031.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 05:46:15,354 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=289046.0, num_to_drop=1, layers_to_drop={0} 2023-05-17 05:46:18,324 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=289049.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 05:46:20,440 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=289052.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 05:46:25,501 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.5286, 5.3458, 5.4769, 5.4894, 5.1250, 5.1996, 4.9886, 5.3170], device='cuda:0'), covar=tensor([0.0536, 0.0509, 0.0738, 0.0465, 0.1667, 0.1236, 0.0465, 0.1160], device='cuda:0'), in_proj_covar=tensor([0.0563, 0.0728, 0.0637, 0.0656, 0.0881, 0.0776, 0.0588, 0.0502], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0005, 0.0004, 0.0004, 0.0005, 0.0005, 0.0004, 0.0003], device='cuda:0') 2023-05-17 05:46:26,232 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([6.0442, 5.9797, 5.7531, 5.1393, 5.1882, 5.9142, 5.5453, 5.3142], device='cuda:0'), covar=tensor([0.0741, 0.1070, 0.0773, 0.1557, 0.0695, 0.0633, 0.1459, 0.0921], device='cuda:0'), in_proj_covar=tensor([0.0646, 0.0578, 0.0535, 0.0651, 0.0423, 0.0751, 0.0795, 0.0580], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0004, 0.0004, 0.0003], device='cuda:0') 2023-05-17 05:46:26,825 INFO [finetune.py:992] (0/2) Epoch 16, batch 5550, loss[loss=0.1578, simple_loss=0.2567, pruned_loss=0.02947, over 12139.00 frames. ], tot_loss[loss=0.1638, simple_loss=0.2536, pruned_loss=0.03697, over 2371434.69 frames. ], batch size: 34, lr: 3.40e-03, grad_scale: 8.0 2023-05-17 05:46:48,376 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.813e+02 2.627e+02 3.073e+02 3.584e+02 7.360e+02, threshold=6.145e+02, percent-clipped=6.0 2023-05-17 05:46:55,631 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=289100.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 05:47:00,857 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=289107.0, num_to_drop=1, layers_to_drop={1} 2023-05-17 05:47:03,493 INFO [finetune.py:992] (0/2) Epoch 16, batch 5600, loss[loss=0.1614, simple_loss=0.2574, pruned_loss=0.03269, over 11814.00 frames. ], tot_loss[loss=0.164, simple_loss=0.2539, pruned_loss=0.03703, over 2369829.07 frames. ], batch size: 44, lr: 3.40e-03, grad_scale: 8.0 2023-05-17 05:47:21,419 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=289136.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 05:47:33,666 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=289153.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 05:47:39,347 INFO [finetune.py:992] (0/2) Epoch 16, batch 5650, loss[loss=0.1709, simple_loss=0.2703, pruned_loss=0.03574, over 12282.00 frames. ], tot_loss[loss=0.1643, simple_loss=0.2542, pruned_loss=0.03719, over 2366678.92 frames. ], batch size: 37, lr: 3.40e-03, grad_scale: 8.0 2023-05-17 05:47:41,109 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 2023-05-17 05:47:59,908 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.617e+02 2.624e+02 3.112e+02 3.713e+02 1.702e+03, threshold=6.223e+02, percent-clipped=2.0 2023-05-17 05:48:05,876 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=289197.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 05:48:16,112 INFO [finetune.py:992] (0/2) Epoch 16, batch 5700, loss[loss=0.1364, simple_loss=0.2164, pruned_loss=0.02819, over 12015.00 frames. ], tot_loss[loss=0.1642, simple_loss=0.254, pruned_loss=0.03719, over 2361131.46 frames. ], batch size: 28, lr: 3.40e-03, grad_scale: 8.0 2023-05-17 05:48:31,506 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.3598, 4.7672, 4.1824, 5.0792, 4.5339, 3.2098, 4.3302, 3.3773], device='cuda:0'), covar=tensor([0.0873, 0.0781, 0.1505, 0.0533, 0.1292, 0.1601, 0.1032, 0.3068], device='cuda:0'), in_proj_covar=tensor([0.0316, 0.0385, 0.0367, 0.0334, 0.0378, 0.0279, 0.0355, 0.0375], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-17 05:48:52,408 INFO [finetune.py:992] (0/2) Epoch 16, batch 5750, loss[loss=0.2153, simple_loss=0.3003, pruned_loss=0.06508, over 12131.00 frames. ], tot_loss[loss=0.1637, simple_loss=0.2535, pruned_loss=0.03694, over 2375521.24 frames. ], batch size: 39, lr: 3.40e-03, grad_scale: 8.0 2023-05-17 05:48:55,396 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=289265.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 05:49:13,132 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.000e+02 2.626e+02 2.935e+02 3.698e+02 1.071e+03, threshold=5.870e+02, percent-clipped=1.0 2023-05-17 05:49:16,589 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.39 vs. limit=2.0 2023-05-17 05:49:28,344 INFO [finetune.py:992] (0/2) Epoch 16, batch 5800, loss[loss=0.1819, simple_loss=0.2767, pruned_loss=0.04357, over 10322.00 frames. ], tot_loss[loss=0.164, simple_loss=0.2536, pruned_loss=0.03714, over 2370576.14 frames. ], batch size: 68, lr: 3.40e-03, grad_scale: 8.0 2023-05-17 05:49:29,892 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=289313.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 05:50:01,762 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.1140, 4.8934, 4.8462, 4.9026, 4.5920, 5.0835, 5.0190, 5.2251], device='cuda:0'), covar=tensor([0.0187, 0.0182, 0.0205, 0.0389, 0.0751, 0.0350, 0.0159, 0.0182], device='cuda:0'), in_proj_covar=tensor([0.0205, 0.0205, 0.0201, 0.0258, 0.0249, 0.0231, 0.0185, 0.0242], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0005, 0.0004, 0.0005, 0.0003, 0.0005], device='cuda:0') 2023-05-17 05:50:04,471 INFO [finetune.py:992] (0/2) Epoch 16, batch 5850, loss[loss=0.1685, simple_loss=0.2693, pruned_loss=0.03384, over 11613.00 frames. ], tot_loss[loss=0.1646, simple_loss=0.2544, pruned_loss=0.03737, over 2367021.91 frames. ], batch size: 48, lr: 3.40e-03, grad_scale: 8.0 2023-05-17 05:50:25,899 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.667e+02 2.733e+02 3.224e+02 3.691e+02 7.666e+02, threshold=6.448e+02, percent-clipped=6.0 2023-05-17 05:50:35,029 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=289402.0, num_to_drop=1, layers_to_drop={2} 2023-05-17 05:50:41,318 INFO [finetune.py:992] (0/2) Epoch 16, batch 5900, loss[loss=0.1728, simple_loss=0.2676, pruned_loss=0.03901, over 12013.00 frames. ], tot_loss[loss=0.1647, simple_loss=0.2543, pruned_loss=0.03757, over 2365245.28 frames. ], batch size: 40, lr: 3.40e-03, grad_scale: 8.0 2023-05-17 05:50:58,940 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=3.04 vs. limit=5.0 2023-05-17 05:51:11,562 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=289453.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 05:51:17,115 INFO [finetune.py:992] (0/2) Epoch 16, batch 5950, loss[loss=0.1545, simple_loss=0.2511, pruned_loss=0.02895, over 12410.00 frames. ], tot_loss[loss=0.1642, simple_loss=0.2539, pruned_loss=0.03722, over 2368974.79 frames. ], batch size: 32, lr: 3.40e-03, grad_scale: 8.0 2023-05-17 05:51:36,605 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2023-05-17 05:51:37,579 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.801e+02 2.502e+02 2.971e+02 3.587e+02 5.831e+02, threshold=5.943e+02, percent-clipped=0.0 2023-05-17 05:51:39,868 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=289492.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 05:51:46,188 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=289501.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 05:51:47,742 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=289503.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 05:51:53,293 INFO [finetune.py:992] (0/2) Epoch 16, batch 6000, loss[loss=0.1522, simple_loss=0.235, pruned_loss=0.03472, over 11854.00 frames. ], tot_loss[loss=0.1646, simple_loss=0.2545, pruned_loss=0.03735, over 2370808.97 frames. ], batch size: 26, lr: 3.40e-03, grad_scale: 8.0 2023-05-17 05:51:53,294 INFO [finetune.py:1017] (0/2) Computing validation loss 2023-05-17 05:52:11,668 INFO [finetune.py:1026] (0/2) Epoch 16, validation: loss=0.3117, simple_loss=0.3889, pruned_loss=0.1172, over 1020973.00 frames. 2023-05-17 05:52:11,670 INFO [finetune.py:1027] (0/2) Maximum memory allocated so far is 12827MB 2023-05-17 05:52:12,508 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.3696, 5.1348, 5.3340, 5.3435, 4.9413, 5.0149, 4.7199, 5.2064], device='cuda:0'), covar=tensor([0.0783, 0.0607, 0.0813, 0.0562, 0.2115, 0.1267, 0.0639, 0.1314], device='cuda:0'), in_proj_covar=tensor([0.0567, 0.0729, 0.0638, 0.0656, 0.0881, 0.0777, 0.0590, 0.0505], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0005, 0.0004, 0.0004, 0.0005, 0.0005, 0.0004, 0.0003], device='cuda:0') 2023-05-17 05:52:31,375 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.48 vs. limit=2.0 2023-05-17 05:52:47,252 INFO [finetune.py:992] (0/2) Epoch 16, batch 6050, loss[loss=0.1517, simple_loss=0.241, pruned_loss=0.0312, over 12252.00 frames. ], tot_loss[loss=0.165, simple_loss=0.255, pruned_loss=0.03751, over 2375890.90 frames. ], batch size: 32, lr: 3.40e-03, grad_scale: 8.0 2023-05-17 05:52:49,482 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=289564.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 05:53:07,828 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.885e+02 2.709e+02 3.140e+02 3.726e+02 6.638e+02, threshold=6.280e+02, percent-clipped=1.0 2023-05-17 05:53:24,183 INFO [finetune.py:992] (0/2) Epoch 16, batch 6100, loss[loss=0.1517, simple_loss=0.238, pruned_loss=0.03267, over 12271.00 frames. ], tot_loss[loss=0.1651, simple_loss=0.2547, pruned_loss=0.03773, over 2383327.47 frames. ], batch size: 28, lr: 3.40e-03, grad_scale: 8.0 2023-05-17 05:53:32,185 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.0994, 4.7411, 5.0729, 4.4531, 4.7638, 4.4914, 5.1146, 4.8915], device='cuda:0'), covar=tensor([0.0322, 0.0396, 0.0336, 0.0314, 0.0417, 0.0379, 0.0211, 0.0310], device='cuda:0'), in_proj_covar=tensor([0.0277, 0.0284, 0.0303, 0.0277, 0.0278, 0.0279, 0.0251, 0.0228], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-17 05:53:49,572 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.2497, 3.7758, 4.1534, 4.4010, 2.8135, 3.6862, 2.4693, 3.9383], device='cuda:0'), covar=tensor([0.1574, 0.0858, 0.0786, 0.0614, 0.1272, 0.0703, 0.1982, 0.0992], device='cuda:0'), in_proj_covar=tensor([0.0230, 0.0271, 0.0302, 0.0362, 0.0246, 0.0247, 0.0265, 0.0375], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-17 05:54:00,126 INFO [finetune.py:992] (0/2) Epoch 16, batch 6150, loss[loss=0.1768, simple_loss=0.2697, pruned_loss=0.0419, over 12108.00 frames. ], tot_loss[loss=0.1645, simple_loss=0.2544, pruned_loss=0.03726, over 2384623.17 frames. ], batch size: 42, lr: 3.40e-03, grad_scale: 8.0 2023-05-17 05:54:16,856 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.3775, 4.0759, 4.3353, 4.6890, 3.2600, 3.9900, 2.7763, 4.1876], device='cuda:0'), covar=tensor([0.1589, 0.0748, 0.0759, 0.0575, 0.1112, 0.0633, 0.1684, 0.1207], device='cuda:0'), in_proj_covar=tensor([0.0231, 0.0272, 0.0303, 0.0364, 0.0247, 0.0247, 0.0265, 0.0376], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-17 05:54:18,332 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.0594, 4.4916, 4.0509, 4.8847, 4.3438, 2.8967, 4.1398, 2.9473], device='cuda:0'), covar=tensor([0.0960, 0.0828, 0.1449, 0.0479, 0.1379, 0.1771, 0.1178, 0.3468], device='cuda:0'), in_proj_covar=tensor([0.0314, 0.0384, 0.0363, 0.0334, 0.0376, 0.0277, 0.0353, 0.0373], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-17 05:54:20,931 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.585e+02 2.594e+02 3.012e+02 3.792e+02 6.022e+02, threshold=6.025e+02, percent-clipped=0.0 2023-05-17 05:54:24,159 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.4013, 3.5498, 3.2724, 3.7170, 3.4610, 2.6846, 3.2971, 2.8584], device='cuda:0'), covar=tensor([0.1002, 0.1236, 0.1652, 0.0858, 0.1427, 0.1707, 0.1491, 0.3048], device='cuda:0'), in_proj_covar=tensor([0.0314, 0.0384, 0.0364, 0.0334, 0.0376, 0.0278, 0.0353, 0.0373], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-17 05:54:29,731 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=289702.0, num_to_drop=1, layers_to_drop={0} 2023-05-17 05:54:36,045 INFO [finetune.py:992] (0/2) Epoch 16, batch 6200, loss[loss=0.1672, simple_loss=0.2691, pruned_loss=0.03261, over 10528.00 frames. ], tot_loss[loss=0.1643, simple_loss=0.2543, pruned_loss=0.03718, over 2384047.00 frames. ], batch size: 68, lr: 3.40e-03, grad_scale: 16.0 2023-05-17 05:55:04,770 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=289750.0, num_to_drop=1, layers_to_drop={0} 2023-05-17 05:55:09,211 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.65 vs. limit=2.0 2023-05-17 05:55:12,968 INFO [finetune.py:992] (0/2) Epoch 16, batch 6250, loss[loss=0.17, simple_loss=0.2585, pruned_loss=0.04071, over 12028.00 frames. ], tot_loss[loss=0.1638, simple_loss=0.2538, pruned_loss=0.03686, over 2385353.20 frames. ], batch size: 40, lr: 3.40e-03, grad_scale: 16.0 2023-05-17 05:55:33,876 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.999e+02 2.669e+02 3.033e+02 3.557e+02 7.068e+02, threshold=6.066e+02, percent-clipped=2.0 2023-05-17 05:55:35,481 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=289792.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 05:55:49,224 INFO [finetune.py:992] (0/2) Epoch 16, batch 6300, loss[loss=0.1655, simple_loss=0.2613, pruned_loss=0.03481, over 11871.00 frames. ], tot_loss[loss=0.1644, simple_loss=0.2546, pruned_loss=0.03711, over 2382239.31 frames. ], batch size: 49, lr: 3.39e-03, grad_scale: 16.0 2023-05-17 05:56:10,324 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=289840.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 05:56:20,445 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.2974, 3.6180, 3.8025, 4.1456, 2.8318, 3.5612, 2.6021, 3.6170], device='cuda:0'), covar=tensor([0.1460, 0.0791, 0.0878, 0.0688, 0.1150, 0.0705, 0.1737, 0.0939], device='cuda:0'), in_proj_covar=tensor([0.0231, 0.0271, 0.0303, 0.0363, 0.0246, 0.0247, 0.0265, 0.0375], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-17 05:56:23,887 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=289859.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 05:56:25,164 INFO [finetune.py:992] (0/2) Epoch 16, batch 6350, loss[loss=0.1634, simple_loss=0.2617, pruned_loss=0.03258, over 12355.00 frames. ], tot_loss[loss=0.1648, simple_loss=0.2546, pruned_loss=0.03748, over 2376657.49 frames. ], batch size: 35, lr: 3.39e-03, grad_scale: 16.0 2023-05-17 05:56:26,217 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 2023-05-17 05:56:30,210 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.0032, 4.6721, 4.9760, 4.3461, 4.6564, 4.4900, 5.0100, 4.7186], device='cuda:0'), covar=tensor([0.0311, 0.0406, 0.0358, 0.0319, 0.0491, 0.0332, 0.0245, 0.0431], device='cuda:0'), in_proj_covar=tensor([0.0277, 0.0284, 0.0303, 0.0277, 0.0277, 0.0278, 0.0251, 0.0227], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-17 05:56:46,420 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.888e+02 2.684e+02 3.149e+02 3.789e+02 6.168e+02, threshold=6.298e+02, percent-clipped=1.0 2023-05-17 05:57:02,186 INFO [finetune.py:992] (0/2) Epoch 16, batch 6400, loss[loss=0.1951, simple_loss=0.2821, pruned_loss=0.05407, over 12121.00 frames. ], tot_loss[loss=0.1648, simple_loss=0.2547, pruned_loss=0.03745, over 2381880.75 frames. ], batch size: 39, lr: 3.39e-03, grad_scale: 16.0 2023-05-17 05:57:08,647 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=289920.0, num_to_drop=1, layers_to_drop={1} 2023-05-17 05:57:37,813 INFO [finetune.py:992] (0/2) Epoch 16, batch 6450, loss[loss=0.1594, simple_loss=0.2522, pruned_loss=0.03331, over 12147.00 frames. ], tot_loss[loss=0.1657, simple_loss=0.2553, pruned_loss=0.03804, over 2368884.86 frames. ], batch size: 34, lr: 3.39e-03, grad_scale: 16.0 2023-05-17 05:57:38,083 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.4448, 4.2807, 4.3836, 4.6264, 3.2059, 4.0523, 3.0872, 4.3005], device='cuda:0'), covar=tensor([0.1456, 0.0561, 0.0786, 0.0489, 0.1132, 0.0586, 0.1523, 0.1078], device='cuda:0'), in_proj_covar=tensor([0.0230, 0.0270, 0.0302, 0.0363, 0.0246, 0.0247, 0.0265, 0.0376], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-17 05:57:42,965 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.9354, 5.8990, 5.7227, 5.1902, 5.2362, 5.8298, 5.4316, 5.2623], device='cuda:0'), covar=tensor([0.0766, 0.1104, 0.0689, 0.1589, 0.0757, 0.0716, 0.1492, 0.1011], device='cuda:0'), in_proj_covar=tensor([0.0646, 0.0584, 0.0538, 0.0658, 0.0425, 0.0758, 0.0804, 0.0585], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0004, 0.0004, 0.0003], device='cuda:0') 2023-05-17 05:57:52,392 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=289981.0, num_to_drop=1, layers_to_drop={0} 2023-05-17 05:57:58,797 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.900e+02 2.638e+02 3.095e+02 3.629e+02 6.932e+02, threshold=6.190e+02, percent-clipped=4.0 2023-05-17 05:58:04,709 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=289998.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 05:58:06,192 INFO [checkpoint.py:75] (0/2) Saving checkpoint to pruned_transducer_stateless7/exp_giga_finetune/checkpoint-190000.pt 2023-05-17 05:58:17,163 INFO [finetune.py:992] (0/2) Epoch 16, batch 6500, loss[loss=0.1871, simple_loss=0.2926, pruned_loss=0.04075, over 12346.00 frames. ], tot_loss[loss=0.1658, simple_loss=0.2553, pruned_loss=0.03809, over 2369264.04 frames. ], batch size: 36, lr: 3.39e-03, grad_scale: 16.0 2023-05-17 05:58:45,912 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=290050.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 05:58:52,833 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=290059.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 05:58:53,999 INFO [finetune.py:992] (0/2) Epoch 16, batch 6550, loss[loss=0.1585, simple_loss=0.2553, pruned_loss=0.03083, over 12294.00 frames. ], tot_loss[loss=0.165, simple_loss=0.2546, pruned_loss=0.03771, over 2369340.49 frames. ], batch size: 33, lr: 3.39e-03, grad_scale: 16.0 2023-05-17 05:59:05,616 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.2420, 4.9252, 5.2371, 4.5957, 4.8742, 4.6724, 5.2369, 4.9938], device='cuda:0'), covar=tensor([0.0267, 0.0349, 0.0273, 0.0279, 0.0439, 0.0329, 0.0209, 0.0361], device='cuda:0'), in_proj_covar=tensor([0.0276, 0.0283, 0.0304, 0.0276, 0.0277, 0.0278, 0.0251, 0.0226], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-17 05:59:14,847 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.948e+02 2.611e+02 2.892e+02 3.613e+02 1.747e+03, threshold=5.784e+02, percent-clipped=1.0 2023-05-17 05:59:29,839 INFO [finetune.py:992] (0/2) Epoch 16, batch 6600, loss[loss=0.1873, simple_loss=0.286, pruned_loss=0.04426, over 12355.00 frames. ], tot_loss[loss=0.1657, simple_loss=0.2555, pruned_loss=0.03794, over 2369642.48 frames. ], batch size: 38, lr: 3.39e-03, grad_scale: 16.0 2023-05-17 05:59:30,085 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=290111.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 05:59:53,020 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.6113, 3.7345, 3.3548, 3.2340, 2.9805, 2.8659, 3.7005, 2.4356], device='cuda:0'), covar=tensor([0.0400, 0.0121, 0.0182, 0.0201, 0.0349, 0.0369, 0.0114, 0.0470], device='cuda:0'), in_proj_covar=tensor([0.0199, 0.0168, 0.0173, 0.0195, 0.0205, 0.0205, 0.0178, 0.0211], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-17 06:00:04,597 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=290159.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 06:00:05,882 INFO [finetune.py:992] (0/2) Epoch 16, batch 6650, loss[loss=0.1846, simple_loss=0.2717, pruned_loss=0.04873, over 10806.00 frames. ], tot_loss[loss=0.1654, simple_loss=0.2552, pruned_loss=0.03779, over 2371087.92 frames. ], batch size: 68, lr: 3.39e-03, grad_scale: 16.0 2023-05-17 06:00:27,286 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.015e+02 2.630e+02 3.118e+02 3.870e+02 7.746e+02, threshold=6.235e+02, percent-clipped=1.0 2023-05-17 06:00:40,570 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=290207.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 06:00:43,345 INFO [finetune.py:992] (0/2) Epoch 16, batch 6700, loss[loss=0.1789, simple_loss=0.259, pruned_loss=0.04941, over 12350.00 frames. ], tot_loss[loss=0.1648, simple_loss=0.255, pruned_loss=0.03736, over 2372702.12 frames. ], batch size: 31, lr: 3.39e-03, grad_scale: 16.0 2023-05-17 06:00:57,033 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.4191, 2.5025, 3.0720, 4.2192, 2.3662, 4.2430, 4.3408, 4.4142], device='cuda:0'), covar=tensor([0.0166, 0.1377, 0.0609, 0.0238, 0.1390, 0.0345, 0.0210, 0.0151], device='cuda:0'), in_proj_covar=tensor([0.0128, 0.0211, 0.0189, 0.0127, 0.0195, 0.0189, 0.0183, 0.0131], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002], device='cuda:0') 2023-05-17 06:01:18,671 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.2718, 4.6620, 2.8878, 2.6757, 3.9948, 2.5623, 3.9379, 3.0881], device='cuda:0'), covar=tensor([0.0747, 0.0525, 0.1129, 0.1510, 0.0299, 0.1377, 0.0495, 0.0884], device='cuda:0'), in_proj_covar=tensor([0.0193, 0.0263, 0.0179, 0.0207, 0.0146, 0.0186, 0.0202, 0.0177], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-17 06:01:19,142 INFO [finetune.py:992] (0/2) Epoch 16, batch 6750, loss[loss=0.177, simple_loss=0.2653, pruned_loss=0.04438, over 12057.00 frames. ], tot_loss[loss=0.1648, simple_loss=0.2546, pruned_loss=0.03754, over 2369436.41 frames. ], batch size: 37, lr: 3.39e-03, grad_scale: 16.0 2023-05-17 06:01:21,520 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.2444, 3.7045, 3.3230, 3.1771, 2.9218, 2.8600, 3.6286, 2.1655], device='cuda:0'), covar=tensor([0.0565, 0.0126, 0.0202, 0.0209, 0.0411, 0.0391, 0.0154, 0.0643], device='cuda:0'), in_proj_covar=tensor([0.0200, 0.0168, 0.0172, 0.0195, 0.0205, 0.0206, 0.0180, 0.0210], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-17 06:01:29,918 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=290276.0, num_to_drop=1, layers_to_drop={1} 2023-05-17 06:01:39,841 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.771e+02 2.624e+02 3.096e+02 3.902e+02 1.134e+03, threshold=6.192e+02, percent-clipped=3.0 2023-05-17 06:01:55,100 INFO [finetune.py:992] (0/2) Epoch 16, batch 6800, loss[loss=0.1494, simple_loss=0.2323, pruned_loss=0.03321, over 12256.00 frames. ], tot_loss[loss=0.1648, simple_loss=0.2546, pruned_loss=0.0375, over 2374025.87 frames. ], batch size: 32, lr: 3.39e-03, grad_scale: 16.0 2023-05-17 06:02:00,188 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=290318.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 06:02:26,150 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=290354.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 06:02:26,932 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.6886, 2.7958, 3.2552, 4.5084, 2.5329, 4.4383, 4.6305, 4.6735], device='cuda:0'), covar=tensor([0.0151, 0.1212, 0.0540, 0.0216, 0.1286, 0.0331, 0.0154, 0.0144], device='cuda:0'), in_proj_covar=tensor([0.0128, 0.0212, 0.0188, 0.0127, 0.0195, 0.0189, 0.0184, 0.0131], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002], device='cuda:0') 2023-05-17 06:02:30,947 INFO [finetune.py:992] (0/2) Epoch 16, batch 6850, loss[loss=0.2108, simple_loss=0.2897, pruned_loss=0.06591, over 8330.00 frames. ], tot_loss[loss=0.1658, simple_loss=0.2558, pruned_loss=0.03788, over 2375958.96 frames. ], batch size: 97, lr: 3.39e-03, grad_scale: 16.0 2023-05-17 06:02:34,084 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.5611, 2.9650, 3.7605, 4.6222, 3.9551, 4.6156, 4.0398, 3.2302], device='cuda:0'), covar=tensor([0.0056, 0.0364, 0.0140, 0.0039, 0.0128, 0.0086, 0.0124, 0.0385], device='cuda:0'), in_proj_covar=tensor([0.0091, 0.0124, 0.0106, 0.0080, 0.0106, 0.0117, 0.0101, 0.0141], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-17 06:02:43,631 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.88 vs. limit=2.0 2023-05-17 06:02:44,017 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=290379.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 06:02:51,661 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.889e+02 2.699e+02 3.058e+02 3.912e+02 8.049e+02, threshold=6.116e+02, percent-clipped=2.0 2023-05-17 06:03:03,386 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=290406.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 06:03:06,722 INFO [finetune.py:992] (0/2) Epoch 16, batch 6900, loss[loss=0.1686, simple_loss=0.2578, pruned_loss=0.0397, over 11194.00 frames. ], tot_loss[loss=0.166, simple_loss=0.2558, pruned_loss=0.0381, over 2368845.84 frames. ], batch size: 55, lr: 3.39e-03, grad_scale: 16.0 2023-05-17 06:03:16,557 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.36 vs. limit=2.0 2023-05-17 06:03:28,204 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.85 vs. limit=2.0 2023-05-17 06:03:29,263 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.3366, 3.2214, 3.2994, 3.4683, 2.7656, 3.2126, 2.6976, 2.9991], device='cuda:0'), covar=tensor([0.1397, 0.0743, 0.0805, 0.0542, 0.0917, 0.0683, 0.1453, 0.0954], device='cuda:0'), in_proj_covar=tensor([0.0229, 0.0270, 0.0303, 0.0362, 0.0246, 0.0247, 0.0264, 0.0375], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-17 06:03:43,375 INFO [finetune.py:992] (0/2) Epoch 16, batch 6950, loss[loss=0.2123, simple_loss=0.2929, pruned_loss=0.06588, over 7982.00 frames. ], tot_loss[loss=0.1653, simple_loss=0.255, pruned_loss=0.03783, over 2359177.24 frames. ], batch size: 98, lr: 3.39e-03, grad_scale: 16.0 2023-05-17 06:04:04,627 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.896e+02 2.638e+02 3.087e+02 3.748e+02 8.122e+02, threshold=6.174e+02, percent-clipped=3.0 2023-05-17 06:04:19,807 INFO [finetune.py:992] (0/2) Epoch 16, batch 7000, loss[loss=0.1591, simple_loss=0.2467, pruned_loss=0.03575, over 12143.00 frames. ], tot_loss[loss=0.1649, simple_loss=0.2546, pruned_loss=0.03759, over 2363273.33 frames. ], batch size: 38, lr: 3.39e-03, grad_scale: 16.0 2023-05-17 06:04:38,000 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([6.2049, 6.1058, 5.8833, 5.4065, 5.3163, 6.0640, 5.7668, 5.4533], device='cuda:0'), covar=tensor([0.0601, 0.0917, 0.0650, 0.1632, 0.0677, 0.0692, 0.1568, 0.1058], device='cuda:0'), in_proj_covar=tensor([0.0639, 0.0578, 0.0532, 0.0652, 0.0424, 0.0747, 0.0792, 0.0580], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-05-17 06:04:56,226 INFO [finetune.py:992] (0/2) Epoch 16, batch 7050, loss[loss=0.1469, simple_loss=0.242, pruned_loss=0.02587, over 12368.00 frames. ], tot_loss[loss=0.1641, simple_loss=0.2534, pruned_loss=0.03738, over 2367354.74 frames. ], batch size: 35, lr: 3.39e-03, grad_scale: 16.0 2023-05-17 06:05:07,169 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=290576.0, num_to_drop=1, layers_to_drop={2} 2023-05-17 06:05:12,496 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 2023-05-17 06:05:16,972 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.957e+02 2.502e+02 2.920e+02 3.536e+02 8.522e+02, threshold=5.839e+02, percent-clipped=1.0 2023-05-17 06:05:26,760 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.2963, 3.5514, 3.1570, 3.6414, 3.4012, 2.6191, 3.2099, 2.7770], device='cuda:0'), covar=tensor([0.1085, 0.1030, 0.1773, 0.0893, 0.1468, 0.1794, 0.1460, 0.3097], device='cuda:0'), in_proj_covar=tensor([0.0313, 0.0382, 0.0364, 0.0333, 0.0375, 0.0277, 0.0352, 0.0371], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-17 06:05:29,515 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.9416, 5.7213, 5.4510, 5.2402, 5.8547, 5.0569, 5.3110, 5.2840], device='cuda:0'), covar=tensor([0.1583, 0.1094, 0.1168, 0.1901, 0.0972, 0.2368, 0.1859, 0.1153], device='cuda:0'), in_proj_covar=tensor([0.0359, 0.0497, 0.0398, 0.0448, 0.0468, 0.0429, 0.0396, 0.0389], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-17 06:05:32,974 INFO [finetune.py:992] (0/2) Epoch 16, batch 7100, loss[loss=0.1378, simple_loss=0.2204, pruned_loss=0.02754, over 12009.00 frames. ], tot_loss[loss=0.1651, simple_loss=0.2547, pruned_loss=0.03778, over 2369032.40 frames. ], batch size: 28, lr: 3.39e-03, grad_scale: 16.0 2023-05-17 06:05:42,871 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=290624.0, num_to_drop=1, layers_to_drop={0} 2023-05-17 06:06:03,926 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=290654.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 06:06:08,614 INFO [finetune.py:992] (0/2) Epoch 16, batch 7150, loss[loss=0.147, simple_loss=0.2367, pruned_loss=0.02864, over 12181.00 frames. ], tot_loss[loss=0.1659, simple_loss=0.2556, pruned_loss=0.03814, over 2371826.73 frames. ], batch size: 31, lr: 3.39e-03, grad_scale: 16.0 2023-05-17 06:06:18,123 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=290674.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 06:06:29,243 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.735e+02 2.710e+02 3.160e+02 3.653e+02 5.571e+02, threshold=6.320e+02, percent-clipped=0.0 2023-05-17 06:06:37,799 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=290702.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 06:06:40,751 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=290706.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 06:06:44,189 INFO [finetune.py:992] (0/2) Epoch 16, batch 7200, loss[loss=0.1597, simple_loss=0.2509, pruned_loss=0.03424, over 11844.00 frames. ], tot_loss[loss=0.1671, simple_loss=0.2568, pruned_loss=0.03867, over 2364202.15 frames. ], batch size: 44, lr: 3.39e-03, grad_scale: 16.0 2023-05-17 06:07:15,622 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=290754.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 06:07:17,934 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.1804, 4.3825, 2.7503, 2.4878, 3.8029, 2.5138, 3.8097, 2.9559], device='cuda:0'), covar=tensor([0.0725, 0.0544, 0.1149, 0.1623, 0.0294, 0.1371, 0.0477, 0.0893], device='cuda:0'), in_proj_covar=tensor([0.0191, 0.0262, 0.0178, 0.0203, 0.0144, 0.0183, 0.0199, 0.0175], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-17 06:07:18,663 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.6324, 2.8119, 4.3848, 4.4226, 2.7599, 2.5128, 2.8915, 2.1170], device='cuda:0'), covar=tensor([0.1659, 0.3125, 0.0512, 0.0523, 0.1487, 0.2628, 0.2916, 0.4143], device='cuda:0'), in_proj_covar=tensor([0.0304, 0.0397, 0.0280, 0.0306, 0.0280, 0.0321, 0.0398, 0.0381], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-17 06:07:20,455 INFO [finetune.py:992] (0/2) Epoch 16, batch 7250, loss[loss=0.164, simple_loss=0.2573, pruned_loss=0.03542, over 12114.00 frames. ], tot_loss[loss=0.1668, simple_loss=0.2565, pruned_loss=0.03853, over 2366221.29 frames. ], batch size: 33, lr: 3.39e-03, grad_scale: 16.0 2023-05-17 06:07:41,804 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.543e+02 2.496e+02 3.019e+02 3.951e+02 6.788e+02, threshold=6.038e+02, percent-clipped=2.0 2023-05-17 06:07:57,144 INFO [finetune.py:992] (0/2) Epoch 16, batch 7300, loss[loss=0.1644, simple_loss=0.2484, pruned_loss=0.04015, over 12179.00 frames. ], tot_loss[loss=0.1662, simple_loss=0.2558, pruned_loss=0.03824, over 2367840.99 frames. ], batch size: 31, lr: 3.39e-03, grad_scale: 16.0 2023-05-17 06:08:32,820 INFO [finetune.py:992] (0/2) Epoch 16, batch 7350, loss[loss=0.1773, simple_loss=0.2704, pruned_loss=0.0421, over 12153.00 frames. ], tot_loss[loss=0.1658, simple_loss=0.2555, pruned_loss=0.03807, over 2372196.13 frames. ], batch size: 34, lr: 3.39e-03, grad_scale: 16.0 2023-05-17 06:08:44,978 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=4.66 vs. limit=5.0 2023-05-17 06:08:54,039 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.696e+02 2.777e+02 3.176e+02 3.962e+02 8.033e+02, threshold=6.352e+02, percent-clipped=3.0 2023-05-17 06:09:01,824 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.36 vs. limit=2.0 2023-05-17 06:09:09,094 INFO [finetune.py:992] (0/2) Epoch 16, batch 7400, loss[loss=0.1822, simple_loss=0.2673, pruned_loss=0.04858, over 10572.00 frames. ], tot_loss[loss=0.1668, simple_loss=0.2566, pruned_loss=0.03851, over 2368289.81 frames. ], batch size: 68, lr: 3.39e-03, grad_scale: 16.0 2023-05-17 06:09:45,437 INFO [finetune.py:992] (0/2) Epoch 16, batch 7450, loss[loss=0.1759, simple_loss=0.2644, pruned_loss=0.04372, over 12053.00 frames. ], tot_loss[loss=0.167, simple_loss=0.2565, pruned_loss=0.03872, over 2366060.04 frames. ], batch size: 42, lr: 3.39e-03, grad_scale: 16.0 2023-05-17 06:09:54,489 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=290974.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 06:10:05,518 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.836e+02 2.703e+02 3.035e+02 3.836e+02 7.685e+02, threshold=6.070e+02, percent-clipped=2.0 2023-05-17 06:10:14,910 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.5644, 2.8202, 3.2517, 4.3768, 2.3246, 4.3435, 4.5678, 4.5961], device='cuda:0'), covar=tensor([0.0144, 0.1201, 0.0529, 0.0176, 0.1481, 0.0333, 0.0154, 0.0126], device='cuda:0'), in_proj_covar=tensor([0.0129, 0.0212, 0.0189, 0.0128, 0.0196, 0.0188, 0.0184, 0.0131], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002], device='cuda:0') 2023-05-17 06:10:20,232 INFO [finetune.py:992] (0/2) Epoch 16, batch 7500, loss[loss=0.1495, simple_loss=0.2402, pruned_loss=0.02942, over 12283.00 frames. ], tot_loss[loss=0.1676, simple_loss=0.2571, pruned_loss=0.03905, over 2371065.17 frames. ], batch size: 33, lr: 3.39e-03, grad_scale: 16.0 2023-05-17 06:10:28,228 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=291022.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 06:10:34,209 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.4918, 4.8685, 3.2192, 2.6629, 4.1950, 2.8262, 4.1780, 3.4574], device='cuda:0'), covar=tensor([0.0709, 0.0636, 0.1001, 0.1641, 0.0270, 0.1225, 0.0496, 0.0735], device='cuda:0'), in_proj_covar=tensor([0.0193, 0.0266, 0.0180, 0.0206, 0.0146, 0.0186, 0.0202, 0.0177], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-17 06:10:57,127 INFO [finetune.py:992] (0/2) Epoch 16, batch 7550, loss[loss=0.1792, simple_loss=0.2794, pruned_loss=0.03948, over 11992.00 frames. ], tot_loss[loss=0.1677, simple_loss=0.2572, pruned_loss=0.03907, over 2365998.86 frames. ], batch size: 42, lr: 3.39e-03, grad_scale: 16.0 2023-05-17 06:11:02,882 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=291069.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 06:11:17,746 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.887e+02 2.906e+02 3.450e+02 4.180e+02 9.677e+02, threshold=6.900e+02, percent-clipped=7.0 2023-05-17 06:11:32,980 INFO [finetune.py:992] (0/2) Epoch 16, batch 7600, loss[loss=0.1736, simple_loss=0.2716, pruned_loss=0.03776, over 12113.00 frames. ], tot_loss[loss=0.1672, simple_loss=0.2566, pruned_loss=0.0389, over 2361680.63 frames. ], batch size: 38, lr: 3.39e-03, grad_scale: 16.0 2023-05-17 06:11:46,602 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=291130.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 06:12:03,060 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=2.03 vs. limit=2.0 2023-05-17 06:12:08,459 INFO [finetune.py:992] (0/2) Epoch 16, batch 7650, loss[loss=0.1583, simple_loss=0.2534, pruned_loss=0.03155, over 12332.00 frames. ], tot_loss[loss=0.1678, simple_loss=0.2569, pruned_loss=0.03932, over 2349125.47 frames. ], batch size: 36, lr: 3.39e-03, grad_scale: 16.0 2023-05-17 06:12:30,421 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.929e+02 2.451e+02 2.997e+02 3.576e+02 6.341e+02, threshold=5.994e+02, percent-clipped=0.0 2023-05-17 06:12:45,707 INFO [finetune.py:992] (0/2) Epoch 16, batch 7700, loss[loss=0.1656, simple_loss=0.2559, pruned_loss=0.03769, over 12189.00 frames. ], tot_loss[loss=0.1679, simple_loss=0.2573, pruned_loss=0.03924, over 2349604.98 frames. ], batch size: 35, lr: 3.38e-03, grad_scale: 16.0 2023-05-17 06:13:21,120 INFO [finetune.py:992] (0/2) Epoch 16, batch 7750, loss[loss=0.1718, simple_loss=0.2634, pruned_loss=0.04012, over 12113.00 frames. ], tot_loss[loss=0.1664, simple_loss=0.2557, pruned_loss=0.03857, over 2358343.24 frames. ], batch size: 38, lr: 3.38e-03, grad_scale: 16.0 2023-05-17 06:13:41,738 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.245e+02 2.900e+02 3.369e+02 4.087e+02 1.776e+03, threshold=6.739e+02, percent-clipped=5.0 2023-05-17 06:13:57,204 INFO [finetune.py:992] (0/2) Epoch 16, batch 7800, loss[loss=0.1671, simple_loss=0.2631, pruned_loss=0.03552, over 12192.00 frames. ], tot_loss[loss=0.1673, simple_loss=0.2566, pruned_loss=0.03903, over 2355685.90 frames. ], batch size: 35, lr: 3.38e-03, grad_scale: 16.0 2023-05-17 06:14:04,282 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.4405, 2.4041, 3.0116, 4.2767, 2.1685, 4.2757, 4.4489, 4.4763], device='cuda:0'), covar=tensor([0.0150, 0.1494, 0.0641, 0.0172, 0.1483, 0.0309, 0.0183, 0.0123], device='cuda:0'), in_proj_covar=tensor([0.0127, 0.0209, 0.0187, 0.0126, 0.0192, 0.0185, 0.0182, 0.0129], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-17 06:14:14,918 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 2023-05-17 06:14:23,146 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.2841, 4.2545, 4.1088, 4.4325, 3.2000, 4.0977, 2.7121, 4.2246], device='cuda:0'), covar=tensor([0.1638, 0.0622, 0.0862, 0.0779, 0.1124, 0.0549, 0.1807, 0.1095], device='cuda:0'), in_proj_covar=tensor([0.0229, 0.0269, 0.0300, 0.0362, 0.0245, 0.0246, 0.0263, 0.0374], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-17 06:14:26,048 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.70 vs. limit=2.0 2023-05-17 06:14:32,960 INFO [finetune.py:992] (0/2) Epoch 16, batch 7850, loss[loss=0.1541, simple_loss=0.2458, pruned_loss=0.03117, over 12032.00 frames. ], tot_loss[loss=0.1672, simple_loss=0.2565, pruned_loss=0.03893, over 2364795.80 frames. ], batch size: 31, lr: 3.38e-03, grad_scale: 16.0 2023-05-17 06:14:35,224 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.2555, 4.1251, 4.0381, 4.4091, 3.1907, 4.0361, 2.5034, 4.1868], device='cuda:0'), covar=tensor([0.1644, 0.0683, 0.1004, 0.0592, 0.1161, 0.0576, 0.1983, 0.1209], device='cuda:0'), in_proj_covar=tensor([0.0229, 0.0270, 0.0301, 0.0362, 0.0246, 0.0246, 0.0264, 0.0374], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-17 06:14:53,387 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.955e+02 2.682e+02 3.106e+02 4.145e+02 1.209e+03, threshold=6.213e+02, percent-clipped=5.0 2023-05-17 06:15:08,702 INFO [finetune.py:992] (0/2) Epoch 16, batch 7900, loss[loss=0.166, simple_loss=0.2594, pruned_loss=0.0363, over 12196.00 frames. ], tot_loss[loss=0.1673, simple_loss=0.2565, pruned_loss=0.03906, over 2361987.52 frames. ], batch size: 35, lr: 3.38e-03, grad_scale: 16.0 2023-05-17 06:15:18,931 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=291425.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 06:15:19,066 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.4120, 4.7581, 3.1050, 2.4754, 4.1736, 2.4946, 4.0969, 3.1589], device='cuda:0'), covar=tensor([0.0738, 0.0611, 0.1062, 0.1759, 0.0335, 0.1558, 0.0487, 0.0896], device='cuda:0'), in_proj_covar=tensor([0.0193, 0.0264, 0.0180, 0.0206, 0.0144, 0.0184, 0.0202, 0.0176], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-17 06:15:45,453 INFO [finetune.py:992] (0/2) Epoch 16, batch 7950, loss[loss=0.1453, simple_loss=0.2296, pruned_loss=0.03052, over 12259.00 frames. ], tot_loss[loss=0.168, simple_loss=0.2571, pruned_loss=0.03944, over 2356557.87 frames. ], batch size: 28, lr: 3.38e-03, grad_scale: 16.0 2023-05-17 06:15:50,081 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 2023-05-17 06:16:06,549 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.010e+02 2.592e+02 3.141e+02 3.877e+02 1.057e+03, threshold=6.283e+02, percent-clipped=3.0 2023-05-17 06:16:20,623 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=3.39 vs. limit=5.0 2023-05-17 06:16:21,529 INFO [finetune.py:992] (0/2) Epoch 16, batch 8000, loss[loss=0.1607, simple_loss=0.2551, pruned_loss=0.03314, over 12136.00 frames. ], tot_loss[loss=0.168, simple_loss=0.2572, pruned_loss=0.03937, over 2354915.79 frames. ], batch size: 36, lr: 3.38e-03, grad_scale: 16.0 2023-05-17 06:16:53,155 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.3616, 3.5107, 3.2281, 3.0900, 2.8725, 2.6510, 3.5650, 2.3932], device='cuda:0'), covar=tensor([0.0526, 0.0144, 0.0230, 0.0243, 0.0411, 0.0420, 0.0152, 0.0527], device='cuda:0'), in_proj_covar=tensor([0.0198, 0.0166, 0.0171, 0.0193, 0.0203, 0.0202, 0.0177, 0.0206], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-17 06:16:57,151 INFO [finetune.py:992] (0/2) Epoch 16, batch 8050, loss[loss=0.193, simple_loss=0.2798, pruned_loss=0.05311, over 12093.00 frames. ], tot_loss[loss=0.1677, simple_loss=0.257, pruned_loss=0.03917, over 2360620.26 frames. ], batch size: 40, lr: 3.38e-03, grad_scale: 16.0 2023-05-17 06:17:17,794 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.054e+02 2.728e+02 3.105e+02 3.762e+02 7.004e+02, threshold=6.210e+02, percent-clipped=2.0 2023-05-17 06:17:26,664 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.74 vs. limit=2.0 2023-05-17 06:17:33,163 INFO [finetune.py:992] (0/2) Epoch 16, batch 8100, loss[loss=0.1683, simple_loss=0.265, pruned_loss=0.03576, over 12327.00 frames. ], tot_loss[loss=0.1662, simple_loss=0.2556, pruned_loss=0.03843, over 2373862.58 frames. ], batch size: 34, lr: 3.38e-03, grad_scale: 16.0 2023-05-17 06:18:09,656 INFO [finetune.py:992] (0/2) Epoch 16, batch 8150, loss[loss=0.1634, simple_loss=0.2452, pruned_loss=0.04075, over 12322.00 frames. ], tot_loss[loss=0.1659, simple_loss=0.2552, pruned_loss=0.03826, over 2377112.33 frames. ], batch size: 31, lr: 3.38e-03, grad_scale: 8.0 2023-05-17 06:18:30,834 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.012e+02 2.696e+02 3.139e+02 3.881e+02 6.152e+02, threshold=6.277e+02, percent-clipped=0.0 2023-05-17 06:18:32,762 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.59 vs. limit=2.0 2023-05-17 06:18:35,376 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.2241, 4.1507, 3.9628, 4.3420, 2.9739, 3.9668, 2.5223, 4.0720], device='cuda:0'), covar=tensor([0.1719, 0.0677, 0.0980, 0.0725, 0.1264, 0.0595, 0.1965, 0.1159], device='cuda:0'), in_proj_covar=tensor([0.0231, 0.0272, 0.0304, 0.0365, 0.0247, 0.0247, 0.0266, 0.0377], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-17 06:18:45,008 INFO [finetune.py:992] (0/2) Epoch 16, batch 8200, loss[loss=0.1765, simple_loss=0.2664, pruned_loss=0.04331, over 12357.00 frames. ], tot_loss[loss=0.1669, simple_loss=0.2566, pruned_loss=0.03858, over 2380242.00 frames. ], batch size: 38, lr: 3.38e-03, grad_scale: 8.0 2023-05-17 06:18:55,054 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=291725.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 06:19:20,304 INFO [finetune.py:992] (0/2) Epoch 16, batch 8250, loss[loss=0.1583, simple_loss=0.2544, pruned_loss=0.03104, over 12160.00 frames. ], tot_loss[loss=0.166, simple_loss=0.2558, pruned_loss=0.03808, over 2386309.51 frames. ], batch size: 34, lr: 3.38e-03, grad_scale: 8.0 2023-05-17 06:19:29,621 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=291773.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 06:19:31,870 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=291776.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 06:19:43,006 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.576e+02 2.597e+02 2.979e+02 3.573e+02 8.446e+02, threshold=5.958e+02, percent-clipped=2.0 2023-05-17 06:19:48,486 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 2023-05-17 06:19:54,304 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.5387, 4.5167, 4.2793, 4.6383, 3.4977, 4.2243, 2.7022, 4.3941], device='cuda:0'), covar=tensor([0.1409, 0.0541, 0.0868, 0.0598, 0.0985, 0.0536, 0.1738, 0.1088], device='cuda:0'), in_proj_covar=tensor([0.0231, 0.0271, 0.0303, 0.0365, 0.0247, 0.0247, 0.0266, 0.0376], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-17 06:19:57,159 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.2057, 2.9428, 2.7616, 2.7750, 2.4267, 2.3998, 2.7805, 2.0351], device='cuda:0'), covar=tensor([0.0419, 0.0160, 0.0228, 0.0222, 0.0450, 0.0318, 0.0222, 0.0500], device='cuda:0'), in_proj_covar=tensor([0.0200, 0.0167, 0.0173, 0.0195, 0.0206, 0.0203, 0.0178, 0.0209], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-17 06:19:57,576 INFO [finetune.py:992] (0/2) Epoch 16, batch 8300, loss[loss=0.214, simple_loss=0.2911, pruned_loss=0.0684, over 10497.00 frames. ], tot_loss[loss=0.1648, simple_loss=0.2547, pruned_loss=0.03743, over 2388387.00 frames. ], batch size: 68, lr: 3.38e-03, grad_scale: 8.0 2023-05-17 06:20:10,463 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.2900, 4.6797, 2.9491, 2.4477, 4.0455, 2.3768, 3.9633, 2.9877], device='cuda:0'), covar=tensor([0.0656, 0.0429, 0.1083, 0.1597, 0.0275, 0.1504, 0.0431, 0.0911], device='cuda:0'), in_proj_covar=tensor([0.0192, 0.0262, 0.0179, 0.0204, 0.0145, 0.0183, 0.0201, 0.0175], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-17 06:20:16,134 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=291837.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 06:20:33,124 INFO [finetune.py:992] (0/2) Epoch 16, batch 8350, loss[loss=0.1733, simple_loss=0.2435, pruned_loss=0.05152, over 12174.00 frames. ], tot_loss[loss=0.1648, simple_loss=0.2545, pruned_loss=0.0375, over 2387453.78 frames. ], batch size: 29, lr: 3.38e-03, grad_scale: 8.0 2023-05-17 06:20:54,476 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.516e+02 2.600e+02 3.057e+02 3.626e+02 1.055e+03, threshold=6.113e+02, percent-clipped=3.0 2023-05-17 06:21:08,756 INFO [finetune.py:992] (0/2) Epoch 16, batch 8400, loss[loss=0.1618, simple_loss=0.2405, pruned_loss=0.04161, over 12279.00 frames. ], tot_loss[loss=0.1668, simple_loss=0.2562, pruned_loss=0.03873, over 2366484.43 frames. ], batch size: 28, lr: 3.38e-03, grad_scale: 8.0 2023-05-17 06:21:23,564 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.32 vs. limit=2.0 2023-05-17 06:21:25,500 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.62 vs. limit=2.0 2023-05-17 06:21:37,607 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 2023-05-17 06:21:45,799 INFO [finetune.py:992] (0/2) Epoch 16, batch 8450, loss[loss=0.1707, simple_loss=0.2616, pruned_loss=0.03989, over 11202.00 frames. ], tot_loss[loss=0.1664, simple_loss=0.2557, pruned_loss=0.03851, over 2368304.03 frames. ], batch size: 55, lr: 3.38e-03, grad_scale: 8.0 2023-05-17 06:22:06,990 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.918e+02 2.756e+02 3.165e+02 3.955e+02 9.030e+02, threshold=6.330e+02, percent-clipped=2.0 2023-05-17 06:22:13,793 INFO [checkpoint.py:75] (0/2) Saving checkpoint to pruned_transducer_stateless7/exp_giga_finetune/checkpoint-192000.pt 2023-05-17 06:22:24,602 INFO [finetune.py:992] (0/2) Epoch 16, batch 8500, loss[loss=0.1549, simple_loss=0.2449, pruned_loss=0.03246, over 11191.00 frames. ], tot_loss[loss=0.1661, simple_loss=0.2554, pruned_loss=0.03839, over 2367990.39 frames. ], batch size: 55, lr: 3.38e-03, grad_scale: 8.0 2023-05-17 06:22:52,853 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.79 vs. limit=2.0 2023-05-17 06:22:56,155 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=292054.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 06:23:01,025 INFO [finetune.py:992] (0/2) Epoch 16, batch 8550, loss[loss=0.2025, simple_loss=0.2871, pruned_loss=0.05898, over 8450.00 frames. ], tot_loss[loss=0.1657, simple_loss=0.2554, pruned_loss=0.03799, over 2368408.94 frames. ], batch size: 99, lr: 3.38e-03, grad_scale: 8.0 2023-05-17 06:23:22,689 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.783e+02 2.666e+02 3.356e+02 3.786e+02 5.808e+02, threshold=6.713e+02, percent-clipped=0.0 2023-05-17 06:23:36,955 INFO [finetune.py:992] (0/2) Epoch 16, batch 8600, loss[loss=0.1867, simple_loss=0.2714, pruned_loss=0.05096, over 12041.00 frames. ], tot_loss[loss=0.1657, simple_loss=0.2553, pruned_loss=0.03806, over 2369587.22 frames. ], batch size: 40, lr: 3.38e-03, grad_scale: 8.0 2023-05-17 06:23:39,957 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=292115.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 06:23:41,690 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.58 vs. limit=2.0 2023-05-17 06:23:52,071 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=292132.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 06:23:54,466 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.3329, 4.0782, 4.0757, 4.4132, 2.9844, 3.9568, 2.4479, 4.1602], device='cuda:0'), covar=tensor([0.1536, 0.0694, 0.0971, 0.0596, 0.1226, 0.0619, 0.2043, 0.1050], device='cuda:0'), in_proj_covar=tensor([0.0232, 0.0273, 0.0306, 0.0368, 0.0249, 0.0249, 0.0268, 0.0379], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-17 06:24:12,744 INFO [finetune.py:992] (0/2) Epoch 16, batch 8650, loss[loss=0.1718, simple_loss=0.256, pruned_loss=0.04377, over 12091.00 frames. ], tot_loss[loss=0.1658, simple_loss=0.2554, pruned_loss=0.03804, over 2368338.90 frames. ], batch size: 32, lr: 3.38e-03, grad_scale: 8.0 2023-05-17 06:24:34,124 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.761e+02 2.729e+02 3.150e+02 3.727e+02 5.480e+02, threshold=6.300e+02, percent-clipped=0.0 2023-05-17 06:24:42,099 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.34 vs. limit=2.0 2023-05-17 06:24:49,465 INFO [finetune.py:992] (0/2) Epoch 16, batch 8700, loss[loss=0.1802, simple_loss=0.2807, pruned_loss=0.0398, over 12306.00 frames. ], tot_loss[loss=0.1647, simple_loss=0.2547, pruned_loss=0.03739, over 2379011.64 frames. ], batch size: 34, lr: 3.38e-03, grad_scale: 8.0 2023-05-17 06:25:10,885 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 2023-05-17 06:25:25,802 INFO [finetune.py:992] (0/2) Epoch 16, batch 8750, loss[loss=0.1424, simple_loss=0.2251, pruned_loss=0.02983, over 11801.00 frames. ], tot_loss[loss=0.165, simple_loss=0.2549, pruned_loss=0.03756, over 2377361.64 frames. ], batch size: 26, lr: 3.38e-03, grad_scale: 8.0 2023-05-17 06:25:29,091 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=3.19 vs. limit=5.0 2023-05-17 06:25:45,431 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=2.00 vs. limit=2.0 2023-05-17 06:25:47,146 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.959e+02 2.707e+02 3.123e+02 3.591e+02 5.330e+02, threshold=6.245e+02, percent-clipped=0.0 2023-05-17 06:26:01,304 INFO [finetune.py:992] (0/2) Epoch 16, batch 8800, loss[loss=0.1612, simple_loss=0.2545, pruned_loss=0.03398, over 12191.00 frames. ], tot_loss[loss=0.1658, simple_loss=0.2556, pruned_loss=0.03801, over 2374285.41 frames. ], batch size: 35, lr: 3.38e-03, grad_scale: 8.0 2023-05-17 06:26:14,620 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=3.43 vs. limit=5.0 2023-05-17 06:26:19,600 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.99 vs. limit=2.0 2023-05-17 06:26:37,616 INFO [finetune.py:992] (0/2) Epoch 16, batch 8850, loss[loss=0.1583, simple_loss=0.241, pruned_loss=0.03779, over 12187.00 frames. ], tot_loss[loss=0.1656, simple_loss=0.2554, pruned_loss=0.0379, over 2373566.83 frames. ], batch size: 31, lr: 3.38e-03, grad_scale: 8.0 2023-05-17 06:26:59,615 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.980e+02 2.756e+02 3.464e+02 4.372e+02 8.358e+02, threshold=6.928e+02, percent-clipped=4.0 2023-05-17 06:27:13,635 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=292410.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 06:27:14,278 INFO [finetune.py:992] (0/2) Epoch 16, batch 8900, loss[loss=0.1403, simple_loss=0.2263, pruned_loss=0.02712, over 12085.00 frames. ], tot_loss[loss=0.1659, simple_loss=0.2555, pruned_loss=0.03813, over 2368310.67 frames. ], batch size: 32, lr: 3.38e-03, grad_scale: 8.0 2023-05-17 06:27:29,253 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=292432.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 06:27:49,598 INFO [finetune.py:992] (0/2) Epoch 16, batch 8950, loss[loss=0.1634, simple_loss=0.264, pruned_loss=0.03134, over 12149.00 frames. ], tot_loss[loss=0.1659, simple_loss=0.2554, pruned_loss=0.03815, over 2370491.86 frames. ], batch size: 34, lr: 3.38e-03, grad_scale: 8.0 2023-05-17 06:28:03,313 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=292480.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 06:28:03,807 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.34 vs. limit=2.0 2023-05-17 06:28:11,807 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.862e+02 2.832e+02 3.435e+02 4.189e+02 1.010e+03, threshold=6.871e+02, percent-clipped=2.0 2023-05-17 06:28:14,642 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.7125, 4.5513, 4.5015, 4.5996, 4.2531, 4.7552, 4.6707, 4.8167], device='cuda:0'), covar=tensor([0.0235, 0.0196, 0.0247, 0.0405, 0.0834, 0.0409, 0.0173, 0.0241], device='cuda:0'), in_proj_covar=tensor([0.0210, 0.0208, 0.0206, 0.0261, 0.0253, 0.0234, 0.0187, 0.0245], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0005, 0.0004, 0.0005, 0.0003, 0.0005], device='cuda:0') 2023-05-17 06:28:26,610 INFO [finetune.py:992] (0/2) Epoch 16, batch 9000, loss[loss=0.2089, simple_loss=0.2849, pruned_loss=0.06647, over 8349.00 frames. ], tot_loss[loss=0.166, simple_loss=0.2554, pruned_loss=0.03834, over 2372890.63 frames. ], batch size: 98, lr: 3.38e-03, grad_scale: 8.0 2023-05-17 06:28:26,611 INFO [finetune.py:1017] (0/2) Computing validation loss 2023-05-17 06:28:44,910 INFO [finetune.py:1026] (0/2) Epoch 16, validation: loss=0.3156, simple_loss=0.3914, pruned_loss=0.1199, over 1020973.00 frames. 2023-05-17 06:28:44,910 INFO [finetune.py:1027] (0/2) Maximum memory allocated so far is 12827MB 2023-05-17 06:28:53,119 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.95 vs. limit=2.0 2023-05-17 06:29:14,503 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.98 vs. limit=2.0 2023-05-17 06:29:20,399 INFO [finetune.py:992] (0/2) Epoch 16, batch 9050, loss[loss=0.1875, simple_loss=0.2679, pruned_loss=0.05351, over 12287.00 frames. ], tot_loss[loss=0.1658, simple_loss=0.2552, pruned_loss=0.0382, over 2371595.86 frames. ], batch size: 34, lr: 3.38e-03, grad_scale: 8.0 2023-05-17 06:29:22,173 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=2.67 vs. limit=5.0 2023-05-17 06:29:22,913 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.98 vs. limit=2.0 2023-05-17 06:29:34,002 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.1427, 4.1666, 4.2746, 4.5746, 3.0268, 4.0231, 2.6659, 4.2975], device='cuda:0'), covar=tensor([0.1633, 0.0653, 0.0764, 0.0478, 0.1188, 0.0597, 0.1800, 0.1094], device='cuda:0'), in_proj_covar=tensor([0.0229, 0.0270, 0.0303, 0.0363, 0.0245, 0.0247, 0.0265, 0.0374], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-17 06:29:42,888 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.669e+02 2.660e+02 3.096e+02 3.918e+02 9.243e+02, threshold=6.193e+02, percent-clipped=1.0 2023-05-17 06:29:57,495 INFO [finetune.py:992] (0/2) Epoch 16, batch 9100, loss[loss=0.1516, simple_loss=0.2461, pruned_loss=0.02858, over 12187.00 frames. ], tot_loss[loss=0.1656, simple_loss=0.2552, pruned_loss=0.03803, over 2370296.67 frames. ], batch size: 31, lr: 3.38e-03, grad_scale: 8.0 2023-05-17 06:30:33,346 INFO [finetune.py:992] (0/2) Epoch 16, batch 9150, loss[loss=0.1591, simple_loss=0.2542, pruned_loss=0.03195, over 12280.00 frames. ], tot_loss[loss=0.1657, simple_loss=0.2555, pruned_loss=0.038, over 2368974.17 frames. ], batch size: 33, lr: 3.37e-03, grad_scale: 8.0 2023-05-17 06:30:54,420 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.740e+02 2.517e+02 2.822e+02 3.782e+02 1.055e+03, threshold=5.645e+02, percent-clipped=1.0 2023-05-17 06:30:55,556 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2023-05-17 06:31:08,254 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=292710.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 06:31:08,882 INFO [finetune.py:992] (0/2) Epoch 16, batch 9200, loss[loss=0.1766, simple_loss=0.2687, pruned_loss=0.04222, over 10527.00 frames. ], tot_loss[loss=0.165, simple_loss=0.2549, pruned_loss=0.03752, over 2372159.05 frames. ], batch size: 68, lr: 3.37e-03, grad_scale: 8.0 2023-05-17 06:31:37,478 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.6624, 3.1414, 5.1259, 2.5950, 2.8187, 3.6578, 3.1794, 3.6899], device='cuda:0'), covar=tensor([0.0416, 0.1354, 0.0289, 0.1248, 0.1965, 0.1583, 0.1448, 0.1385], device='cuda:0'), in_proj_covar=tensor([0.0242, 0.0242, 0.0260, 0.0186, 0.0241, 0.0296, 0.0229, 0.0270], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-17 06:31:43,741 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=292758.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 06:31:45,807 INFO [finetune.py:992] (0/2) Epoch 16, batch 9250, loss[loss=0.1472, simple_loss=0.2354, pruned_loss=0.02953, over 12036.00 frames. ], tot_loss[loss=0.1655, simple_loss=0.2554, pruned_loss=0.03776, over 2365245.66 frames. ], batch size: 31, lr: 3.37e-03, grad_scale: 8.0 2023-05-17 06:31:52,383 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.3619, 3.3160, 3.0761, 2.9674, 2.7129, 2.6223, 3.3235, 2.1923], device='cuda:0'), covar=tensor([0.0425, 0.0143, 0.0234, 0.0230, 0.0421, 0.0368, 0.0154, 0.0555], device='cuda:0'), in_proj_covar=tensor([0.0203, 0.0169, 0.0174, 0.0198, 0.0208, 0.0207, 0.0182, 0.0211], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-17 06:32:07,220 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.896e+02 2.575e+02 3.009e+02 3.644e+02 6.909e+02, threshold=6.018e+02, percent-clipped=2.0 2023-05-17 06:32:17,261 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=4.64 vs. limit=5.0 2023-05-17 06:32:21,776 INFO [finetune.py:992] (0/2) Epoch 16, batch 9300, loss[loss=0.1707, simple_loss=0.2576, pruned_loss=0.04186, over 12288.00 frames. ], tot_loss[loss=0.1655, simple_loss=0.2555, pruned_loss=0.03776, over 2365511.35 frames. ], batch size: 33, lr: 3.37e-03, grad_scale: 8.0 2023-05-17 06:32:25,949 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.4223, 3.5129, 3.1472, 3.0105, 2.7560, 2.6760, 3.5048, 2.2566], device='cuda:0'), covar=tensor([0.0450, 0.0169, 0.0238, 0.0256, 0.0447, 0.0395, 0.0188, 0.0560], device='cuda:0'), in_proj_covar=tensor([0.0204, 0.0169, 0.0175, 0.0199, 0.0209, 0.0207, 0.0182, 0.0212], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-17 06:32:57,234 INFO [finetune.py:992] (0/2) Epoch 16, batch 9350, loss[loss=0.1568, simple_loss=0.2452, pruned_loss=0.03422, over 12302.00 frames. ], tot_loss[loss=0.1651, simple_loss=0.2555, pruned_loss=0.03738, over 2373555.01 frames. ], batch size: 34, lr: 3.37e-03, grad_scale: 8.0 2023-05-17 06:33:19,941 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.707e+02 2.513e+02 3.048e+02 3.850e+02 7.180e+02, threshold=6.096e+02, percent-clipped=3.0 2023-05-17 06:33:28,119 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.7056, 2.9367, 4.4874, 4.6802, 3.0106, 2.7535, 3.0586, 2.1669], device='cuda:0'), covar=tensor([0.1671, 0.2972, 0.0483, 0.0461, 0.1319, 0.2509, 0.2752, 0.4262], device='cuda:0'), in_proj_covar=tensor([0.0305, 0.0394, 0.0279, 0.0305, 0.0280, 0.0321, 0.0399, 0.0381], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-17 06:33:34,202 INFO [finetune.py:992] (0/2) Epoch 16, batch 9400, loss[loss=0.1635, simple_loss=0.2518, pruned_loss=0.03758, over 11758.00 frames. ], tot_loss[loss=0.165, simple_loss=0.2553, pruned_loss=0.03736, over 2374309.90 frames. ], batch size: 44, lr: 3.37e-03, grad_scale: 8.0 2023-05-17 06:33:56,476 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.9447, 3.0339, 4.7819, 4.9824, 3.0567, 2.8409, 3.1363, 2.4620], device='cuda:0'), covar=tensor([0.1599, 0.3141, 0.0486, 0.0436, 0.1340, 0.2509, 0.2733, 0.3922], device='cuda:0'), in_proj_covar=tensor([0.0305, 0.0394, 0.0280, 0.0305, 0.0280, 0.0321, 0.0400, 0.0382], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-17 06:34:09,234 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.96 vs. limit=2.0 2023-05-17 06:34:09,552 INFO [finetune.py:992] (0/2) Epoch 16, batch 9450, loss[loss=0.1667, simple_loss=0.2542, pruned_loss=0.03963, over 12132.00 frames. ], tot_loss[loss=0.1646, simple_loss=0.2548, pruned_loss=0.03715, over 2370036.16 frames. ], batch size: 38, lr: 3.37e-03, grad_scale: 8.0 2023-05-17 06:34:14,103 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=292967.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 06:34:30,774 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.988e+02 2.666e+02 3.237e+02 3.941e+02 7.125e+02, threshold=6.474e+02, percent-clipped=3.0 2023-05-17 06:34:45,419 INFO [finetune.py:992] (0/2) Epoch 16, batch 9500, loss[loss=0.1814, simple_loss=0.2715, pruned_loss=0.04565, over 12296.00 frames. ], tot_loss[loss=0.1649, simple_loss=0.255, pruned_loss=0.0374, over 2375346.03 frames. ], batch size: 33, lr: 3.37e-03, grad_scale: 8.0 2023-05-17 06:34:58,089 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=293028.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 06:35:22,139 INFO [finetune.py:992] (0/2) Epoch 16, batch 9550, loss[loss=0.1452, simple_loss=0.2426, pruned_loss=0.02392, over 12259.00 frames. ], tot_loss[loss=0.1649, simple_loss=0.2551, pruned_loss=0.03739, over 2376623.53 frames. ], batch size: 32, lr: 3.37e-03, grad_scale: 8.0 2023-05-17 06:35:36,637 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=293081.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 06:35:43,594 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.048e+02 2.670e+02 3.195e+02 3.630e+02 6.304e+02, threshold=6.391e+02, percent-clipped=0.0 2023-05-17 06:35:56,031 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.4429, 2.6409, 2.9450, 4.3232, 2.1726, 4.3290, 4.4936, 4.4205], device='cuda:0'), covar=tensor([0.0148, 0.1273, 0.0612, 0.0181, 0.1502, 0.0253, 0.0131, 0.0143], device='cuda:0'), in_proj_covar=tensor([0.0128, 0.0209, 0.0188, 0.0126, 0.0194, 0.0186, 0.0183, 0.0131], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002], device='cuda:0') 2023-05-17 06:35:57,894 INFO [finetune.py:992] (0/2) Epoch 16, batch 9600, loss[loss=0.2069, simple_loss=0.2833, pruned_loss=0.06527, over 8320.00 frames. ], tot_loss[loss=0.1653, simple_loss=0.255, pruned_loss=0.03776, over 2372393.00 frames. ], batch size: 99, lr: 3.37e-03, grad_scale: 8.0 2023-05-17 06:36:02,329 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=293117.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 06:36:04,570 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.8607, 3.6398, 5.3253, 2.7373, 2.9693, 3.7459, 3.3754, 3.8041], device='cuda:0'), covar=tensor([0.0443, 0.1021, 0.0276, 0.1168, 0.1962, 0.1547, 0.1357, 0.1333], device='cuda:0'), in_proj_covar=tensor([0.0242, 0.0243, 0.0260, 0.0187, 0.0240, 0.0297, 0.0230, 0.0270], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-17 06:36:20,482 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=293142.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 06:36:27,793 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.6647, 4.1635, 4.2789, 4.5377, 4.4537, 4.5217, 4.4140, 2.2060], device='cuda:0'), covar=tensor([0.0185, 0.0141, 0.0184, 0.0105, 0.0082, 0.0220, 0.0151, 0.1303], device='cuda:0'), in_proj_covar=tensor([0.0072, 0.0083, 0.0085, 0.0076, 0.0062, 0.0096, 0.0085, 0.0101], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-17 06:36:33,909 INFO [finetune.py:992] (0/2) Epoch 16, batch 9650, loss[loss=0.2611, simple_loss=0.3287, pruned_loss=0.09675, over 8364.00 frames. ], tot_loss[loss=0.1654, simple_loss=0.2553, pruned_loss=0.03778, over 2374474.24 frames. ], batch size: 98, lr: 3.37e-03, grad_scale: 8.0 2023-05-17 06:36:47,557 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=293178.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 06:36:57,418 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.924e+02 2.792e+02 3.150e+02 3.953e+02 6.811e+02, threshold=6.300e+02, percent-clipped=2.0 2023-05-17 06:37:07,589 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.5374, 5.4487, 5.2974, 4.7547, 4.8975, 5.4355, 5.0699, 4.9002], device='cuda:0'), covar=tensor([0.0739, 0.1075, 0.0766, 0.1729, 0.1040, 0.0832, 0.1563, 0.1028], device='cuda:0'), in_proj_covar=tensor([0.0649, 0.0586, 0.0541, 0.0649, 0.0431, 0.0755, 0.0800, 0.0587], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0004, 0.0004, 0.0003], device='cuda:0') 2023-05-17 06:37:09,045 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.9523, 5.8671, 5.4210, 5.4596, 5.9443, 5.2647, 5.3926, 5.4705], device='cuda:0'), covar=tensor([0.1682, 0.1013, 0.1139, 0.1853, 0.1046, 0.2267, 0.2117, 0.1153], device='cuda:0'), in_proj_covar=tensor([0.0372, 0.0512, 0.0408, 0.0464, 0.0482, 0.0444, 0.0412, 0.0401], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-17 06:37:11,097 INFO [finetune.py:992] (0/2) Epoch 16, batch 9700, loss[loss=0.1692, simple_loss=0.2629, pruned_loss=0.03775, over 10412.00 frames. ], tot_loss[loss=0.1644, simple_loss=0.2543, pruned_loss=0.03727, over 2373801.18 frames. ], batch size: 68, lr: 3.37e-03, grad_scale: 4.0 2023-05-17 06:37:25,142 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=2.82 vs. limit=5.0 2023-05-17 06:37:46,874 INFO [finetune.py:992] (0/2) Epoch 16, batch 9750, loss[loss=0.1656, simple_loss=0.2573, pruned_loss=0.03692, over 12149.00 frames. ], tot_loss[loss=0.1642, simple_loss=0.2542, pruned_loss=0.03708, over 2379164.55 frames. ], batch size: 34, lr: 3.37e-03, grad_scale: 4.0 2023-05-17 06:38:08,866 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.743e+02 2.520e+02 2.905e+02 3.730e+02 7.475e+02, threshold=5.811e+02, percent-clipped=2.0 2023-05-17 06:38:22,491 INFO [finetune.py:992] (0/2) Epoch 16, batch 9800, loss[loss=0.151, simple_loss=0.2337, pruned_loss=0.03414, over 12117.00 frames. ], tot_loss[loss=0.1639, simple_loss=0.2538, pruned_loss=0.03701, over 2382400.23 frames. ], batch size: 30, lr: 3.37e-03, grad_scale: 4.0 2023-05-17 06:38:31,934 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=293323.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 06:38:55,316 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=2.70 vs. limit=5.0 2023-05-17 06:38:59,105 INFO [finetune.py:992] (0/2) Epoch 16, batch 9850, loss[loss=0.1635, simple_loss=0.2472, pruned_loss=0.03989, over 12083.00 frames. ], tot_loss[loss=0.1642, simple_loss=0.2538, pruned_loss=0.03729, over 2372052.72 frames. ], batch size: 32, lr: 3.37e-03, grad_scale: 4.0 2023-05-17 06:38:59,349 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=293361.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 06:39:01,391 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=293364.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 06:39:21,032 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.867e+02 2.575e+02 3.091e+02 3.647e+02 7.324e+02, threshold=6.182e+02, percent-clipped=5.0 2023-05-17 06:39:26,460 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.89 vs. limit=2.0 2023-05-17 06:39:34,876 INFO [finetune.py:992] (0/2) Epoch 16, batch 9900, loss[loss=0.1515, simple_loss=0.2503, pruned_loss=0.02637, over 11616.00 frames. ], tot_loss[loss=0.1646, simple_loss=0.2544, pruned_loss=0.03736, over 2374364.14 frames. ], batch size: 48, lr: 3.37e-03, grad_scale: 4.0 2023-05-17 06:39:43,013 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=293422.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 06:39:45,181 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=293425.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 06:39:53,740 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=293437.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 06:40:10,788 INFO [finetune.py:992] (0/2) Epoch 16, batch 9950, loss[loss=0.1698, simple_loss=0.2674, pruned_loss=0.03613, over 12279.00 frames. ], tot_loss[loss=0.1643, simple_loss=0.2538, pruned_loss=0.03741, over 2377862.10 frames. ], batch size: 37, lr: 3.37e-03, grad_scale: 4.0 2023-05-17 06:40:20,702 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=293473.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 06:40:26,268 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.34 vs. limit=2.0 2023-05-17 06:40:28,222 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.6426, 3.2688, 5.1360, 2.5714, 2.7589, 3.7366, 3.1480, 3.8024], device='cuda:0'), covar=tensor([0.0481, 0.1215, 0.0379, 0.1249, 0.2050, 0.1586, 0.1454, 0.1222], device='cuda:0'), in_proj_covar=tensor([0.0244, 0.0244, 0.0262, 0.0187, 0.0242, 0.0298, 0.0230, 0.0271], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-17 06:40:34,295 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.863e+02 2.574e+02 3.007e+02 3.575e+02 1.265e+03, threshold=6.015e+02, percent-clipped=3.0 2023-05-17 06:40:47,684 INFO [finetune.py:992] (0/2) Epoch 16, batch 10000, loss[loss=0.1573, simple_loss=0.2513, pruned_loss=0.03163, over 12206.00 frames. ], tot_loss[loss=0.1645, simple_loss=0.2545, pruned_loss=0.0373, over 2378850.57 frames. ], batch size: 35, lr: 3.37e-03, grad_scale: 8.0 2023-05-17 06:41:09,376 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.3300, 2.5007, 2.9815, 4.1978, 2.4387, 4.2434, 4.3668, 4.3255], device='cuda:0'), covar=tensor([0.0128, 0.1268, 0.0592, 0.0190, 0.1277, 0.0275, 0.0142, 0.0130], device='cuda:0'), in_proj_covar=tensor([0.0127, 0.0208, 0.0188, 0.0127, 0.0193, 0.0184, 0.0181, 0.0130], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-17 06:41:23,672 INFO [finetune.py:992] (0/2) Epoch 16, batch 10050, loss[loss=0.1712, simple_loss=0.2603, pruned_loss=0.04103, over 11257.00 frames. ], tot_loss[loss=0.1643, simple_loss=0.254, pruned_loss=0.03733, over 2379245.48 frames. ], batch size: 55, lr: 3.37e-03, grad_scale: 8.0 2023-05-17 06:41:45,651 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.849e+02 2.608e+02 3.078e+02 3.728e+02 5.894e+02, threshold=6.156e+02, percent-clipped=0.0 2023-05-17 06:41:59,645 INFO [finetune.py:992] (0/2) Epoch 16, batch 10100, loss[loss=0.1481, simple_loss=0.2428, pruned_loss=0.0267, over 12401.00 frames. ], tot_loss[loss=0.1649, simple_loss=0.2543, pruned_loss=0.03775, over 2383931.47 frames. ], batch size: 32, lr: 3.37e-03, grad_scale: 8.0 2023-05-17 06:41:59,846 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.1313, 4.7072, 4.8636, 5.0149, 4.8591, 4.9234, 4.8532, 2.6828], device='cuda:0'), covar=tensor([0.0083, 0.0078, 0.0090, 0.0066, 0.0049, 0.0104, 0.0088, 0.0783], device='cuda:0'), in_proj_covar=tensor([0.0072, 0.0083, 0.0085, 0.0077, 0.0062, 0.0096, 0.0085, 0.0101], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-17 06:42:09,569 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=293623.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 06:42:35,652 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.91 vs. limit=2.0 2023-05-17 06:42:36,648 INFO [finetune.py:992] (0/2) Epoch 16, batch 10150, loss[loss=0.168, simple_loss=0.2564, pruned_loss=0.0398, over 7932.00 frames. ], tot_loss[loss=0.1656, simple_loss=0.2551, pruned_loss=0.03803, over 2373123.52 frames. ], batch size: 98, lr: 3.37e-03, grad_scale: 8.0 2023-05-17 06:42:43,785 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=293671.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 06:42:50,416 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.6097, 5.4250, 5.5482, 5.5741, 5.1959, 5.2408, 5.0064, 5.4504], device='cuda:0'), covar=tensor([0.0697, 0.0564, 0.0817, 0.0601, 0.1968, 0.1284, 0.0533, 0.1252], device='cuda:0'), in_proj_covar=tensor([0.0567, 0.0726, 0.0640, 0.0654, 0.0892, 0.0781, 0.0590, 0.0504], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0005, 0.0004, 0.0004, 0.0005, 0.0005, 0.0004, 0.0003], device='cuda:0') 2023-05-17 06:42:58,806 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.069e+02 2.676e+02 3.125e+02 3.829e+02 1.121e+03, threshold=6.250e+02, percent-clipped=3.0 2023-05-17 06:43:03,665 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.36 vs. limit=2.0 2023-05-17 06:43:12,372 INFO [finetune.py:992] (0/2) Epoch 16, batch 10200, loss[loss=0.1706, simple_loss=0.2597, pruned_loss=0.04072, over 12344.00 frames. ], tot_loss[loss=0.1663, simple_loss=0.2559, pruned_loss=0.03837, over 2370817.58 frames. ], batch size: 31, lr: 3.37e-03, grad_scale: 8.0 2023-05-17 06:43:15,448 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.8327, 3.0559, 4.7010, 4.9226, 2.9681, 2.7110, 3.2093, 2.3686], device='cuda:0'), covar=tensor([0.1661, 0.2799, 0.0476, 0.0402, 0.1367, 0.2536, 0.2617, 0.3930], device='cuda:0'), in_proj_covar=tensor([0.0307, 0.0394, 0.0280, 0.0307, 0.0281, 0.0322, 0.0400, 0.0382], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-17 06:43:16,646 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=293717.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 06:43:18,799 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=293720.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 06:43:31,062 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=293737.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 06:43:49,495 INFO [finetune.py:992] (0/2) Epoch 16, batch 10250, loss[loss=0.1517, simple_loss=0.2345, pruned_loss=0.03444, over 12183.00 frames. ], tot_loss[loss=0.1653, simple_loss=0.2552, pruned_loss=0.03764, over 2377718.61 frames. ], batch size: 31, lr: 3.37e-03, grad_scale: 8.0 2023-05-17 06:43:58,265 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=293773.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 06:44:06,588 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=293785.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 06:44:11,367 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.933e+02 2.592e+02 2.928e+02 3.521e+02 7.905e+02, threshold=5.855e+02, percent-clipped=2.0 2023-05-17 06:44:24,806 INFO [finetune.py:992] (0/2) Epoch 16, batch 10300, loss[loss=0.1554, simple_loss=0.2516, pruned_loss=0.02961, over 12351.00 frames. ], tot_loss[loss=0.1655, simple_loss=0.2556, pruned_loss=0.03769, over 2371767.68 frames. ], batch size: 36, lr: 3.37e-03, grad_scale: 8.0 2023-05-17 06:44:28,211 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.89 vs. limit=2.0 2023-05-17 06:44:31,783 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=293821.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 06:45:00,319 INFO [finetune.py:992] (0/2) Epoch 16, batch 10350, loss[loss=0.1759, simple_loss=0.2658, pruned_loss=0.04302, over 11593.00 frames. ], tot_loss[loss=0.1652, simple_loss=0.2552, pruned_loss=0.03765, over 2370515.18 frames. ], batch size: 48, lr: 3.37e-03, grad_scale: 8.0 2023-05-17 06:45:22,196 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.697e+02 2.689e+02 3.056e+02 3.910e+02 6.982e+02, threshold=6.113e+02, percent-clipped=5.0 2023-05-17 06:45:36,316 INFO [finetune.py:992] (0/2) Epoch 16, batch 10400, loss[loss=0.1652, simple_loss=0.2474, pruned_loss=0.04155, over 12349.00 frames. ], tot_loss[loss=0.1651, simple_loss=0.2549, pruned_loss=0.03766, over 2372185.71 frames. ], batch size: 31, lr: 3.37e-03, grad_scale: 8.0 2023-05-17 06:46:12,628 INFO [finetune.py:992] (0/2) Epoch 16, batch 10450, loss[loss=0.1698, simple_loss=0.2502, pruned_loss=0.04469, over 12115.00 frames. ], tot_loss[loss=0.1649, simple_loss=0.2546, pruned_loss=0.03758, over 2362762.62 frames. ], batch size: 30, lr: 3.37e-03, grad_scale: 8.0 2023-05-17 06:46:31,107 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=293987.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 06:46:33,213 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.0596, 4.8460, 4.9993, 5.0393, 4.5939, 4.7297, 4.4751, 4.9435], device='cuda:0'), covar=tensor([0.0745, 0.0674, 0.0953, 0.0651, 0.2326, 0.1433, 0.0708, 0.1114], device='cuda:0'), in_proj_covar=tensor([0.0572, 0.0733, 0.0644, 0.0663, 0.0899, 0.0789, 0.0597, 0.0509], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0005, 0.0004, 0.0004, 0.0005, 0.0005, 0.0004, 0.0003], device='cuda:0') 2023-05-17 06:46:34,540 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.957e+02 2.724e+02 3.226e+02 3.643e+02 5.561e+02, threshold=6.451e+02, percent-clipped=0.0 2023-05-17 06:46:40,451 INFO [checkpoint.py:75] (0/2) Saving checkpoint to pruned_transducer_stateless7/exp_giga_finetune/checkpoint-194000.pt 2023-05-17 06:46:51,156 INFO [finetune.py:992] (0/2) Epoch 16, batch 10500, loss[loss=0.1716, simple_loss=0.2702, pruned_loss=0.03652, over 12346.00 frames. ], tot_loss[loss=0.1647, simple_loss=0.2548, pruned_loss=0.03737, over 2363334.89 frames. ], batch size: 31, lr: 3.37e-03, grad_scale: 8.0 2023-05-17 06:46:55,611 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=294017.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 06:46:57,797 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=294020.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 06:47:18,432 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=294048.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 06:47:27,950 INFO [finetune.py:992] (0/2) Epoch 16, batch 10550, loss[loss=0.1464, simple_loss=0.2317, pruned_loss=0.03053, over 12260.00 frames. ], tot_loss[loss=0.1631, simple_loss=0.2532, pruned_loss=0.03655, over 2376322.55 frames. ], batch size: 32, lr: 3.37e-03, grad_scale: 8.0 2023-05-17 06:47:28,149 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.0772, 4.9198, 4.8861, 4.8830, 4.5597, 5.0427, 5.0181, 5.2257], device='cuda:0'), covar=tensor([0.0224, 0.0176, 0.0224, 0.0431, 0.0857, 0.0298, 0.0178, 0.0191], device='cuda:0'), in_proj_covar=tensor([0.0209, 0.0208, 0.0204, 0.0261, 0.0255, 0.0233, 0.0188, 0.0245], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0005, 0.0004, 0.0005, 0.0004, 0.0005], device='cuda:0') 2023-05-17 06:47:30,800 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=294065.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 06:47:32,953 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=294068.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 06:47:47,464 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.2245, 5.0641, 5.0398, 5.0397, 4.7500, 5.2237, 5.1629, 5.3199], device='cuda:0'), covar=tensor([0.0230, 0.0175, 0.0190, 0.0415, 0.0839, 0.0353, 0.0180, 0.0210], device='cuda:0'), in_proj_covar=tensor([0.0208, 0.0207, 0.0203, 0.0260, 0.0254, 0.0232, 0.0187, 0.0244], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0005, 0.0004, 0.0005, 0.0003, 0.0005], device='cuda:0') 2023-05-17 06:47:50,104 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.736e+02 2.579e+02 2.988e+02 3.523e+02 1.065e+03, threshold=5.977e+02, percent-clipped=6.0 2023-05-17 06:47:50,306 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.1138, 2.0532, 3.0918, 4.0153, 2.0587, 4.2009, 4.2378, 4.2453], device='cuda:0'), covar=tensor([0.0176, 0.1617, 0.0509, 0.0191, 0.1541, 0.0204, 0.0145, 0.0150], device='cuda:0'), in_proj_covar=tensor([0.0128, 0.0209, 0.0189, 0.0127, 0.0193, 0.0185, 0.0182, 0.0130], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-17 06:48:03,336 INFO [finetune.py:992] (0/2) Epoch 16, batch 10600, loss[loss=0.1663, simple_loss=0.2545, pruned_loss=0.03905, over 12358.00 frames. ], tot_loss[loss=0.1629, simple_loss=0.2529, pruned_loss=0.03641, over 2380487.52 frames. ], batch size: 35, lr: 3.36e-03, grad_scale: 8.0 2023-05-17 06:48:37,357 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.0253, 4.9047, 4.9862, 5.0503, 4.6834, 4.7417, 4.5116, 4.9253], device='cuda:0'), covar=tensor([0.0779, 0.0568, 0.0947, 0.0598, 0.1869, 0.1342, 0.0565, 0.1155], device='cuda:0'), in_proj_covar=tensor([0.0565, 0.0726, 0.0637, 0.0655, 0.0887, 0.0779, 0.0589, 0.0504], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0005, 0.0005, 0.0004, 0.0003], device='cuda:0') 2023-05-17 06:48:38,736 INFO [finetune.py:992] (0/2) Epoch 16, batch 10650, loss[loss=0.1788, simple_loss=0.2737, pruned_loss=0.04197, over 11990.00 frames. ], tot_loss[loss=0.1643, simple_loss=0.2545, pruned_loss=0.037, over 2373105.84 frames. ], batch size: 42, lr: 3.36e-03, grad_scale: 8.0 2023-05-17 06:48:45,625 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=294170.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 06:48:48,515 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.1629, 2.6577, 3.7359, 3.1830, 3.4911, 3.2506, 2.6850, 3.6157], device='cuda:0'), covar=tensor([0.0162, 0.0393, 0.0152, 0.0235, 0.0200, 0.0219, 0.0376, 0.0136], device='cuda:0'), in_proj_covar=tensor([0.0186, 0.0211, 0.0200, 0.0195, 0.0227, 0.0174, 0.0204, 0.0195], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-17 06:49:01,147 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.814e+02 2.732e+02 2.981e+02 3.538e+02 9.530e+02, threshold=5.962e+02, percent-clipped=5.0 2023-05-17 06:49:16,475 INFO [finetune.py:992] (0/2) Epoch 16, batch 10700, loss[loss=0.1384, simple_loss=0.2241, pruned_loss=0.02633, over 12178.00 frames. ], tot_loss[loss=0.1638, simple_loss=0.2537, pruned_loss=0.03692, over 2367038.65 frames. ], batch size: 31, lr: 3.36e-03, grad_scale: 8.0 2023-05-17 06:49:30,615 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=294231.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 06:49:51,338 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.39 vs. limit=2.0 2023-05-17 06:49:51,534 INFO [finetune.py:992] (0/2) Epoch 16, batch 10750, loss[loss=0.1393, simple_loss=0.227, pruned_loss=0.02587, over 12337.00 frames. ], tot_loss[loss=0.163, simple_loss=0.2529, pruned_loss=0.03654, over 2371776.72 frames. ], batch size: 31, lr: 3.36e-03, grad_scale: 8.0 2023-05-17 06:50:10,671 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=4.57 vs. limit=5.0 2023-05-17 06:50:13,690 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.912e+02 2.614e+02 3.041e+02 3.851e+02 7.162e+02, threshold=6.082e+02, percent-clipped=2.0 2023-05-17 06:50:27,244 INFO [finetune.py:992] (0/2) Epoch 16, batch 10800, loss[loss=0.1782, simple_loss=0.275, pruned_loss=0.0407, over 12081.00 frames. ], tot_loss[loss=0.1636, simple_loss=0.2534, pruned_loss=0.03693, over 2364652.94 frames. ], batch size: 42, lr: 3.36e-03, grad_scale: 8.0 2023-05-17 06:50:47,898 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=294339.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 06:50:50,730 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=294343.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 06:51:04,181 INFO [finetune.py:992] (0/2) Epoch 16, batch 10850, loss[loss=0.1902, simple_loss=0.2793, pruned_loss=0.0505, over 12062.00 frames. ], tot_loss[loss=0.165, simple_loss=0.2546, pruned_loss=0.03773, over 2353651.64 frames. ], batch size: 37, lr: 3.36e-03, grad_scale: 8.0 2023-05-17 06:51:08,745 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.4465, 2.3632, 3.1838, 4.3413, 2.2613, 4.3292, 4.4938, 4.4725], device='cuda:0'), covar=tensor([0.0164, 0.1479, 0.0548, 0.0155, 0.1467, 0.0273, 0.0145, 0.0132], device='cuda:0'), in_proj_covar=tensor([0.0130, 0.0212, 0.0191, 0.0128, 0.0196, 0.0188, 0.0184, 0.0131], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002], device='cuda:0') 2023-05-17 06:51:27,119 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.747e+02 2.525e+02 3.170e+02 3.668e+02 7.601e+02, threshold=6.341e+02, percent-clipped=1.0 2023-05-17 06:51:33,597 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=294400.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 06:51:41,260 INFO [finetune.py:992] (0/2) Epoch 16, batch 10900, loss[loss=0.1523, simple_loss=0.2326, pruned_loss=0.03599, over 12283.00 frames. ], tot_loss[loss=0.1656, simple_loss=0.2546, pruned_loss=0.03833, over 2350609.01 frames. ], batch size: 28, lr: 3.36e-03, grad_scale: 8.0 2023-05-17 06:51:49,756 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=294423.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 06:51:49,947 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=2.60 vs. limit=5.0 2023-05-17 06:52:16,363 INFO [finetune.py:992] (0/2) Epoch 16, batch 10950, loss[loss=0.1439, simple_loss=0.2292, pruned_loss=0.0293, over 12152.00 frames. ], tot_loss[loss=0.1659, simple_loss=0.2548, pruned_loss=0.03848, over 2355024.04 frames. ], batch size: 29, lr: 3.36e-03, grad_scale: 8.0 2023-05-17 06:52:33,483 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=294484.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 06:52:37,033 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=294488.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 06:52:39,754 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.050e+02 2.775e+02 3.300e+02 3.931e+02 7.532e+02, threshold=6.601e+02, percent-clipped=2.0 2023-05-17 06:52:53,153 INFO [finetune.py:992] (0/2) Epoch 16, batch 11000, loss[loss=0.1752, simple_loss=0.2704, pruned_loss=0.04, over 12150.00 frames. ], tot_loss[loss=0.1688, simple_loss=0.2578, pruned_loss=0.03985, over 2328329.42 frames. ], batch size: 36, lr: 3.36e-03, grad_scale: 8.0 2023-05-17 06:53:03,648 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=294526.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 06:53:20,247 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=294549.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 06:53:28,478 INFO [finetune.py:992] (0/2) Epoch 16, batch 11050, loss[loss=0.1506, simple_loss=0.2316, pruned_loss=0.03478, over 12111.00 frames. ], tot_loss[loss=0.1718, simple_loss=0.2605, pruned_loss=0.04157, over 2285506.08 frames. ], batch size: 30, lr: 3.36e-03, grad_scale: 8.0 2023-05-17 06:53:35,569 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=294571.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 06:53:51,511 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.224e+02 2.846e+02 3.526e+02 4.373e+02 7.159e+02, threshold=7.053e+02, percent-clipped=1.0 2023-05-17 06:54:03,228 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=4.46 vs. limit=5.0 2023-05-17 06:54:04,828 INFO [finetune.py:992] (0/2) Epoch 16, batch 11100, loss[loss=0.1639, simple_loss=0.2563, pruned_loss=0.0357, over 12126.00 frames. ], tot_loss[loss=0.1753, simple_loss=0.264, pruned_loss=0.04333, over 2259936.45 frames. ], batch size: 33, lr: 3.36e-03, grad_scale: 8.0 2023-05-17 06:54:19,734 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=294632.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 06:54:27,456 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=294643.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 06:54:39,428 INFO [finetune.py:992] (0/2) Epoch 16, batch 11150, loss[loss=0.2634, simple_loss=0.333, pruned_loss=0.09686, over 6646.00 frames. ], tot_loss[loss=0.1799, simple_loss=0.2679, pruned_loss=0.04593, over 2213721.08 frames. ], batch size: 98, lr: 3.36e-03, grad_scale: 8.0 2023-05-17 06:54:39,751 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.74 vs. limit=2.0 2023-05-17 06:55:01,757 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=294691.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 06:55:02,372 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.366e+02 3.566e+02 4.262e+02 5.350e+02 1.116e+03, threshold=8.524e+02, percent-clipped=9.0 2023-05-17 06:55:04,435 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=294695.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 06:55:13,299 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 2023-05-17 06:55:15,658 INFO [finetune.py:992] (0/2) Epoch 16, batch 11200, loss[loss=0.2487, simple_loss=0.3343, pruned_loss=0.08155, over 10253.00 frames. ], tot_loss[loss=0.1872, simple_loss=0.2741, pruned_loss=0.05013, over 2148116.44 frames. ], batch size: 68, lr: 3.36e-03, grad_scale: 8.0 2023-05-17 06:55:30,868 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.0771, 2.3686, 3.6609, 3.0112, 3.3869, 3.2401, 2.4795, 3.5464], device='cuda:0'), covar=tensor([0.0166, 0.0448, 0.0177, 0.0294, 0.0197, 0.0194, 0.0441, 0.0159], device='cuda:0'), in_proj_covar=tensor([0.0185, 0.0210, 0.0199, 0.0194, 0.0225, 0.0172, 0.0202, 0.0194], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-17 06:55:39,619 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.9327, 2.2177, 2.8839, 2.7104, 3.0349, 2.9878, 2.9583, 2.3689], device='cuda:0'), covar=tensor([0.0101, 0.0390, 0.0180, 0.0122, 0.0115, 0.0127, 0.0144, 0.0407], device='cuda:0'), in_proj_covar=tensor([0.0091, 0.0123, 0.0104, 0.0080, 0.0104, 0.0117, 0.0102, 0.0140], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-17 06:55:51,925 INFO [finetune.py:992] (0/2) Epoch 16, batch 11250, loss[loss=0.2116, simple_loss=0.3049, pruned_loss=0.05917, over 10528.00 frames. ], tot_loss[loss=0.1938, simple_loss=0.2804, pruned_loss=0.05363, over 2105306.19 frames. ], batch size: 69, lr: 3.36e-03, grad_scale: 8.0 2023-05-17 06:56:03,960 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=294779.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 06:56:12,886 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.093e+02 3.710e+02 4.466e+02 5.909e+02 1.059e+03, threshold=8.932e+02, percent-clipped=2.0 2023-05-17 06:56:24,987 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.0580, 3.8437, 4.0026, 4.3574, 2.9679, 3.7736, 2.5291, 3.9827], device='cuda:0'), covar=tensor([0.1873, 0.0951, 0.0956, 0.0703, 0.1323, 0.0754, 0.2045, 0.1265], device='cuda:0'), in_proj_covar=tensor([0.0229, 0.0270, 0.0299, 0.0362, 0.0244, 0.0245, 0.0263, 0.0373], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-17 06:56:26,094 INFO [finetune.py:992] (0/2) Epoch 16, batch 11300, loss[loss=0.3164, simple_loss=0.3806, pruned_loss=0.1261, over 7146.00 frames. ], tot_loss[loss=0.2009, simple_loss=0.2866, pruned_loss=0.05767, over 2042421.62 frames. ], batch size: 99, lr: 3.36e-03, grad_scale: 8.0 2023-05-17 06:56:26,333 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.7204, 2.1267, 2.9666, 2.5850, 2.8437, 2.9360, 2.0863, 2.9548], device='cuda:0'), covar=tensor([0.0119, 0.0436, 0.0127, 0.0240, 0.0174, 0.0168, 0.0468, 0.0143], device='cuda:0'), in_proj_covar=tensor([0.0184, 0.0209, 0.0198, 0.0193, 0.0224, 0.0171, 0.0201, 0.0193], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-17 06:56:37,560 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=294826.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 06:56:49,895 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=294844.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 06:57:01,352 INFO [finetune.py:992] (0/2) Epoch 16, batch 11350, loss[loss=0.2568, simple_loss=0.3207, pruned_loss=0.09644, over 7357.00 frames. ], tot_loss[loss=0.2069, simple_loss=0.2917, pruned_loss=0.06111, over 1984236.17 frames. ], batch size: 99, lr: 3.36e-03, grad_scale: 8.0 2023-05-17 06:57:09,997 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=294874.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 06:57:23,765 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.315e+02 3.427e+02 4.075e+02 4.942e+02 2.146e+03, threshold=8.150e+02, percent-clipped=2.0 2023-05-17 06:57:36,780 INFO [finetune.py:992] (0/2) Epoch 16, batch 11400, loss[loss=0.2644, simple_loss=0.324, pruned_loss=0.1024, over 6905.00 frames. ], tot_loss[loss=0.2122, simple_loss=0.2963, pruned_loss=0.06401, over 1934296.73 frames. ], batch size: 98, lr: 3.36e-03, grad_scale: 8.0 2023-05-17 06:57:47,564 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=294927.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 06:58:11,670 INFO [finetune.py:992] (0/2) Epoch 16, batch 11450, loss[loss=0.2029, simple_loss=0.2929, pruned_loss=0.05643, over 10362.00 frames. ], tot_loss[loss=0.2169, simple_loss=0.2999, pruned_loss=0.06693, over 1886704.02 frames. ], batch size: 68, lr: 3.36e-03, grad_scale: 8.0 2023-05-17 06:58:25,391 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-05-17 06:58:30,637 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=294989.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 06:58:32,475 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.574e+02 3.335e+02 3.934e+02 4.570e+02 8.629e+02, threshold=7.868e+02, percent-clipped=1.0 2023-05-17 06:58:34,694 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=294995.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 06:58:39,150 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.90 vs. limit=2.0 2023-05-17 06:58:41,223 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.79 vs. limit=2.0 2023-05-17 06:58:46,051 INFO [finetune.py:992] (0/2) Epoch 16, batch 11500, loss[loss=0.2826, simple_loss=0.3405, pruned_loss=0.1124, over 6533.00 frames. ], tot_loss[loss=0.2212, simple_loss=0.3031, pruned_loss=0.06962, over 1850751.53 frames. ], batch size: 99, lr: 3.36e-03, grad_scale: 8.0 2023-05-17 06:59:08,845 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=295043.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 06:59:13,804 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=295050.0, num_to_drop=1, layers_to_drop={2} 2023-05-17 06:59:21,040 INFO [finetune.py:992] (0/2) Epoch 16, batch 11550, loss[loss=0.2532, simple_loss=0.3221, pruned_loss=0.09221, over 6987.00 frames. ], tot_loss[loss=0.2244, simple_loss=0.3055, pruned_loss=0.07166, over 1811690.21 frames. ], batch size: 98, lr: 3.36e-03, grad_scale: 8.0 2023-05-17 06:59:24,794 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.2274, 2.9220, 2.9632, 3.2145, 2.5701, 3.0155, 2.6228, 2.5255], device='cuda:0'), covar=tensor([0.1789, 0.1120, 0.0853, 0.0522, 0.1212, 0.0929, 0.1621, 0.0676], device='cuda:0'), in_proj_covar=tensor([0.0229, 0.0271, 0.0300, 0.0361, 0.0244, 0.0248, 0.0263, 0.0373], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-17 06:59:25,980 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=295068.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 06:59:34,707 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=295079.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 06:59:43,396 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.687e+02 3.365e+02 3.923e+02 4.459e+02 8.081e+02, threshold=7.845e+02, percent-clipped=1.0 2023-05-17 06:59:55,844 INFO [finetune.py:992] (0/2) Epoch 16, batch 11600, loss[loss=0.2804, simple_loss=0.3498, pruned_loss=0.1055, over 6867.00 frames. ], tot_loss[loss=0.226, simple_loss=0.3067, pruned_loss=0.07262, over 1794756.50 frames. ], batch size: 99, lr: 3.36e-03, grad_scale: 8.0 2023-05-17 07:00:06,114 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.5034, 4.4692, 4.3738, 4.0182, 4.0909, 4.4714, 4.2046, 4.0848], device='cuda:0'), covar=tensor([0.0771, 0.0910, 0.0680, 0.1491, 0.2019, 0.0874, 0.1389, 0.1103], device='cuda:0'), in_proj_covar=tensor([0.0620, 0.0551, 0.0511, 0.0618, 0.0409, 0.0708, 0.0757, 0.0558], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002], device='cuda:0') 2023-05-17 07:00:06,704 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=295127.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 07:00:08,146 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=295129.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 07:00:20,452 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=295144.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 07:00:23,838 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.72 vs. limit=2.0 2023-05-17 07:00:31,012 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.65 vs. limit=2.0 2023-05-17 07:00:32,616 INFO [finetune.py:992] (0/2) Epoch 16, batch 11650, loss[loss=0.2705, simple_loss=0.3323, pruned_loss=0.1044, over 6646.00 frames. ], tot_loss[loss=0.2265, simple_loss=0.306, pruned_loss=0.07348, over 1760270.44 frames. ], batch size: 97, lr: 3.36e-03, grad_scale: 8.0 2023-05-17 07:00:53,737 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.481e+02 3.491e+02 3.968e+02 4.748e+02 1.073e+03, threshold=7.935e+02, percent-clipped=3.0 2023-05-17 07:00:53,852 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=295192.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 07:01:08,328 INFO [finetune.py:992] (0/2) Epoch 16, batch 11700, loss[loss=0.2748, simple_loss=0.3325, pruned_loss=0.1086, over 6873.00 frames. ], tot_loss[loss=0.227, simple_loss=0.3061, pruned_loss=0.07392, over 1746371.91 frames. ], batch size: 99, lr: 3.36e-03, grad_scale: 16.0 2023-05-17 07:01:17,778 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=295225.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 07:01:19,077 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=295227.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 07:01:21,477 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=4.83 vs. limit=5.0 2023-05-17 07:01:29,856 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.0373, 2.2789, 2.3072, 2.1922, 2.0903, 2.0920, 2.1128, 1.8193], device='cuda:0'), covar=tensor([0.0331, 0.0191, 0.0212, 0.0218, 0.0312, 0.0243, 0.0236, 0.0415], device='cuda:0'), in_proj_covar=tensor([0.0197, 0.0164, 0.0168, 0.0193, 0.0204, 0.0201, 0.0177, 0.0205], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-17 07:01:36,773 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.9280, 3.0357, 4.4308, 2.4383, 2.5155, 3.4286, 3.0224, 3.4470], device='cuda:0'), covar=tensor([0.0599, 0.1242, 0.0205, 0.1370, 0.2054, 0.1409, 0.1416, 0.1095], device='cuda:0'), in_proj_covar=tensor([0.0236, 0.0236, 0.0250, 0.0181, 0.0234, 0.0287, 0.0221, 0.0259], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:0') 2023-05-17 07:01:42,644 INFO [finetune.py:992] (0/2) Epoch 16, batch 11750, loss[loss=0.2159, simple_loss=0.3014, pruned_loss=0.06525, over 10120.00 frames. ], tot_loss[loss=0.2278, simple_loss=0.3062, pruned_loss=0.07472, over 1714869.75 frames. ], batch size: 68, lr: 3.36e-03, grad_scale: 16.0 2023-05-17 07:01:52,754 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=295275.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 07:01:55,606 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=295279.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 07:02:00,347 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=295286.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 07:02:04,212 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.194e+02 3.411e+02 3.937e+02 4.615e+02 7.621e+02, threshold=7.874e+02, percent-clipped=0.0 2023-05-17 07:02:08,432 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.0112, 2.1289, 2.1585, 2.1010, 1.8782, 1.9666, 2.0007, 1.7194], device='cuda:0'), covar=tensor([0.0328, 0.0204, 0.0224, 0.0254, 0.0328, 0.0239, 0.0236, 0.0418], device='cuda:0'), in_proj_covar=tensor([0.0197, 0.0164, 0.0168, 0.0193, 0.0204, 0.0201, 0.0177, 0.0205], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-17 07:02:16,902 INFO [finetune.py:992] (0/2) Epoch 16, batch 11800, loss[loss=0.2794, simple_loss=0.3398, pruned_loss=0.1095, over 7004.00 frames. ], tot_loss[loss=0.2307, simple_loss=0.3083, pruned_loss=0.07653, over 1690196.79 frames. ], batch size: 99, lr: 3.36e-03, grad_scale: 16.0 2023-05-17 07:02:33,782 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=295333.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 07:02:38,408 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=295340.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 07:02:39,981 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.44 vs. limit=2.0 2023-05-17 07:02:41,650 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=295345.0, num_to_drop=1, layers_to_drop={0} 2023-05-17 07:02:50,377 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.7622, 3.6723, 3.7547, 3.4788, 3.6591, 3.5185, 3.7059, 3.5456], device='cuda:0'), covar=tensor([0.0557, 0.0424, 0.0483, 0.0327, 0.0459, 0.0395, 0.0468, 0.1548], device='cuda:0'), in_proj_covar=tensor([0.0260, 0.0262, 0.0282, 0.0259, 0.0260, 0.0259, 0.0235, 0.0209], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-17 07:02:51,083 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=295359.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 07:02:52,296 INFO [finetune.py:992] (0/2) Epoch 16, batch 11850, loss[loss=0.2583, simple_loss=0.3277, pruned_loss=0.09446, over 6292.00 frames. ], tot_loss[loss=0.2313, simple_loss=0.3094, pruned_loss=0.07656, over 1692512.69 frames. ], batch size: 99, lr: 3.36e-03, grad_scale: 16.0 2023-05-17 07:03:00,584 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.4250, 3.1386, 3.0452, 3.3024, 2.6831, 3.1270, 2.5668, 2.7354], device='cuda:0'), covar=tensor([0.1436, 0.0926, 0.0883, 0.0657, 0.0999, 0.0822, 0.1783, 0.0670], device='cuda:0'), in_proj_covar=tensor([0.0226, 0.0268, 0.0297, 0.0355, 0.0241, 0.0244, 0.0262, 0.0369], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-17 07:03:14,240 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.314e+02 3.456e+02 4.054e+02 4.912e+02 9.762e+02, threshold=8.107e+02, percent-clipped=2.0 2023-05-17 07:03:15,883 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=295394.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 07:03:20,699 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=5.10 vs. limit=5.0 2023-05-17 07:03:27,564 INFO [finetune.py:992] (0/2) Epoch 16, batch 11900, loss[loss=0.202, simple_loss=0.2948, pruned_loss=0.05464, over 11167.00 frames. ], tot_loss[loss=0.2295, simple_loss=0.3085, pruned_loss=0.07527, over 1682076.84 frames. ], batch size: 55, lr: 3.36e-03, grad_scale: 16.0 2023-05-17 07:03:33,734 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=295420.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 07:03:36,476 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=295424.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 07:04:02,354 INFO [finetune.py:992] (0/2) Epoch 16, batch 11950, loss[loss=0.2046, simple_loss=0.2891, pruned_loss=0.06011, over 10228.00 frames. ], tot_loss[loss=0.2253, simple_loss=0.3057, pruned_loss=0.07242, over 1678744.43 frames. ], batch size: 68, lr: 3.36e-03, grad_scale: 16.0 2023-05-17 07:04:03,878 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.9030, 2.1762, 2.8268, 2.8500, 2.8489, 2.9641, 2.8114, 2.4784], device='cuda:0'), covar=tensor([0.0102, 0.0414, 0.0203, 0.0100, 0.0171, 0.0143, 0.0167, 0.0347], device='cuda:0'), in_proj_covar=tensor([0.0089, 0.0123, 0.0103, 0.0079, 0.0103, 0.0117, 0.0100, 0.0139], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-17 07:04:19,433 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.0344, 1.9552, 2.3069, 2.0746, 2.1540, 2.3078, 1.8460, 2.2678], device='cuda:0'), covar=tensor([0.0127, 0.0376, 0.0185, 0.0228, 0.0171, 0.0182, 0.0314, 0.0160], device='cuda:0'), in_proj_covar=tensor([0.0174, 0.0199, 0.0184, 0.0182, 0.0210, 0.0162, 0.0191, 0.0181], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-17 07:04:22,162 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.7932, 3.1574, 2.3920, 2.2412, 2.7651, 2.3834, 3.0546, 2.6323], device='cuda:0'), covar=tensor([0.0668, 0.0601, 0.1134, 0.1537, 0.0317, 0.1245, 0.0510, 0.0842], device='cuda:0'), in_proj_covar=tensor([0.0184, 0.0247, 0.0172, 0.0196, 0.0139, 0.0179, 0.0192, 0.0169], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-17 07:04:23,241 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.056e+02 3.064e+02 3.708e+02 4.359e+02 1.064e+03, threshold=7.417e+02, percent-clipped=1.0 2023-05-17 07:04:32,408 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.0409, 1.9486, 2.3371, 2.0980, 2.1525, 2.3213, 1.8736, 2.2674], device='cuda:0'), covar=tensor([0.0136, 0.0365, 0.0185, 0.0232, 0.0195, 0.0183, 0.0322, 0.0166], device='cuda:0'), in_proj_covar=tensor([0.0174, 0.0200, 0.0185, 0.0182, 0.0210, 0.0162, 0.0191, 0.0182], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-17 07:04:37,510 INFO [finetune.py:992] (0/2) Epoch 16, batch 12000, loss[loss=0.1763, simple_loss=0.2655, pruned_loss=0.04351, over 7590.00 frames. ], tot_loss[loss=0.2201, simple_loss=0.3015, pruned_loss=0.06931, over 1663819.66 frames. ], batch size: 99, lr: 3.36e-03, grad_scale: 16.0 2023-05-17 07:04:37,511 INFO [finetune.py:1017] (0/2) Computing validation loss 2023-05-17 07:04:50,113 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.4693, 4.3931, 4.5464, 4.5302, 4.2622, 4.2908, 4.1337, 4.3002], device='cuda:0'), covar=tensor([0.0790, 0.0599, 0.0747, 0.0604, 0.1649, 0.1435, 0.0649, 0.1254], device='cuda:0'), in_proj_covar=tensor([0.0514, 0.0663, 0.0585, 0.0598, 0.0799, 0.0710, 0.0539, 0.0460], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0003], device='cuda:0') 2023-05-17 07:04:55,572 INFO [finetune.py:1026] (0/2) Epoch 16, validation: loss=0.2865, simple_loss=0.3618, pruned_loss=0.1056, over 1020973.00 frames. 2023-05-17 07:04:55,572 INFO [finetune.py:1027] (0/2) Maximum memory allocated so far is 12827MB 2023-05-17 07:05:08,087 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.44 vs. limit=2.0 2023-05-17 07:05:27,395 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=295556.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 07:05:28,690 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=295558.0, num_to_drop=1, layers_to_drop={1} 2023-05-17 07:05:30,588 INFO [finetune.py:992] (0/2) Epoch 16, batch 12050, loss[loss=0.1919, simple_loss=0.2913, pruned_loss=0.04623, over 12359.00 frames. ], tot_loss[loss=0.2148, simple_loss=0.2973, pruned_loss=0.06612, over 1680822.17 frames. ], batch size: 36, lr: 3.36e-03, grad_scale: 16.0 2023-05-17 07:05:43,668 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=295581.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 07:05:50,514 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.099e+02 2.927e+02 3.374e+02 4.273e+02 1.460e+03, threshold=6.749e+02, percent-clipped=2.0 2023-05-17 07:05:52,776 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=5.05 vs. limit=5.0 2023-05-17 07:06:02,835 INFO [finetune.py:992] (0/2) Epoch 16, batch 12100, loss[loss=0.2033, simple_loss=0.2824, pruned_loss=0.06208, over 6944.00 frames. ], tot_loss[loss=0.2131, simple_loss=0.2961, pruned_loss=0.06506, over 1679021.13 frames. ], batch size: 99, lr: 3.35e-03, grad_scale: 16.0 2023-05-17 07:06:06,802 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=295617.0, num_to_drop=1, layers_to_drop={1} 2023-05-17 07:06:08,027 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=295619.0, num_to_drop=1, layers_to_drop={2} 2023-05-17 07:06:14,567 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=295629.0, num_to_drop=1, layers_to_drop={0} 2023-05-17 07:06:18,190 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=295635.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 07:06:24,486 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=295645.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 07:06:35,550 INFO [finetune.py:992] (0/2) Epoch 16, batch 12150, loss[loss=0.2189, simple_loss=0.3136, pruned_loss=0.06206, over 10166.00 frames. ], tot_loss[loss=0.2143, simple_loss=0.2973, pruned_loss=0.06567, over 1686122.93 frames. ], batch size: 68, lr: 3.35e-03, grad_scale: 16.0 2023-05-17 07:06:49,157 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2023-05-17 07:06:50,812 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=295685.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 07:06:53,241 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=295689.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 07:06:54,020 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=295690.0, num_to_drop=1, layers_to_drop={2} 2023-05-17 07:06:55,025 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.036e+02 3.228e+02 3.632e+02 4.475e+02 1.263e+03, threshold=7.264e+02, percent-clipped=3.0 2023-05-17 07:06:55,837 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=295693.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 07:07:05,039 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.92 vs. limit=2.0 2023-05-17 07:07:07,168 INFO [finetune.py:992] (0/2) Epoch 16, batch 12200, loss[loss=0.2263, simple_loss=0.299, pruned_loss=0.07685, over 6747.00 frames. ], tot_loss[loss=0.2147, simple_loss=0.2976, pruned_loss=0.06594, over 1678801.99 frames. ], batch size: 98, lr: 3.35e-03, grad_scale: 16.0 2023-05-17 07:07:09,781 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=295715.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 07:07:15,362 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=295724.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 07:07:29,108 INFO [checkpoint.py:75] (0/2) Saving checkpoint to pruned_transducer_stateless7/exp_giga_finetune/epoch-16.pt 2023-05-17 07:07:51,915 INFO [finetune.py:992] (0/2) Epoch 17, batch 0, loss[loss=0.1793, simple_loss=0.2702, pruned_loss=0.04425, over 12145.00 frames. ], tot_loss[loss=0.1793, simple_loss=0.2702, pruned_loss=0.04425, over 12145.00 frames. ], batch size: 36, lr: 3.35e-03, grad_scale: 16.0 2023-05-17 07:07:51,916 INFO [finetune.py:1017] (0/2) Computing validation loss 2023-05-17 07:08:09,392 INFO [finetune.py:1026] (0/2) Epoch 17, validation: loss=0.2904, simple_loss=0.3634, pruned_loss=0.1087, over 1020973.00 frames. 2023-05-17 07:08:09,393 INFO [finetune.py:1027] (0/2) Maximum memory allocated so far is 12827MB 2023-05-17 07:08:10,296 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=295746.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 07:08:28,637 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=295772.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 07:08:38,165 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.6542, 3.6287, 3.2687, 3.2067, 2.9259, 2.8309, 3.7023, 2.4730], device='cuda:0'), covar=tensor([0.0425, 0.0170, 0.0234, 0.0253, 0.0441, 0.0365, 0.0145, 0.0544], device='cuda:0'), in_proj_covar=tensor([0.0193, 0.0159, 0.0164, 0.0189, 0.0200, 0.0196, 0.0172, 0.0201], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-17 07:08:43,012 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.786e+02 3.043e+02 3.475e+02 4.357e+02 6.788e+02, threshold=6.950e+02, percent-clipped=0.0 2023-05-17 07:08:45,101 INFO [finetune.py:992] (0/2) Epoch 17, batch 50, loss[loss=0.1618, simple_loss=0.2515, pruned_loss=0.03607, over 12150.00 frames. ], tot_loss[loss=0.1765, simple_loss=0.2666, pruned_loss=0.04317, over 530384.96 frames. ], batch size: 34, lr: 3.35e-03, grad_scale: 16.0 2023-05-17 07:09:03,710 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=295820.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 07:09:21,371 INFO [finetune.py:992] (0/2) Epoch 17, batch 100, loss[loss=0.1457, simple_loss=0.2332, pruned_loss=0.02907, over 12170.00 frames. ], tot_loss[loss=0.1725, simple_loss=0.2637, pruned_loss=0.04067, over 937297.94 frames. ], batch size: 29, lr: 3.35e-03, grad_scale: 16.0 2023-05-17 07:09:22,248 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.5601, 4.4675, 4.3030, 4.5632, 3.3439, 4.1886, 2.9399, 4.4120], device='cuda:0'), covar=tensor([0.1634, 0.0649, 0.0962, 0.0829, 0.1199, 0.0612, 0.1841, 0.1190], device='cuda:0'), in_proj_covar=tensor([0.0228, 0.0268, 0.0296, 0.0355, 0.0241, 0.0244, 0.0262, 0.0367], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-17 07:09:41,454 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=295873.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 07:09:47,782 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=295881.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 07:09:47,856 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=295881.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 07:09:52,911 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.3452, 4.7592, 3.2117, 2.9894, 4.0209, 2.6220, 4.1056, 3.2976], device='cuda:0'), covar=tensor([0.0760, 0.0649, 0.1137, 0.1508, 0.0359, 0.1454, 0.0468, 0.0881], device='cuda:0'), in_proj_covar=tensor([0.0186, 0.0249, 0.0175, 0.0198, 0.0140, 0.0181, 0.0194, 0.0171], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-17 07:09:56,378 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.060e+02 2.730e+02 3.190e+02 3.765e+02 6.955e+02, threshold=6.380e+02, percent-clipped=1.0 2023-05-17 07:09:58,564 INFO [finetune.py:992] (0/2) Epoch 17, batch 150, loss[loss=0.142, simple_loss=0.2325, pruned_loss=0.02571, over 12333.00 frames. ], tot_loss[loss=0.1714, simple_loss=0.2622, pruned_loss=0.04034, over 1251255.98 frames. ], batch size: 31, lr: 3.35e-03, grad_scale: 16.0 2023-05-17 07:10:10,890 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=295912.0, num_to_drop=1, layers_to_drop={3} 2023-05-17 07:10:12,357 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=295914.0, num_to_drop=1, layers_to_drop={2} 2023-05-17 07:10:22,922 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=295929.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 07:10:26,678 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=295934.0, num_to_drop=1, layers_to_drop={0} 2023-05-17 07:10:27,368 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=295935.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 07:10:34,448 INFO [finetune.py:992] (0/2) Epoch 17, batch 200, loss[loss=0.1806, simple_loss=0.2749, pruned_loss=0.04316, over 12037.00 frames. ], tot_loss[loss=0.1704, simple_loss=0.2609, pruned_loss=0.03993, over 1498265.81 frames. ], batch size: 40, lr: 3.35e-03, grad_scale: 16.0 2023-05-17 07:10:49,626 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.1595, 4.7626, 5.1249, 4.4996, 4.8109, 4.5490, 5.1647, 4.8530], device='cuda:0'), covar=tensor([0.0283, 0.0384, 0.0297, 0.0279, 0.0409, 0.0364, 0.0230, 0.0413], device='cuda:0'), in_proj_covar=tensor([0.0262, 0.0263, 0.0283, 0.0260, 0.0260, 0.0260, 0.0237, 0.0211], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-17 07:11:01,716 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=295983.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 07:11:03,255 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=295985.0, num_to_drop=1, layers_to_drop={1} 2023-05-17 07:11:06,046 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=295989.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 07:11:08,101 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.015e+02 2.624e+02 3.123e+02 3.965e+02 6.314e+02, threshold=6.245e+02, percent-clipped=0.0 2023-05-17 07:11:10,244 INFO [finetune.py:992] (0/2) Epoch 17, batch 250, loss[loss=0.1904, simple_loss=0.281, pruned_loss=0.04991, over 11350.00 frames. ], tot_loss[loss=0.1684, simple_loss=0.2588, pruned_loss=0.03898, over 1692600.86 frames. ], batch size: 55, lr: 3.35e-03, grad_scale: 16.0 2023-05-17 07:11:14,136 INFO [checkpoint.py:75] (0/2) Saving checkpoint to pruned_transducer_stateless7/exp_giga_finetune/checkpoint-196000.pt 2023-05-17 07:11:28,040 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=296015.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 07:11:44,378 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.8175, 2.2377, 3.3293, 2.7986, 3.1902, 3.0273, 2.3046, 3.2313], device='cuda:0'), covar=tensor([0.0174, 0.0435, 0.0210, 0.0266, 0.0192, 0.0203, 0.0429, 0.0173], device='cuda:0'), in_proj_covar=tensor([0.0176, 0.0200, 0.0187, 0.0183, 0.0212, 0.0162, 0.0194, 0.0183], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-17 07:11:44,870 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=296037.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 07:11:47,729 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=296041.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 07:11:50,535 INFO [finetune.py:992] (0/2) Epoch 17, batch 300, loss[loss=0.1805, simple_loss=0.2717, pruned_loss=0.04464, over 12207.00 frames. ], tot_loss[loss=0.1691, simple_loss=0.2591, pruned_loss=0.03956, over 1829226.97 frames. ], batch size: 35, lr: 3.35e-03, grad_scale: 16.0 2023-05-17 07:11:53,651 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=296049.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 07:11:56,538 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.3717, 3.6460, 3.7484, 4.1754, 2.8919, 3.4702, 2.4674, 3.7548], device='cuda:0'), covar=tensor([0.1744, 0.1137, 0.1225, 0.0937, 0.1440, 0.0916, 0.2329, 0.1146], device='cuda:0'), in_proj_covar=tensor([0.0231, 0.0272, 0.0299, 0.0360, 0.0244, 0.0247, 0.0265, 0.0373], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-17 07:12:01,435 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.6672, 3.7695, 3.3639, 3.2115, 3.0548, 2.9881, 3.7140, 2.4409], device='cuda:0'), covar=tensor([0.0432, 0.0121, 0.0209, 0.0254, 0.0385, 0.0325, 0.0142, 0.0582], device='cuda:0'), in_proj_covar=tensor([0.0195, 0.0161, 0.0166, 0.0191, 0.0202, 0.0198, 0.0175, 0.0203], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-17 07:12:03,473 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=296063.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 07:12:23,841 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.927e+02 2.686e+02 3.153e+02 3.747e+02 6.952e+02, threshold=6.306e+02, percent-clipped=1.0 2023-05-17 07:12:25,966 INFO [finetune.py:992] (0/2) Epoch 17, batch 350, loss[loss=0.1747, simple_loss=0.2731, pruned_loss=0.03811, over 11794.00 frames. ], tot_loss[loss=0.1687, simple_loss=0.2591, pruned_loss=0.03909, over 1956200.20 frames. ], batch size: 44, lr: 3.35e-03, grad_scale: 16.0 2023-05-17 07:12:31,119 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.0400, 5.9101, 5.5252, 5.4576, 5.9729, 5.2404, 5.5292, 5.5232], device='cuda:0'), covar=tensor([0.1650, 0.0919, 0.1120, 0.2140, 0.0974, 0.2187, 0.1953, 0.1191], device='cuda:0'), in_proj_covar=tensor([0.0361, 0.0495, 0.0399, 0.0447, 0.0466, 0.0430, 0.0398, 0.0388], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-17 07:12:36,703 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=296110.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 07:13:01,364 INFO [finetune.py:992] (0/2) Epoch 17, batch 400, loss[loss=0.1435, simple_loss=0.2309, pruned_loss=0.02803, over 12301.00 frames. ], tot_loss[loss=0.1681, simple_loss=0.2586, pruned_loss=0.03877, over 2046704.20 frames. ], batch size: 34, lr: 3.35e-03, grad_scale: 16.0 2023-05-17 07:13:24,194 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=296176.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 07:13:37,024 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.732e+02 2.556e+02 3.066e+02 3.538e+02 2.337e+03, threshold=6.132e+02, percent-clipped=1.0 2023-05-17 07:13:38,578 INFO [finetune.py:992] (0/2) Epoch 17, batch 450, loss[loss=0.1646, simple_loss=0.2659, pruned_loss=0.03165, over 12341.00 frames. ], tot_loss[loss=0.167, simple_loss=0.2574, pruned_loss=0.03835, over 2125271.18 frames. ], batch size: 36, lr: 3.35e-03, grad_scale: 8.0 2023-05-17 07:13:41,690 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.8505, 4.4486, 4.8062, 4.2005, 4.4915, 4.2636, 4.8428, 4.5409], device='cuda:0'), covar=tensor([0.0329, 0.0463, 0.0345, 0.0311, 0.0472, 0.0416, 0.0248, 0.0583], device='cuda:0'), in_proj_covar=tensor([0.0268, 0.0269, 0.0289, 0.0266, 0.0267, 0.0266, 0.0242, 0.0216], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-17 07:13:51,322 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=296212.0, num_to_drop=1, layers_to_drop={0} 2023-05-17 07:13:52,655 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=296214.0, num_to_drop=1, layers_to_drop={1} 2023-05-17 07:14:03,362 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=296229.0, num_to_drop=1, layers_to_drop={0} 2023-05-17 07:14:14,660 INFO [finetune.py:992] (0/2) Epoch 17, batch 500, loss[loss=0.1459, simple_loss=0.2277, pruned_loss=0.03207, over 12174.00 frames. ], tot_loss[loss=0.1671, simple_loss=0.2577, pruned_loss=0.03826, over 2184326.26 frames. ], batch size: 29, lr: 3.35e-03, grad_scale: 8.0 2023-05-17 07:14:17,832 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 2023-05-17 07:14:25,395 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=296260.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 07:14:26,788 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=296262.0, num_to_drop=1, layers_to_drop={1} 2023-05-17 07:14:43,379 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=296285.0, num_to_drop=1, layers_to_drop={2} 2023-05-17 07:14:48,990 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.711e+02 2.639e+02 3.014e+02 3.514e+02 6.174e+02, threshold=6.028e+02, percent-clipped=1.0 2023-05-17 07:14:50,406 INFO [finetune.py:992] (0/2) Epoch 17, batch 550, loss[loss=0.1883, simple_loss=0.2771, pruned_loss=0.04978, over 12051.00 frames. ], tot_loss[loss=0.1672, simple_loss=0.2579, pruned_loss=0.03826, over 2232524.90 frames. ], batch size: 40, lr: 3.35e-03, grad_scale: 8.0 2023-05-17 07:14:53,319 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.4031, 2.0660, 2.8109, 2.5109, 2.8814, 2.6768, 1.9576, 2.8826], device='cuda:0'), covar=tensor([0.0224, 0.0495, 0.0256, 0.0322, 0.0211, 0.0261, 0.0514, 0.0216], device='cuda:0'), in_proj_covar=tensor([0.0180, 0.0205, 0.0191, 0.0187, 0.0216, 0.0167, 0.0197, 0.0187], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-17 07:15:01,198 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.0013, 4.6767, 4.6546, 4.9215, 4.7851, 4.8585, 4.8130, 2.4322], device='cuda:0'), covar=tensor([0.0120, 0.0077, 0.0124, 0.0068, 0.0050, 0.0115, 0.0092, 0.0965], device='cuda:0'), in_proj_covar=tensor([0.0071, 0.0080, 0.0084, 0.0075, 0.0061, 0.0095, 0.0083, 0.0100], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-17 07:15:04,719 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=296315.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 07:15:18,765 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=296333.0, num_to_drop=1, layers_to_drop={0} 2023-05-17 07:15:24,346 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=296341.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 07:15:27,026 INFO [finetune.py:992] (0/2) Epoch 17, batch 600, loss[loss=0.2306, simple_loss=0.3093, pruned_loss=0.07599, over 8364.00 frames. ], tot_loss[loss=0.1677, simple_loss=0.258, pruned_loss=0.03866, over 2250188.12 frames. ], batch size: 99, lr: 3.35e-03, grad_scale: 8.0 2023-05-17 07:15:43,812 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.4393, 3.5007, 3.2470, 3.1083, 2.8623, 2.6956, 3.5328, 2.3502], device='cuda:0'), covar=tensor([0.0430, 0.0202, 0.0208, 0.0237, 0.0429, 0.0407, 0.0145, 0.0502], device='cuda:0'), in_proj_covar=tensor([0.0198, 0.0163, 0.0168, 0.0193, 0.0204, 0.0200, 0.0177, 0.0206], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-17 07:15:48,816 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.5392, 3.7168, 3.3241, 3.1795, 2.9813, 2.7884, 3.7002, 2.4418], device='cuda:0'), covar=tensor([0.0442, 0.0131, 0.0237, 0.0230, 0.0455, 0.0418, 0.0158, 0.0515], device='cuda:0'), in_proj_covar=tensor([0.0198, 0.0163, 0.0168, 0.0194, 0.0205, 0.0200, 0.0177, 0.0206], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-17 07:15:49,591 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=296376.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 07:15:51,200 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=2.93 vs. limit=5.0 2023-05-17 07:15:58,522 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=296389.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 07:16:01,312 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.730e+02 2.646e+02 2.973e+02 3.749e+02 7.703e+02, threshold=5.946e+02, percent-clipped=3.0 2023-05-17 07:16:02,802 INFO [finetune.py:992] (0/2) Epoch 17, batch 650, loss[loss=0.1741, simple_loss=0.2704, pruned_loss=0.03895, over 12312.00 frames. ], tot_loss[loss=0.1666, simple_loss=0.2569, pruned_loss=0.03821, over 2288393.08 frames. ], batch size: 34, lr: 3.35e-03, grad_scale: 8.0 2023-05-17 07:16:10,569 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=296405.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 07:16:25,770 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.0947, 4.7000, 5.0707, 4.4033, 4.7486, 4.4624, 5.1148, 4.7567], device='cuda:0'), covar=tensor([0.0306, 0.0399, 0.0309, 0.0277, 0.0411, 0.0396, 0.0206, 0.0395], device='cuda:0'), in_proj_covar=tensor([0.0267, 0.0268, 0.0288, 0.0265, 0.0266, 0.0266, 0.0241, 0.0215], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-17 07:16:31,585 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=296434.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 07:16:39,020 INFO [finetune.py:992] (0/2) Epoch 17, batch 700, loss[loss=0.1606, simple_loss=0.2544, pruned_loss=0.03346, over 12026.00 frames. ], tot_loss[loss=0.1655, simple_loss=0.2554, pruned_loss=0.03783, over 2308874.49 frames. ], batch size: 40, lr: 3.35e-03, grad_scale: 8.0 2023-05-17 07:16:51,644 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.2519, 4.0528, 4.1443, 4.3809, 2.9106, 3.8927, 2.6442, 4.1848], device='cuda:0'), covar=tensor([0.1636, 0.0778, 0.0860, 0.0732, 0.1350, 0.0665, 0.1945, 0.1185], device='cuda:0'), in_proj_covar=tensor([0.0233, 0.0274, 0.0301, 0.0364, 0.0246, 0.0250, 0.0266, 0.0377], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-17 07:17:02,152 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=296476.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 07:17:14,326 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=296492.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 07:17:14,814 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.539e+02 2.527e+02 2.946e+02 3.308e+02 5.689e+02, threshold=5.892e+02, percent-clipped=0.0 2023-05-17 07:17:16,315 INFO [finetune.py:992] (0/2) Epoch 17, batch 750, loss[loss=0.171, simple_loss=0.2677, pruned_loss=0.03716, over 12045.00 frames. ], tot_loss[loss=0.1643, simple_loss=0.2538, pruned_loss=0.03743, over 2329123.84 frames. ], batch size: 37, lr: 3.35e-03, grad_scale: 8.0 2023-05-17 07:17:16,568 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=296495.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 07:17:17,959 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=296497.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 07:17:24,421 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.3629, 4.9342, 5.3140, 4.6111, 4.9779, 4.7184, 5.3785, 5.0191], device='cuda:0'), covar=tensor([0.0278, 0.0369, 0.0306, 0.0277, 0.0415, 0.0366, 0.0220, 0.0309], device='cuda:0'), in_proj_covar=tensor([0.0267, 0.0269, 0.0290, 0.0266, 0.0267, 0.0267, 0.0242, 0.0216], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-17 07:17:30,213 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=296514.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 07:17:37,191 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=296524.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 07:17:40,859 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=296529.0, num_to_drop=1, layers_to_drop={0} 2023-05-17 07:17:42,558 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=4.01 vs. limit=5.0 2023-05-17 07:17:52,171 INFO [finetune.py:992] (0/2) Epoch 17, batch 800, loss[loss=0.1824, simple_loss=0.2769, pruned_loss=0.04396, over 12154.00 frames. ], tot_loss[loss=0.1645, simple_loss=0.2541, pruned_loss=0.0375, over 2347553.74 frames. ], batch size: 34, lr: 3.35e-03, grad_scale: 8.0 2023-05-17 07:17:58,023 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=296553.0, num_to_drop=1, layers_to_drop={2} 2023-05-17 07:18:01,445 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=296558.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 07:18:13,049 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.68 vs. limit=2.0 2023-05-17 07:18:13,585 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=296575.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 07:18:14,877 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=296577.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 07:18:26,282 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.770e+02 2.820e+02 3.244e+02 3.783e+02 6.620e+02, threshold=6.487e+02, percent-clipped=2.0 2023-05-17 07:18:27,670 INFO [finetune.py:992] (0/2) Epoch 17, batch 850, loss[loss=0.1487, simple_loss=0.245, pruned_loss=0.02621, over 12199.00 frames. ], tot_loss[loss=0.1654, simple_loss=0.2551, pruned_loss=0.03785, over 2332748.91 frames. ], batch size: 35, lr: 3.35e-03, grad_scale: 8.0 2023-05-17 07:18:41,352 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=296613.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 07:18:43,535 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.5194, 2.6580, 3.7949, 4.5107, 3.8572, 4.4659, 3.8496, 3.2647], device='cuda:0'), covar=tensor([0.0038, 0.0384, 0.0128, 0.0040, 0.0116, 0.0082, 0.0136, 0.0350], device='cuda:0'), in_proj_covar=tensor([0.0089, 0.0123, 0.0102, 0.0079, 0.0103, 0.0116, 0.0100, 0.0139], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-17 07:19:04,460 INFO [finetune.py:992] (0/2) Epoch 17, batch 900, loss[loss=0.1786, simple_loss=0.2707, pruned_loss=0.0432, over 10705.00 frames. ], tot_loss[loss=0.1651, simple_loss=0.2552, pruned_loss=0.03745, over 2342281.68 frames. ], batch size: 68, lr: 3.35e-03, grad_scale: 8.0 2023-05-17 07:19:23,079 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=296671.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 07:19:25,440 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=296674.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 07:19:35,354 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=296688.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 07:19:38,589 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.675e+02 2.605e+02 2.984e+02 3.568e+02 6.116e+02, threshold=5.969e+02, percent-clipped=0.0 2023-05-17 07:19:40,042 INFO [finetune.py:992] (0/2) Epoch 17, batch 950, loss[loss=0.1598, simple_loss=0.2509, pruned_loss=0.03433, over 12078.00 frames. ], tot_loss[loss=0.164, simple_loss=0.2541, pruned_loss=0.03695, over 2355376.32 frames. ], batch size: 32, lr: 3.34e-03, grad_scale: 8.0 2023-05-17 07:19:42,211 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.0983, 4.9481, 4.8878, 4.9452, 4.5712, 5.1125, 4.9918, 5.2102], device='cuda:0'), covar=tensor([0.0220, 0.0153, 0.0191, 0.0322, 0.0790, 0.0272, 0.0173, 0.0172], device='cuda:0'), in_proj_covar=tensor([0.0199, 0.0198, 0.0193, 0.0247, 0.0241, 0.0221, 0.0177, 0.0233], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003, 0.0005], device='cuda:0') 2023-05-17 07:19:47,189 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=296705.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 07:20:10,365 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=2.66 vs. limit=5.0 2023-05-17 07:20:15,299 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.5932, 2.7015, 3.3262, 4.3500, 2.3799, 4.4521, 4.5470, 4.5612], device='cuda:0'), covar=tensor([0.0128, 0.1353, 0.0527, 0.0189, 0.1522, 0.0246, 0.0164, 0.0114], device='cuda:0'), in_proj_covar=tensor([0.0124, 0.0209, 0.0186, 0.0124, 0.0194, 0.0181, 0.0178, 0.0127], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-17 07:20:15,808 INFO [finetune.py:992] (0/2) Epoch 17, batch 1000, loss[loss=0.1714, simple_loss=0.2631, pruned_loss=0.03987, over 12110.00 frames. ], tot_loss[loss=0.1634, simple_loss=0.2536, pruned_loss=0.03662, over 2370635.27 frames. ], batch size: 32, lr: 3.34e-03, grad_scale: 8.0 2023-05-17 07:20:18,969 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=296749.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 07:20:21,764 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=296753.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 07:20:45,960 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=2.70 vs. limit=5.0 2023-05-17 07:20:49,286 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=296790.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 07:20:50,778 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.4849, 2.5875, 3.7500, 4.4724, 3.8862, 4.3749, 3.8517, 3.2778], device='cuda:0'), covar=tensor([0.0041, 0.0403, 0.0147, 0.0045, 0.0113, 0.0096, 0.0142, 0.0352], device='cuda:0'), in_proj_covar=tensor([0.0089, 0.0123, 0.0103, 0.0079, 0.0103, 0.0116, 0.0100, 0.0139], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-17 07:20:51,277 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.891e+02 2.591e+02 3.027e+02 3.502e+02 6.154e+02, threshold=6.055e+02, percent-clipped=2.0 2023-05-17 07:20:52,715 INFO [finetune.py:992] (0/2) Epoch 17, batch 1050, loss[loss=0.1616, simple_loss=0.256, pruned_loss=0.03358, over 11719.00 frames. ], tot_loss[loss=0.1634, simple_loss=0.2538, pruned_loss=0.03649, over 2374220.34 frames. ], batch size: 48, lr: 3.34e-03, grad_scale: 8.0 2023-05-17 07:21:04,795 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.4375, 2.9238, 3.6868, 2.3033, 2.6099, 3.0473, 2.8365, 3.1440], device='cuda:0'), covar=tensor([0.0580, 0.1199, 0.0444, 0.1382, 0.1868, 0.1602, 0.1365, 0.1209], device='cuda:0'), in_proj_covar=tensor([0.0241, 0.0240, 0.0259, 0.0187, 0.0242, 0.0296, 0.0228, 0.0266], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-17 07:21:10,454 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.73 vs. limit=2.0 2023-05-17 07:21:24,194 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.1203, 3.9272, 2.6747, 2.3554, 3.3734, 2.4482, 3.5692, 2.8976], device='cuda:0'), covar=tensor([0.0781, 0.0663, 0.1163, 0.1764, 0.0453, 0.1482, 0.0605, 0.0905], device='cuda:0'), in_proj_covar=tensor([0.0191, 0.0259, 0.0178, 0.0203, 0.0144, 0.0186, 0.0201, 0.0176], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-17 07:21:29,013 INFO [finetune.py:992] (0/2) Epoch 17, batch 1100, loss[loss=0.1744, simple_loss=0.2593, pruned_loss=0.04472, over 12118.00 frames. ], tot_loss[loss=0.1635, simple_loss=0.254, pruned_loss=0.03651, over 2365390.13 frames. ], batch size: 33, lr: 3.34e-03, grad_scale: 8.0 2023-05-17 07:21:31,217 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=296848.0, num_to_drop=1, layers_to_drop={1} 2023-05-17 07:21:34,869 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=296853.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 07:21:47,047 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=296870.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 07:21:49,351 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=296873.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 07:22:03,209 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.842e+02 2.567e+02 3.061e+02 3.490e+02 7.482e+02, threshold=6.122e+02, percent-clipped=1.0 2023-05-17 07:22:04,685 INFO [finetune.py:992] (0/2) Epoch 17, batch 1150, loss[loss=0.1924, simple_loss=0.2841, pruned_loss=0.05033, over 12008.00 frames. ], tot_loss[loss=0.1628, simple_loss=0.2529, pruned_loss=0.03634, over 2375146.98 frames. ], batch size: 40, lr: 3.34e-03, grad_scale: 8.0 2023-05-17 07:22:34,226 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=296934.0, num_to_drop=1, layers_to_drop={2} 2023-05-17 07:22:37,012 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=296938.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 07:22:41,566 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=2.94 vs. limit=5.0 2023-05-17 07:22:41,897 INFO [finetune.py:992] (0/2) Epoch 17, batch 1200, loss[loss=0.1658, simple_loss=0.2554, pruned_loss=0.0381, over 11252.00 frames. ], tot_loss[loss=0.1631, simple_loss=0.2533, pruned_loss=0.03651, over 2372834.70 frames. ], batch size: 55, lr: 3.34e-03, grad_scale: 8.0 2023-05-17 07:22:51,410 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.1971, 4.4463, 2.8358, 2.5551, 3.7719, 2.5963, 3.8151, 3.1868], device='cuda:0'), covar=tensor([0.0756, 0.0600, 0.1217, 0.1541, 0.0338, 0.1377, 0.0548, 0.0814], device='cuda:0'), in_proj_covar=tensor([0.0191, 0.0258, 0.0178, 0.0203, 0.0143, 0.0185, 0.0201, 0.0176], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-17 07:22:58,458 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.2055, 2.9313, 2.8115, 2.7199, 2.4841, 2.4148, 2.8247, 1.9559], device='cuda:0'), covar=tensor([0.0460, 0.0220, 0.0261, 0.0276, 0.0487, 0.0392, 0.0253, 0.0532], device='cuda:0'), in_proj_covar=tensor([0.0198, 0.0163, 0.0169, 0.0193, 0.0204, 0.0202, 0.0178, 0.0206], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-17 07:22:59,094 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=296969.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 07:23:00,516 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=296971.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 07:23:09,412 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.55 vs. limit=2.0 2023-05-17 07:23:11,124 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.9797, 4.8513, 4.9263, 5.0037, 4.6315, 4.6500, 4.4020, 4.9538], device='cuda:0'), covar=tensor([0.0846, 0.0727, 0.1223, 0.0673, 0.2175, 0.1511, 0.0708, 0.1084], device='cuda:0'), in_proj_covar=tensor([0.0548, 0.0709, 0.0623, 0.0636, 0.0852, 0.0760, 0.0570, 0.0490], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0005, 0.0004, 0.0003, 0.0003], device='cuda:0') 2023-05-17 07:23:15,910 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.950e+02 2.560e+02 3.113e+02 3.575e+02 6.952e+02, threshold=6.225e+02, percent-clipped=1.0 2023-05-17 07:23:17,376 INFO [finetune.py:992] (0/2) Epoch 17, batch 1250, loss[loss=0.1479, simple_loss=0.2345, pruned_loss=0.03064, over 12096.00 frames. ], tot_loss[loss=0.1632, simple_loss=0.2533, pruned_loss=0.03653, over 2365860.60 frames. ], batch size: 32, lr: 3.34e-03, grad_scale: 8.0 2023-05-17 07:23:20,573 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=296999.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 07:23:34,839 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=297019.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 07:23:52,167 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.3719, 4.0392, 4.0208, 4.2075, 4.1337, 4.3003, 4.2722, 2.3531], device='cuda:0'), covar=tensor([0.0115, 0.0104, 0.0150, 0.0081, 0.0068, 0.0118, 0.0094, 0.1030], device='cuda:0'), in_proj_covar=tensor([0.0072, 0.0081, 0.0085, 0.0075, 0.0062, 0.0095, 0.0083, 0.0101], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-17 07:23:52,752 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=297044.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 07:23:53,381 INFO [finetune.py:992] (0/2) Epoch 17, batch 1300, loss[loss=0.1635, simple_loss=0.2549, pruned_loss=0.03601, over 12044.00 frames. ], tot_loss[loss=0.1632, simple_loss=0.2532, pruned_loss=0.03655, over 2373310.82 frames. ], batch size: 42, lr: 3.34e-03, grad_scale: 8.0 2023-05-17 07:24:26,812 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=297090.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 07:24:28,824 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.641e+02 2.588e+02 2.952e+02 3.635e+02 8.207e+02, threshold=5.903e+02, percent-clipped=3.0 2023-05-17 07:24:30,308 INFO [finetune.py:992] (0/2) Epoch 17, batch 1350, loss[loss=0.1303, simple_loss=0.2164, pruned_loss=0.02209, over 12254.00 frames. ], tot_loss[loss=0.1633, simple_loss=0.2536, pruned_loss=0.03653, over 2379728.84 frames. ], batch size: 28, lr: 3.34e-03, grad_scale: 8.0 2023-05-17 07:25:00,370 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=297138.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 07:25:03,774 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.91 vs. limit=2.0 2023-05-17 07:25:05,295 INFO [finetune.py:992] (0/2) Epoch 17, batch 1400, loss[loss=0.1692, simple_loss=0.2584, pruned_loss=0.04006, over 12189.00 frames. ], tot_loss[loss=0.1648, simple_loss=0.255, pruned_loss=0.0373, over 2370343.04 frames. ], batch size: 35, lr: 3.34e-03, grad_scale: 8.0 2023-05-17 07:25:07,545 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=297148.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 07:25:11,257 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=297153.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 07:25:23,415 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=297170.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 07:25:37,002 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.7958, 4.2485, 3.6338, 4.4636, 4.0274, 2.8075, 3.7975, 2.9070], device='cuda:0'), covar=tensor([0.0999, 0.0865, 0.1575, 0.0582, 0.1302, 0.1781, 0.1206, 0.3377], device='cuda:0'), in_proj_covar=tensor([0.0310, 0.0380, 0.0362, 0.0328, 0.0373, 0.0274, 0.0348, 0.0369], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-17 07:25:39,577 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.070e+02 2.670e+02 3.053e+02 3.744e+02 7.369e+02, threshold=6.105e+02, percent-clipped=2.0 2023-05-17 07:25:40,991 INFO [finetune.py:992] (0/2) Epoch 17, batch 1450, loss[loss=0.1537, simple_loss=0.2419, pruned_loss=0.03281, over 12340.00 frames. ], tot_loss[loss=0.1646, simple_loss=0.2548, pruned_loss=0.03713, over 2373909.68 frames. ], batch size: 31, lr: 3.34e-03, grad_scale: 8.0 2023-05-17 07:25:41,762 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=297196.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 07:25:45,706 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=297201.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 07:25:59,046 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=297218.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 07:26:01,872 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=297222.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 07:26:06,839 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=297229.0, num_to_drop=1, layers_to_drop={3} 2023-05-17 07:26:18,053 INFO [finetune.py:992] (0/2) Epoch 17, batch 1500, loss[loss=0.1614, simple_loss=0.2513, pruned_loss=0.0357, over 12032.00 frames. ], tot_loss[loss=0.1652, simple_loss=0.2555, pruned_loss=0.0374, over 2369073.78 frames. ], batch size: 40, lr: 3.34e-03, grad_scale: 8.0 2023-05-17 07:26:35,570 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=297269.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 07:26:40,082 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.7022, 2.8072, 4.6669, 4.7780, 2.8473, 2.5819, 3.0007, 2.2137], device='cuda:0'), covar=tensor([0.1768, 0.3203, 0.0420, 0.0443, 0.1370, 0.2845, 0.2883, 0.4301], device='cuda:0'), in_proj_covar=tensor([0.0306, 0.0395, 0.0274, 0.0301, 0.0277, 0.0318, 0.0398, 0.0383], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-17 07:26:45,580 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=297283.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 07:26:52,591 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.666e+02 2.625e+02 2.986e+02 3.551e+02 5.451e+02, threshold=5.973e+02, percent-clipped=0.0 2023-05-17 07:26:53,388 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=297294.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 07:26:54,033 INFO [finetune.py:992] (0/2) Epoch 17, batch 1550, loss[loss=0.1751, simple_loss=0.2662, pruned_loss=0.04207, over 11818.00 frames. ], tot_loss[loss=0.1647, simple_loss=0.2551, pruned_loss=0.03717, over 2371219.99 frames. ], batch size: 44, lr: 3.34e-03, grad_scale: 8.0 2023-05-17 07:26:59,842 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.0914, 4.9359, 4.8754, 4.9293, 4.6698, 5.1039, 5.0988, 5.2445], device='cuda:0'), covar=tensor([0.0230, 0.0166, 0.0193, 0.0301, 0.0695, 0.0234, 0.0157, 0.0197], device='cuda:0'), in_proj_covar=tensor([0.0204, 0.0202, 0.0196, 0.0249, 0.0245, 0.0225, 0.0180, 0.0237], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0005, 0.0004, 0.0004, 0.0003, 0.0005], device='cuda:0') 2023-05-17 07:27:09,874 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=297317.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 07:27:26,337 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.9135, 4.5649, 4.8988, 4.3397, 4.6154, 4.4102, 4.9042, 4.5562], device='cuda:0'), covar=tensor([0.0340, 0.0433, 0.0347, 0.0302, 0.0480, 0.0349, 0.0236, 0.0532], device='cuda:0'), in_proj_covar=tensor([0.0278, 0.0279, 0.0301, 0.0276, 0.0277, 0.0277, 0.0253, 0.0225], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-17 07:27:29,156 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=297344.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 07:27:29,714 INFO [finetune.py:992] (0/2) Epoch 17, batch 1600, loss[loss=0.1657, simple_loss=0.2583, pruned_loss=0.03651, over 10494.00 frames. ], tot_loss[loss=0.1648, simple_loss=0.255, pruned_loss=0.03726, over 2377136.62 frames. ], batch size: 68, lr: 3.34e-03, grad_scale: 8.0 2023-05-17 07:28:04,097 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.9961, 3.1608, 4.4415, 2.4035, 2.6722, 3.3575, 2.9183, 3.4427], device='cuda:0'), covar=tensor([0.0607, 0.1193, 0.0396, 0.1279, 0.1944, 0.1415, 0.1454, 0.1247], device='cuda:0'), in_proj_covar=tensor([0.0241, 0.0240, 0.0259, 0.0187, 0.0241, 0.0296, 0.0228, 0.0267], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-17 07:28:04,621 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=297392.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 07:28:05,255 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.500e+02 2.563e+02 3.041e+02 3.694e+02 7.011e+02, threshold=6.081e+02, percent-clipped=1.0 2023-05-17 07:28:06,710 INFO [finetune.py:992] (0/2) Epoch 17, batch 1650, loss[loss=0.1729, simple_loss=0.2588, pruned_loss=0.04347, over 12068.00 frames. ], tot_loss[loss=0.1642, simple_loss=0.2545, pruned_loss=0.03691, over 2374228.54 frames. ], batch size: 37, lr: 3.34e-03, grad_scale: 8.0 2023-05-17 07:28:22,971 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.5712, 2.4899, 3.1438, 4.4354, 2.3123, 4.4031, 4.5667, 4.5828], device='cuda:0'), covar=tensor([0.0144, 0.1530, 0.0595, 0.0164, 0.1542, 0.0231, 0.0162, 0.0126], device='cuda:0'), in_proj_covar=tensor([0.0124, 0.0208, 0.0186, 0.0123, 0.0193, 0.0181, 0.0177, 0.0126], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-17 07:28:28,711 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.4548, 4.9371, 4.2866, 5.0411, 4.6157, 3.0930, 4.2898, 3.1179], device='cuda:0'), covar=tensor([0.0814, 0.0687, 0.1384, 0.0529, 0.1211, 0.1676, 0.1197, 0.3559], device='cuda:0'), in_proj_covar=tensor([0.0311, 0.0382, 0.0363, 0.0330, 0.0374, 0.0276, 0.0350, 0.0371], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-17 07:28:42,802 INFO [finetune.py:992] (0/2) Epoch 17, batch 1700, loss[loss=0.1759, simple_loss=0.2698, pruned_loss=0.04098, over 10449.00 frames. ], tot_loss[loss=0.1643, simple_loss=0.2547, pruned_loss=0.03696, over 2377970.36 frames. ], batch size: 68, lr: 3.34e-03, grad_scale: 8.0 2023-05-17 07:29:16,885 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.931e+02 2.569e+02 2.942e+02 3.457e+02 5.618e+02, threshold=5.884e+02, percent-clipped=0.0 2023-05-17 07:29:18,386 INFO [finetune.py:992] (0/2) Epoch 17, batch 1750, loss[loss=0.1422, simple_loss=0.2338, pruned_loss=0.02529, over 12348.00 frames. ], tot_loss[loss=0.1633, simple_loss=0.2537, pruned_loss=0.03644, over 2380660.57 frames. ], batch size: 31, lr: 3.34e-03, grad_scale: 8.0 2023-05-17 07:29:31,186 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.6935, 5.6880, 5.4285, 4.9932, 4.9788, 5.5793, 5.2336, 5.0287], device='cuda:0'), covar=tensor([0.0718, 0.0888, 0.0687, 0.1775, 0.0879, 0.0840, 0.1592, 0.1136], device='cuda:0'), in_proj_covar=tensor([0.0643, 0.0577, 0.0532, 0.0648, 0.0427, 0.0739, 0.0792, 0.0583], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-05-17 07:29:34,493 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.44 vs. limit=2.0 2023-05-17 07:29:43,700 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=297529.0, num_to_drop=1, layers_to_drop={1} 2023-05-17 07:29:54,855 INFO [finetune.py:992] (0/2) Epoch 17, batch 1800, loss[loss=0.1613, simple_loss=0.2584, pruned_loss=0.03214, over 12267.00 frames. ], tot_loss[loss=0.1629, simple_loss=0.2532, pruned_loss=0.03624, over 2375513.36 frames. ], batch size: 32, lr: 3.34e-03, grad_scale: 8.0 2023-05-17 07:30:12,983 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.4018, 5.2247, 5.3493, 5.3930, 4.9865, 5.0531, 4.8396, 5.2611], device='cuda:0'), covar=tensor([0.0627, 0.0628, 0.0806, 0.0568, 0.2081, 0.1410, 0.0559, 0.1133], device='cuda:0'), in_proj_covar=tensor([0.0549, 0.0708, 0.0624, 0.0635, 0.0857, 0.0762, 0.0568, 0.0488], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0005, 0.0004, 0.0003, 0.0003], device='cuda:0') 2023-05-17 07:30:17,833 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=297577.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 07:30:18,601 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=297578.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 07:30:29,047 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.780e+02 2.682e+02 3.139e+02 3.795e+02 6.234e+02, threshold=6.278e+02, percent-clipped=1.0 2023-05-17 07:30:29,877 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=297594.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 07:30:30,469 INFO [finetune.py:992] (0/2) Epoch 17, batch 1850, loss[loss=0.2158, simple_loss=0.3, pruned_loss=0.06575, over 8227.00 frames. ], tot_loss[loss=0.1635, simple_loss=0.2537, pruned_loss=0.03666, over 2379773.56 frames. ], batch size: 98, lr: 3.34e-03, grad_scale: 8.0 2023-05-17 07:31:04,629 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=297642.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 07:31:06,619 INFO [finetune.py:992] (0/2) Epoch 17, batch 1900, loss[loss=0.1581, simple_loss=0.2532, pruned_loss=0.0315, over 12342.00 frames. ], tot_loss[loss=0.1631, simple_loss=0.253, pruned_loss=0.03655, over 2383514.58 frames. ], batch size: 36, lr: 3.34e-03, grad_scale: 8.0 2023-05-17 07:31:42,145 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.978e+02 2.617e+02 2.974e+02 3.412e+02 6.785e+02, threshold=5.949e+02, percent-clipped=1.0 2023-05-17 07:31:43,104 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.5706, 2.7182, 3.5101, 4.6131, 3.8815, 4.4987, 3.7784, 3.3986], device='cuda:0'), covar=tensor([0.0038, 0.0383, 0.0187, 0.0035, 0.0129, 0.0085, 0.0165, 0.0339], device='cuda:0'), in_proj_covar=tensor([0.0091, 0.0125, 0.0107, 0.0081, 0.0106, 0.0119, 0.0104, 0.0142], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-17 07:31:43,630 INFO [finetune.py:992] (0/2) Epoch 17, batch 1950, loss[loss=0.1681, simple_loss=0.2524, pruned_loss=0.04191, over 12299.00 frames. ], tot_loss[loss=0.1635, simple_loss=0.2532, pruned_loss=0.0369, over 2383583.17 frames. ], batch size: 33, lr: 3.34e-03, grad_scale: 8.0 2023-05-17 07:32:19,265 INFO [finetune.py:992] (0/2) Epoch 17, batch 2000, loss[loss=0.1717, simple_loss=0.2651, pruned_loss=0.03918, over 12307.00 frames. ], tot_loss[loss=0.1634, simple_loss=0.2536, pruned_loss=0.03664, over 2391342.09 frames. ], batch size: 34, lr: 3.34e-03, grad_scale: 8.0 2023-05-17 07:32:38,909 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.5909, 3.3849, 5.1392, 2.7064, 2.7611, 3.8701, 2.9863, 3.7710], device='cuda:0'), covar=tensor([0.0540, 0.1176, 0.0333, 0.1259, 0.2073, 0.1414, 0.1643, 0.1289], device='cuda:0'), in_proj_covar=tensor([0.0241, 0.0239, 0.0259, 0.0187, 0.0241, 0.0297, 0.0228, 0.0267], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-17 07:32:53,426 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.882e+02 2.509e+02 3.015e+02 3.436e+02 6.650e+02, threshold=6.030e+02, percent-clipped=2.0 2023-05-17 07:32:54,841 INFO [finetune.py:992] (0/2) Epoch 17, batch 2050, loss[loss=0.1796, simple_loss=0.269, pruned_loss=0.04513, over 12044.00 frames. ], tot_loss[loss=0.1625, simple_loss=0.2527, pruned_loss=0.03616, over 2396977.90 frames. ], batch size: 40, lr: 3.34e-03, grad_scale: 8.0 2023-05-17 07:32:56,365 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.5703, 5.0807, 5.4990, 4.8186, 5.1636, 4.9025, 5.5697, 5.1845], device='cuda:0'), covar=tensor([0.0276, 0.0411, 0.0284, 0.0262, 0.0409, 0.0352, 0.0203, 0.0270], device='cuda:0'), in_proj_covar=tensor([0.0276, 0.0280, 0.0302, 0.0276, 0.0277, 0.0276, 0.0251, 0.0224], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-17 07:33:26,673 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=3.40 vs. limit=5.0 2023-05-17 07:33:31,900 INFO [finetune.py:992] (0/2) Epoch 17, batch 2100, loss[loss=0.1805, simple_loss=0.2747, pruned_loss=0.04316, over 11187.00 frames. ], tot_loss[loss=0.1632, simple_loss=0.2532, pruned_loss=0.03664, over 2393916.57 frames. ], batch size: 55, lr: 3.34e-03, grad_scale: 8.0 2023-05-17 07:33:55,297 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=297878.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 07:34:05,846 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.665e+02 2.706e+02 3.127e+02 3.743e+02 8.710e+02, threshold=6.254e+02, percent-clipped=2.0 2023-05-17 07:34:07,291 INFO [finetune.py:992] (0/2) Epoch 17, batch 2150, loss[loss=0.1758, simple_loss=0.2679, pruned_loss=0.04188, over 12094.00 frames. ], tot_loss[loss=0.1627, simple_loss=0.2525, pruned_loss=0.03646, over 2398622.45 frames. ], batch size: 32, lr: 3.34e-03, grad_scale: 8.0 2023-05-17 07:34:09,738 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-05-17 07:34:29,305 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=297926.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 07:34:42,804 INFO [finetune.py:992] (0/2) Epoch 17, batch 2200, loss[loss=0.1721, simple_loss=0.2633, pruned_loss=0.04043, over 12351.00 frames. ], tot_loss[loss=0.162, simple_loss=0.2518, pruned_loss=0.03607, over 2391230.11 frames. ], batch size: 36, lr: 3.34e-03, grad_scale: 8.0 2023-05-17 07:34:52,226 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=297957.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 07:35:08,056 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=4.63 vs. limit=5.0 2023-05-17 07:35:17,966 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.713e+02 2.570e+02 3.028e+02 3.621e+02 1.131e+03, threshold=6.056e+02, percent-clipped=3.0 2023-05-17 07:35:19,383 INFO [finetune.py:992] (0/2) Epoch 17, batch 2250, loss[loss=0.1538, simple_loss=0.253, pruned_loss=0.02729, over 12283.00 frames. ], tot_loss[loss=0.1618, simple_loss=0.2514, pruned_loss=0.03608, over 2385612.06 frames. ], batch size: 34, lr: 3.34e-03, grad_scale: 8.0 2023-05-17 07:35:23,243 INFO [checkpoint.py:75] (0/2) Saving checkpoint to pruned_transducer_stateless7/exp_giga_finetune/checkpoint-198000.pt 2023-05-17 07:35:39,278 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=298018.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 07:35:58,215 INFO [finetune.py:992] (0/2) Epoch 17, batch 2300, loss[loss=0.1748, simple_loss=0.2654, pruned_loss=0.04213, over 12124.00 frames. ], tot_loss[loss=0.1626, simple_loss=0.2517, pruned_loss=0.03672, over 2387520.09 frames. ], batch size: 38, lr: 3.34e-03, grad_scale: 8.0 2023-05-17 07:36:32,015 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.871e+02 2.758e+02 3.155e+02 4.025e+02 7.996e+02, threshold=6.311e+02, percent-clipped=3.0 2023-05-17 07:36:33,438 INFO [finetune.py:992] (0/2) Epoch 17, batch 2350, loss[loss=0.1625, simple_loss=0.2551, pruned_loss=0.03495, over 12194.00 frames. ], tot_loss[loss=0.1625, simple_loss=0.2517, pruned_loss=0.03665, over 2386109.38 frames. ], batch size: 35, lr: 3.34e-03, grad_scale: 8.0 2023-05-17 07:37:10,634 INFO [finetune.py:992] (0/2) Epoch 17, batch 2400, loss[loss=0.15, simple_loss=0.2378, pruned_loss=0.03112, over 12169.00 frames. ], tot_loss[loss=0.163, simple_loss=0.2526, pruned_loss=0.03672, over 2380196.12 frames. ], batch size: 29, lr: 3.34e-03, grad_scale: 8.0 2023-05-17 07:37:17,487 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.1032, 4.7086, 4.9282, 4.9423, 4.7149, 4.9817, 4.9029, 2.5572], device='cuda:0'), covar=tensor([0.0130, 0.0072, 0.0079, 0.0055, 0.0051, 0.0098, 0.0079, 0.0885], device='cuda:0'), in_proj_covar=tensor([0.0072, 0.0082, 0.0086, 0.0077, 0.0063, 0.0097, 0.0085, 0.0102], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-17 07:37:44,952 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.897e+02 2.507e+02 3.015e+02 3.482e+02 1.547e+03, threshold=6.029e+02, percent-clipped=2.0 2023-05-17 07:37:46,360 INFO [finetune.py:992] (0/2) Epoch 17, batch 2450, loss[loss=0.1638, simple_loss=0.2574, pruned_loss=0.03508, over 12108.00 frames. ], tot_loss[loss=0.1629, simple_loss=0.2527, pruned_loss=0.03655, over 2392136.65 frames. ], batch size: 33, lr: 3.33e-03, grad_scale: 16.0 2023-05-17 07:38:06,981 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=4.74 vs. limit=5.0 2023-05-17 07:38:13,392 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=3.25 vs. limit=5.0 2023-05-17 07:38:22,193 INFO [finetune.py:992] (0/2) Epoch 17, batch 2500, loss[loss=0.176, simple_loss=0.2708, pruned_loss=0.04061, over 12152.00 frames. ], tot_loss[loss=0.1633, simple_loss=0.2533, pruned_loss=0.03668, over 2394093.93 frames. ], batch size: 38, lr: 3.33e-03, grad_scale: 16.0 2023-05-17 07:38:41,940 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.5729, 2.7844, 3.2977, 4.4846, 2.4983, 4.5190, 4.5937, 4.5589], device='cuda:0'), covar=tensor([0.0169, 0.1231, 0.0506, 0.0183, 0.1363, 0.0223, 0.0148, 0.0132], device='cuda:0'), in_proj_covar=tensor([0.0124, 0.0206, 0.0185, 0.0123, 0.0191, 0.0180, 0.0177, 0.0126], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-17 07:38:57,556 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.904e+02 2.660e+02 3.049e+02 3.576e+02 5.574e+02, threshold=6.099e+02, percent-clipped=0.0 2023-05-17 07:38:59,006 INFO [finetune.py:992] (0/2) Epoch 17, batch 2550, loss[loss=0.1527, simple_loss=0.2461, pruned_loss=0.02961, over 12293.00 frames. ], tot_loss[loss=0.163, simple_loss=0.2531, pruned_loss=0.03645, over 2392234.24 frames. ], batch size: 33, lr: 3.33e-03, grad_scale: 16.0 2023-05-17 07:39:11,979 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=298313.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 07:39:17,737 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.8690, 2.1852, 3.6082, 2.8998, 3.4785, 3.0351, 2.3154, 3.4544], device='cuda:0'), covar=tensor([0.0215, 0.0585, 0.0211, 0.0369, 0.0201, 0.0240, 0.0542, 0.0213], device='cuda:0'), in_proj_covar=tensor([0.0189, 0.0215, 0.0201, 0.0197, 0.0228, 0.0174, 0.0206, 0.0199], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-17 07:39:28,408 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.8748, 5.8580, 5.6384, 5.1781, 5.1502, 5.7518, 5.3991, 5.1721], device='cuda:0'), covar=tensor([0.0801, 0.0952, 0.0732, 0.1739, 0.0781, 0.0820, 0.1649, 0.1112], device='cuda:0'), in_proj_covar=tensor([0.0644, 0.0574, 0.0531, 0.0648, 0.0428, 0.0742, 0.0794, 0.0580], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002], device='cuda:0') 2023-05-17 07:39:30,842 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.54 vs. limit=2.0 2023-05-17 07:39:34,552 INFO [finetune.py:992] (0/2) Epoch 17, batch 2600, loss[loss=0.1761, simple_loss=0.2684, pruned_loss=0.04186, over 11605.00 frames. ], tot_loss[loss=0.1627, simple_loss=0.2528, pruned_loss=0.03631, over 2388911.15 frames. ], batch size: 48, lr: 3.33e-03, grad_scale: 16.0 2023-05-17 07:39:50,324 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.4227, 3.6358, 3.3052, 3.2454, 3.0851, 2.8040, 3.6263, 2.3553], device='cuda:0'), covar=tensor([0.0449, 0.0126, 0.0210, 0.0197, 0.0324, 0.0372, 0.0129, 0.0490], device='cuda:0'), in_proj_covar=tensor([0.0200, 0.0164, 0.0172, 0.0195, 0.0206, 0.0203, 0.0179, 0.0206], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-17 07:39:53,295 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=4.26 vs. limit=5.0 2023-05-17 07:40:09,180 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.852e+02 2.526e+02 3.029e+02 3.726e+02 9.097e+02, threshold=6.058e+02, percent-clipped=2.0 2023-05-17 07:40:10,502 INFO [finetune.py:992] (0/2) Epoch 17, batch 2650, loss[loss=0.1735, simple_loss=0.2714, pruned_loss=0.03778, over 12286.00 frames. ], tot_loss[loss=0.163, simple_loss=0.253, pruned_loss=0.03653, over 2375469.22 frames. ], batch size: 33, lr: 3.33e-03, grad_scale: 16.0 2023-05-17 07:40:47,093 INFO [finetune.py:992] (0/2) Epoch 17, batch 2700, loss[loss=0.143, simple_loss=0.2371, pruned_loss=0.02444, over 12105.00 frames. ], tot_loss[loss=0.1626, simple_loss=0.2527, pruned_loss=0.03628, over 2383966.66 frames. ], batch size: 33, lr: 3.33e-03, grad_scale: 16.0 2023-05-17 07:41:14,058 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=298482.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 07:41:21,939 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.071e+02 2.706e+02 3.064e+02 3.739e+02 6.020e+02, threshold=6.128e+02, percent-clipped=0.0 2023-05-17 07:41:23,396 INFO [finetune.py:992] (0/2) Epoch 17, batch 2750, loss[loss=0.1663, simple_loss=0.2595, pruned_loss=0.0366, over 12102.00 frames. ], tot_loss[loss=0.1617, simple_loss=0.2515, pruned_loss=0.03596, over 2380922.98 frames. ], batch size: 33, lr: 3.33e-03, grad_scale: 16.0 2023-05-17 07:41:32,374 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.52 vs. limit=2.0 2023-05-17 07:41:49,093 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=298530.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 07:41:58,634 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=298543.0, num_to_drop=1, layers_to_drop={2} 2023-05-17 07:41:59,885 INFO [finetune.py:992] (0/2) Epoch 17, batch 2800, loss[loss=0.1587, simple_loss=0.2458, pruned_loss=0.03582, over 12151.00 frames. ], tot_loss[loss=0.1625, simple_loss=0.2523, pruned_loss=0.03634, over 2372177.32 frames. ], batch size: 34, lr: 3.33e-03, grad_scale: 16.0 2023-05-17 07:42:33,432 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=298591.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 07:42:34,702 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.094e+02 2.580e+02 2.945e+02 3.483e+02 6.486e+02, threshold=5.889e+02, percent-clipped=1.0 2023-05-17 07:42:36,185 INFO [finetune.py:992] (0/2) Epoch 17, batch 2850, loss[loss=0.1535, simple_loss=0.2464, pruned_loss=0.03029, over 12352.00 frames. ], tot_loss[loss=0.1626, simple_loss=0.2524, pruned_loss=0.03635, over 2372884.44 frames. ], batch size: 35, lr: 3.33e-03, grad_scale: 16.0 2023-05-17 07:42:48,944 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=298613.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 07:43:00,458 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=298629.0, num_to_drop=1, layers_to_drop={0} 2023-05-17 07:43:11,513 INFO [finetune.py:992] (0/2) Epoch 17, batch 2900, loss[loss=0.1638, simple_loss=0.2409, pruned_loss=0.04338, over 11430.00 frames. ], tot_loss[loss=0.1612, simple_loss=0.2511, pruned_loss=0.03567, over 2380706.93 frames. ], batch size: 25, lr: 3.33e-03, grad_scale: 16.0 2023-05-17 07:43:21,350 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([6.0578, 6.0253, 5.7771, 5.2930, 5.1703, 5.8944, 5.5278, 5.2512], device='cuda:0'), covar=tensor([0.0720, 0.0830, 0.0710, 0.1757, 0.0768, 0.0767, 0.1518, 0.1196], device='cuda:0'), in_proj_covar=tensor([0.0647, 0.0579, 0.0534, 0.0653, 0.0432, 0.0746, 0.0799, 0.0584], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-05-17 07:43:22,739 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=298661.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 07:43:34,973 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.3068, 3.5110, 3.1827, 3.1516, 2.8240, 2.6278, 3.4668, 2.3060], device='cuda:0'), covar=tensor([0.0507, 0.0150, 0.0224, 0.0230, 0.0453, 0.0441, 0.0178, 0.0513], device='cuda:0'), in_proj_covar=tensor([0.0202, 0.0165, 0.0173, 0.0197, 0.0208, 0.0205, 0.0180, 0.0207], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-17 07:43:43,429 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=298689.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 07:43:44,249 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=298690.0, num_to_drop=1, layers_to_drop={3} 2023-05-17 07:43:46,196 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.934e+02 2.709e+02 3.170e+02 3.797e+02 7.098e+02, threshold=6.341e+02, percent-clipped=3.0 2023-05-17 07:43:47,689 INFO [finetune.py:992] (0/2) Epoch 17, batch 2950, loss[loss=0.1344, simple_loss=0.2134, pruned_loss=0.02769, over 12186.00 frames. ], tot_loss[loss=0.1628, simple_loss=0.2523, pruned_loss=0.0366, over 2364740.90 frames. ], batch size: 29, lr: 3.33e-03, grad_scale: 16.0 2023-05-17 07:44:06,358 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.3601, 2.2441, 3.4697, 4.2871, 3.8626, 4.2915, 3.7747, 3.0843], device='cuda:0'), covar=tensor([0.0052, 0.0508, 0.0180, 0.0062, 0.0133, 0.0101, 0.0162, 0.0406], device='cuda:0'), in_proj_covar=tensor([0.0092, 0.0126, 0.0106, 0.0082, 0.0107, 0.0119, 0.0104, 0.0142], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-17 07:44:24,207 INFO [finetune.py:992] (0/2) Epoch 17, batch 3000, loss[loss=0.1655, simple_loss=0.2442, pruned_loss=0.04339, over 12337.00 frames. ], tot_loss[loss=0.1623, simple_loss=0.2519, pruned_loss=0.03635, over 2369313.24 frames. ], batch size: 31, lr: 3.33e-03, grad_scale: 16.0 2023-05-17 07:44:24,208 INFO [finetune.py:1017] (0/2) Computing validation loss 2023-05-17 07:44:42,390 INFO [finetune.py:1026] (0/2) Epoch 17, validation: loss=0.3104, simple_loss=0.3873, pruned_loss=0.1167, over 1020973.00 frames. 2023-05-17 07:44:42,391 INFO [finetune.py:1027] (0/2) Maximum memory allocated so far is 12827MB 2023-05-17 07:44:46,156 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=298750.0, num_to_drop=1, layers_to_drop={0} 2023-05-17 07:45:17,055 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.729e+02 2.736e+02 3.128e+02 3.461e+02 8.040e+02, threshold=6.255e+02, percent-clipped=1.0 2023-05-17 07:45:18,416 INFO [finetune.py:992] (0/2) Epoch 17, batch 3050, loss[loss=0.1706, simple_loss=0.2674, pruned_loss=0.03694, over 11632.00 frames. ], tot_loss[loss=0.1628, simple_loss=0.2527, pruned_loss=0.03643, over 2377647.83 frames. ], batch size: 48, lr: 3.33e-03, grad_scale: 16.0 2023-05-17 07:45:23,573 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=4.96 vs. limit=5.0 2023-05-17 07:45:50,488 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=298838.0, num_to_drop=1, layers_to_drop={2} 2023-05-17 07:45:55,401 INFO [finetune.py:992] (0/2) Epoch 17, batch 3100, loss[loss=0.1574, simple_loss=0.2529, pruned_loss=0.03097, over 12111.00 frames. ], tot_loss[loss=0.1624, simple_loss=0.2522, pruned_loss=0.03625, over 2377307.61 frames. ], batch size: 38, lr: 3.33e-03, grad_scale: 16.0 2023-05-17 07:46:00,758 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.65 vs. limit=2.0 2023-05-17 07:46:24,622 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=298886.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 07:46:29,431 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.724e+02 2.477e+02 2.875e+02 3.677e+02 5.878e+02, threshold=5.751e+02, percent-clipped=0.0 2023-05-17 07:46:30,908 INFO [finetune.py:992] (0/2) Epoch 17, batch 3150, loss[loss=0.1372, simple_loss=0.2225, pruned_loss=0.02596, over 12040.00 frames. ], tot_loss[loss=0.1624, simple_loss=0.2522, pruned_loss=0.03628, over 2378560.84 frames. ], batch size: 28, lr: 3.33e-03, grad_scale: 16.0 2023-05-17 07:46:33,192 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.0179, 4.9182, 4.8208, 4.8655, 4.5632, 5.0739, 5.0129, 5.2554], device='cuda:0'), covar=tensor([0.0553, 0.0187, 0.0290, 0.0437, 0.0914, 0.0332, 0.0159, 0.0215], device='cuda:0'), in_proj_covar=tensor([0.0205, 0.0207, 0.0199, 0.0252, 0.0249, 0.0227, 0.0181, 0.0238], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0005, 0.0004, 0.0004, 0.0003, 0.0005], device='cuda:0') 2023-05-17 07:47:07,188 INFO [finetune.py:992] (0/2) Epoch 17, batch 3200, loss[loss=0.1548, simple_loss=0.2523, pruned_loss=0.02863, over 12122.00 frames. ], tot_loss[loss=0.1627, simple_loss=0.2525, pruned_loss=0.03643, over 2369276.42 frames. ], batch size: 33, lr: 3.33e-03, grad_scale: 16.0 2023-05-17 07:47:20,232 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.6211, 2.6442, 4.1897, 4.4854, 2.8948, 2.6194, 2.8997, 2.1159], device='cuda:0'), covar=tensor([0.1764, 0.3142, 0.0652, 0.0438, 0.1287, 0.2569, 0.2872, 0.4455], device='cuda:0'), in_proj_covar=tensor([0.0310, 0.0399, 0.0279, 0.0307, 0.0282, 0.0324, 0.0404, 0.0389], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-17 07:47:36,541 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=298985.0, num_to_drop=1, layers_to_drop={0} 2023-05-17 07:47:41,856 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.999e+02 2.653e+02 3.106e+02 3.672e+02 8.833e+02, threshold=6.212e+02, percent-clipped=1.0 2023-05-17 07:47:43,239 INFO [finetune.py:992] (0/2) Epoch 17, batch 3250, loss[loss=0.1769, simple_loss=0.2651, pruned_loss=0.04439, over 12030.00 frames. ], tot_loss[loss=0.1629, simple_loss=0.2528, pruned_loss=0.03652, over 2367992.47 frames. ], batch size: 42, lr: 3.33e-03, grad_scale: 16.0 2023-05-17 07:48:00,955 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.5856, 2.5294, 3.6485, 4.5905, 3.9536, 4.5589, 3.9860, 3.1675], device='cuda:0'), covar=tensor([0.0040, 0.0460, 0.0171, 0.0044, 0.0123, 0.0079, 0.0143, 0.0389], device='cuda:0'), in_proj_covar=tensor([0.0092, 0.0125, 0.0106, 0.0082, 0.0106, 0.0118, 0.0103, 0.0142], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-17 07:48:19,326 INFO [finetune.py:992] (0/2) Epoch 17, batch 3300, loss[loss=0.1472, simple_loss=0.2325, pruned_loss=0.03098, over 12189.00 frames. ], tot_loss[loss=0.162, simple_loss=0.2518, pruned_loss=0.03609, over 2370593.55 frames. ], batch size: 31, lr: 3.33e-03, grad_scale: 16.0 2023-05-17 07:48:19,441 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=299045.0, num_to_drop=1, layers_to_drop={2} 2023-05-17 07:48:42,436 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.35 vs. limit=2.0 2023-05-17 07:48:44,842 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.0953, 5.8640, 5.5342, 5.3851, 6.0007, 5.3282, 5.4610, 5.4177], device='cuda:0'), covar=tensor([0.1543, 0.0987, 0.1063, 0.1869, 0.0994, 0.2153, 0.1882, 0.1346], device='cuda:0'), in_proj_covar=tensor([0.0363, 0.0505, 0.0409, 0.0457, 0.0477, 0.0445, 0.0410, 0.0396], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-17 07:48:54,069 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.070e+02 2.781e+02 3.093e+02 3.666e+02 9.135e+02, threshold=6.186e+02, percent-clipped=3.0 2023-05-17 07:48:55,549 INFO [finetune.py:992] (0/2) Epoch 17, batch 3350, loss[loss=0.2099, simple_loss=0.2952, pruned_loss=0.06227, over 11051.00 frames. ], tot_loss[loss=0.1635, simple_loss=0.2534, pruned_loss=0.03674, over 2370404.49 frames. ], batch size: 55, lr: 3.33e-03, grad_scale: 16.0 2023-05-17 07:48:56,088 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.40 vs. limit=2.0 2023-05-17 07:49:05,229 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.9670, 2.4568, 3.4963, 3.0432, 3.4341, 3.1986, 2.5609, 3.4988], device='cuda:0'), covar=tensor([0.0172, 0.0403, 0.0220, 0.0255, 0.0152, 0.0194, 0.0404, 0.0153], device='cuda:0'), in_proj_covar=tensor([0.0191, 0.0216, 0.0203, 0.0199, 0.0231, 0.0176, 0.0208, 0.0201], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-17 07:49:18,497 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=299126.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 07:49:26,876 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=299138.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 07:49:31,594 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.90 vs. limit=2.0 2023-05-17 07:49:31,853 INFO [finetune.py:992] (0/2) Epoch 17, batch 3400, loss[loss=0.1725, simple_loss=0.2621, pruned_loss=0.04143, over 8233.00 frames. ], tot_loss[loss=0.1631, simple_loss=0.2531, pruned_loss=0.03661, over 2356985.08 frames. ], batch size: 99, lr: 3.33e-03, grad_scale: 16.0 2023-05-17 07:49:51,765 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.5620, 2.6563, 3.2621, 4.3537, 2.5871, 4.3383, 4.5368, 4.5567], device='cuda:0'), covar=tensor([0.0120, 0.1198, 0.0516, 0.0160, 0.1268, 0.0305, 0.0132, 0.0124], device='cuda:0'), in_proj_covar=tensor([0.0124, 0.0207, 0.0186, 0.0124, 0.0193, 0.0181, 0.0177, 0.0127], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-17 07:49:57,283 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.3987, 5.2420, 5.3586, 5.3622, 4.9787, 5.0646, 4.7857, 5.3044], device='cuda:0'), covar=tensor([0.0726, 0.0563, 0.0826, 0.0627, 0.2021, 0.1345, 0.0587, 0.1057], device='cuda:0'), in_proj_covar=tensor([0.0557, 0.0715, 0.0631, 0.0652, 0.0864, 0.0770, 0.0574, 0.0494], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0005, 0.0005, 0.0003, 0.0003], device='cuda:0') 2023-05-17 07:50:00,760 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=299186.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 07:50:00,875 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=299186.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 07:50:01,663 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=299187.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 07:50:04,442 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=299191.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 07:50:05,617 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.900e+02 2.674e+02 3.092e+02 3.690e+02 1.169e+03, threshold=6.184e+02, percent-clipped=3.0 2023-05-17 07:50:07,143 INFO [finetune.py:992] (0/2) Epoch 17, batch 3450, loss[loss=0.1355, simple_loss=0.2225, pruned_loss=0.02423, over 12366.00 frames. ], tot_loss[loss=0.1626, simple_loss=0.2524, pruned_loss=0.03642, over 2359804.85 frames. ], batch size: 30, lr: 3.33e-03, grad_scale: 16.0 2023-05-17 07:50:08,871 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.1963, 2.7096, 3.6640, 3.1445, 3.5736, 3.3305, 2.7094, 3.6772], device='cuda:0'), covar=tensor([0.0151, 0.0362, 0.0196, 0.0257, 0.0170, 0.0198, 0.0403, 0.0136], device='cuda:0'), in_proj_covar=tensor([0.0191, 0.0216, 0.0202, 0.0199, 0.0231, 0.0176, 0.0207, 0.0201], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-17 07:50:35,055 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.3784, 3.6980, 3.2774, 3.2716, 3.0198, 2.8914, 3.6619, 2.2403], device='cuda:0'), covar=tensor([0.0515, 0.0160, 0.0232, 0.0217, 0.0359, 0.0355, 0.0147, 0.0605], device='cuda:0'), in_proj_covar=tensor([0.0202, 0.0167, 0.0174, 0.0197, 0.0209, 0.0206, 0.0181, 0.0208], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-17 07:50:36,347 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=299234.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 07:50:42,804 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=299243.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 07:50:44,031 INFO [finetune.py:992] (0/2) Epoch 17, batch 3500, loss[loss=0.1805, simple_loss=0.2695, pruned_loss=0.04579, over 11797.00 frames. ], tot_loss[loss=0.1622, simple_loss=0.2515, pruned_loss=0.03646, over 2366625.69 frames. ], batch size: 44, lr: 3.33e-03, grad_scale: 16.0 2023-05-17 07:50:49,630 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=299252.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 07:51:12,966 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=299285.0, num_to_drop=1, layers_to_drop={2} 2023-05-17 07:51:18,436 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.522e+02 2.567e+02 2.962e+02 3.764e+02 6.367e+02, threshold=5.924e+02, percent-clipped=1.0 2023-05-17 07:51:19,794 INFO [finetune.py:992] (0/2) Epoch 17, batch 3550, loss[loss=0.1512, simple_loss=0.2448, pruned_loss=0.02882, over 12314.00 frames. ], tot_loss[loss=0.1626, simple_loss=0.2519, pruned_loss=0.03665, over 2368936.87 frames. ], batch size: 34, lr: 3.33e-03, grad_scale: 16.0 2023-05-17 07:51:25,714 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([6.0423, 6.0226, 5.7757, 5.3617, 5.1377, 5.8975, 5.5110, 5.2719], device='cuda:0'), covar=tensor([0.0746, 0.0948, 0.0736, 0.1581, 0.0775, 0.0814, 0.1741, 0.1287], device='cuda:0'), in_proj_covar=tensor([0.0653, 0.0582, 0.0537, 0.0656, 0.0437, 0.0754, 0.0809, 0.0589], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-05-17 07:51:26,526 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=299304.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 07:51:40,232 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=299323.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 07:51:47,280 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=299333.0, num_to_drop=1, layers_to_drop={1} 2023-05-17 07:51:49,816 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=4.69 vs. limit=5.0 2023-05-17 07:51:55,724 INFO [finetune.py:992] (0/2) Epoch 17, batch 3600, loss[loss=0.1832, simple_loss=0.2569, pruned_loss=0.05472, over 8262.00 frames. ], tot_loss[loss=0.1624, simple_loss=0.2517, pruned_loss=0.0366, over 2365863.70 frames. ], batch size: 98, lr: 3.33e-03, grad_scale: 16.0 2023-05-17 07:51:55,847 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=299345.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 07:52:25,037 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=299384.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 07:52:31,078 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.023e+02 2.632e+02 3.074e+02 3.664e+02 7.130e+02, threshold=6.147e+02, percent-clipped=1.0 2023-05-17 07:52:31,156 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=299393.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 07:52:32,556 INFO [finetune.py:992] (0/2) Epoch 17, batch 3650, loss[loss=0.1715, simple_loss=0.2755, pruned_loss=0.03373, over 12104.00 frames. ], tot_loss[loss=0.1628, simple_loss=0.2521, pruned_loss=0.03679, over 2360060.73 frames. ], batch size: 33, lr: 3.33e-03, grad_scale: 16.0 2023-05-17 07:52:51,368 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.9693, 5.9722, 5.7554, 5.1833, 5.1415, 5.8730, 5.4315, 5.2150], device='cuda:0'), covar=tensor([0.0828, 0.0921, 0.0716, 0.1750, 0.0739, 0.0767, 0.1754, 0.1116], device='cuda:0'), in_proj_covar=tensor([0.0652, 0.0580, 0.0536, 0.0655, 0.0435, 0.0751, 0.0807, 0.0589], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-05-17 07:53:08,243 INFO [finetune.py:992] (0/2) Epoch 17, batch 3700, loss[loss=0.1561, simple_loss=0.2443, pruned_loss=0.03397, over 12023.00 frames. ], tot_loss[loss=0.1628, simple_loss=0.252, pruned_loss=0.03675, over 2374756.81 frames. ], batch size: 31, lr: 3.33e-03, grad_scale: 16.0 2023-05-17 07:53:32,933 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 2023-05-17 07:53:34,718 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=299482.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 07:53:39,100 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.4893, 5.0075, 5.4717, 4.7141, 5.0914, 4.8464, 5.5395, 5.0913], device='cuda:0'), covar=tensor([0.0278, 0.0399, 0.0271, 0.0281, 0.0417, 0.0355, 0.0193, 0.0336], device='cuda:0'), in_proj_covar=tensor([0.0277, 0.0281, 0.0302, 0.0278, 0.0276, 0.0275, 0.0251, 0.0223], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-17 07:53:42,561 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.909e+02 2.649e+02 2.993e+02 3.473e+02 6.479e+02, threshold=5.985e+02, percent-clipped=1.0 2023-05-17 07:53:44,055 INFO [finetune.py:992] (0/2) Epoch 17, batch 3750, loss[loss=0.1458, simple_loss=0.2485, pruned_loss=0.02154, over 12159.00 frames. ], tot_loss[loss=0.1627, simple_loss=0.2521, pruned_loss=0.03668, over 2375892.26 frames. ], batch size: 36, lr: 3.33e-03, grad_scale: 16.0 2023-05-17 07:54:20,890 INFO [finetune.py:992] (0/2) Epoch 17, batch 3800, loss[loss=0.1433, simple_loss=0.2197, pruned_loss=0.03349, over 12251.00 frames. ], tot_loss[loss=0.1619, simple_loss=0.2511, pruned_loss=0.03636, over 2371909.99 frames. ], batch size: 28, lr: 3.33e-03, grad_scale: 16.0 2023-05-17 07:54:22,390 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=299547.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 07:54:32,417 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.3649, 4.7327, 2.9988, 2.4963, 4.2123, 2.4690, 3.9821, 3.1857], device='cuda:0'), covar=tensor([0.0794, 0.0559, 0.1264, 0.1810, 0.0322, 0.1627, 0.0609, 0.0884], device='cuda:0'), in_proj_covar=tensor([0.0193, 0.0263, 0.0180, 0.0204, 0.0144, 0.0185, 0.0204, 0.0177], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-17 07:54:55,133 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.802e+02 2.650e+02 3.115e+02 3.689e+02 6.414e+02, threshold=6.230e+02, percent-clipped=3.0 2023-05-17 07:54:56,642 INFO [finetune.py:992] (0/2) Epoch 17, batch 3850, loss[loss=0.187, simple_loss=0.2779, pruned_loss=0.04803, over 8062.00 frames. ], tot_loss[loss=0.1615, simple_loss=0.2506, pruned_loss=0.03618, over 2363236.60 frames. ], batch size: 98, lr: 3.33e-03, grad_scale: 16.0 2023-05-17 07:54:59,431 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=299599.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 07:55:32,506 INFO [finetune.py:992] (0/2) Epoch 17, batch 3900, loss[loss=0.1527, simple_loss=0.2263, pruned_loss=0.0395, over 11993.00 frames. ], tot_loss[loss=0.1619, simple_loss=0.2512, pruned_loss=0.03626, over 2363078.91 frames. ], batch size: 28, lr: 3.32e-03, grad_scale: 16.0 2023-05-17 07:55:44,604 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.40 vs. limit=2.0 2023-05-17 07:55:57,577 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=299679.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 07:56:03,488 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.57 vs. limit=2.0 2023-05-17 07:56:08,116 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.885e+02 2.706e+02 3.266e+02 3.999e+02 8.575e+02, threshold=6.532e+02, percent-clipped=2.0 2023-05-17 07:56:09,538 INFO [finetune.py:992] (0/2) Epoch 17, batch 3950, loss[loss=0.1707, simple_loss=0.2597, pruned_loss=0.04091, over 12278.00 frames. ], tot_loss[loss=0.1614, simple_loss=0.2509, pruned_loss=0.03599, over 2371278.70 frames. ], batch size: 37, lr: 3.32e-03, grad_scale: 16.0 2023-05-17 07:56:17,685 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2023-05-17 07:56:44,712 INFO [finetune.py:992] (0/2) Epoch 17, batch 4000, loss[loss=0.1596, simple_loss=0.2398, pruned_loss=0.03974, over 11839.00 frames. ], tot_loss[loss=0.1612, simple_loss=0.2507, pruned_loss=0.03585, over 2377716.73 frames. ], batch size: 26, lr: 3.32e-03, grad_scale: 16.0 2023-05-17 07:57:01,692 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=299769.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 07:57:02,673 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.53 vs. limit=2.0 2023-05-17 07:57:10,656 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=299782.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 07:57:18,496 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.820e+02 2.576e+02 3.064e+02 3.696e+02 7.843e+02, threshold=6.128e+02, percent-clipped=2.0 2023-05-17 07:57:20,605 INFO [finetune.py:992] (0/2) Epoch 17, batch 4050, loss[loss=0.1752, simple_loss=0.2688, pruned_loss=0.04078, over 12077.00 frames. ], tot_loss[loss=0.1621, simple_loss=0.2519, pruned_loss=0.03617, over 2368748.27 frames. ], batch size: 42, lr: 3.32e-03, grad_scale: 16.0 2023-05-17 07:57:22,980 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=299798.0, num_to_drop=1, layers_to_drop={1} 2023-05-17 07:57:47,056 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=299830.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 07:57:47,182 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=299830.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 07:57:57,497 INFO [finetune.py:992] (0/2) Epoch 17, batch 4100, loss[loss=0.201, simple_loss=0.2959, pruned_loss=0.05308, over 12150.00 frames. ], tot_loss[loss=0.1625, simple_loss=0.2523, pruned_loss=0.03638, over 2364831.04 frames. ], batch size: 39, lr: 3.32e-03, grad_scale: 16.0 2023-05-17 07:57:58,993 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=299847.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 07:58:07,683 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=299859.0, num_to_drop=1, layers_to_drop={3} 2023-05-17 07:58:14,098 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.4982, 5.3033, 5.3687, 5.4243, 5.0248, 5.0815, 4.8305, 5.3866], device='cuda:0'), covar=tensor([0.0725, 0.0632, 0.0926, 0.0679, 0.2081, 0.1482, 0.0583, 0.1048], device='cuda:0'), in_proj_covar=tensor([0.0564, 0.0723, 0.0644, 0.0668, 0.0878, 0.0781, 0.0583, 0.0504], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0005, 0.0005, 0.0004, 0.0003], device='cuda:0') 2023-05-17 07:58:31,671 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.917e+02 2.564e+02 2.897e+02 3.446e+02 7.491e+02, threshold=5.795e+02, percent-clipped=2.0 2023-05-17 07:58:33,052 INFO [finetune.py:992] (0/2) Epoch 17, batch 4150, loss[loss=0.1771, simple_loss=0.2731, pruned_loss=0.04056, over 12030.00 frames. ], tot_loss[loss=0.1626, simple_loss=0.2525, pruned_loss=0.03638, over 2367779.83 frames. ], batch size: 42, lr: 3.32e-03, grad_scale: 16.0 2023-05-17 07:58:33,123 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=299895.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 07:58:35,926 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=299899.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 07:58:56,850 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.1662, 5.8990, 5.4803, 5.5073, 6.0572, 5.3739, 5.4380, 5.4737], device='cuda:0'), covar=tensor([0.1666, 0.1091, 0.1368, 0.2014, 0.1001, 0.2341, 0.2117, 0.1261], device='cuda:0'), in_proj_covar=tensor([0.0373, 0.0516, 0.0421, 0.0464, 0.0484, 0.0455, 0.0419, 0.0405], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-17 07:59:07,167 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.12 vs. limit=2.0 2023-05-17 07:59:08,814 INFO [finetune.py:992] (0/2) Epoch 17, batch 4200, loss[loss=0.1618, simple_loss=0.264, pruned_loss=0.02976, over 12354.00 frames. ], tot_loss[loss=0.1622, simple_loss=0.2522, pruned_loss=0.03612, over 2374595.37 frames. ], batch size: 36, lr: 3.32e-03, grad_scale: 16.0 2023-05-17 07:59:11,028 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=299947.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 07:59:29,826 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.6065, 2.7788, 4.4069, 4.5608, 2.7960, 2.4627, 2.8552, 2.0838], device='cuda:0'), covar=tensor([0.1747, 0.3159, 0.0489, 0.0445, 0.1455, 0.2878, 0.2920, 0.4517], device='cuda:0'), in_proj_covar=tensor([0.0309, 0.0397, 0.0279, 0.0306, 0.0282, 0.0324, 0.0405, 0.0387], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-17 07:59:31,196 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.5995, 2.6305, 3.3559, 4.5413, 2.4878, 4.5205, 4.6329, 4.6469], device='cuda:0'), covar=tensor([0.0159, 0.1246, 0.0464, 0.0162, 0.1377, 0.0240, 0.0152, 0.0120], device='cuda:0'), in_proj_covar=tensor([0.0124, 0.0207, 0.0186, 0.0122, 0.0192, 0.0181, 0.0177, 0.0128], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-17 07:59:34,762 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=299979.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 07:59:44,518 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.829e+02 2.549e+02 3.017e+02 3.536e+02 5.011e+02, threshold=6.035e+02, percent-clipped=0.0 2023-05-17 07:59:45,977 INFO [finetune.py:992] (0/2) Epoch 17, batch 4250, loss[loss=0.1449, simple_loss=0.241, pruned_loss=0.02439, over 12147.00 frames. ], tot_loss[loss=0.1623, simple_loss=0.2522, pruned_loss=0.03617, over 2376773.41 frames. ], batch size: 34, lr: 3.32e-03, grad_scale: 16.0 2023-05-17 07:59:49,832 INFO [checkpoint.py:75] (0/2) Saving checkpoint to pruned_transducer_stateless7/exp_giga_finetune/checkpoint-200000.pt 2023-05-17 08:00:12,174 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=300027.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 08:00:24,947 INFO [finetune.py:992] (0/2) Epoch 17, batch 4300, loss[loss=0.1835, simple_loss=0.2701, pruned_loss=0.0484, over 10423.00 frames. ], tot_loss[loss=0.1626, simple_loss=0.2523, pruned_loss=0.03644, over 2375746.08 frames. ], batch size: 68, lr: 3.32e-03, grad_scale: 16.0 2023-05-17 08:00:44,672 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=300073.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 08:00:59,088 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.818e+02 2.688e+02 3.373e+02 4.196e+02 1.008e+03, threshold=6.746e+02, percent-clipped=3.0 2023-05-17 08:01:00,500 INFO [finetune.py:992] (0/2) Epoch 17, batch 4350, loss[loss=0.1861, simple_loss=0.2811, pruned_loss=0.0455, over 12121.00 frames. ], tot_loss[loss=0.1639, simple_loss=0.2534, pruned_loss=0.0372, over 2366481.06 frames. ], batch size: 42, lr: 3.32e-03, grad_scale: 16.0 2023-05-17 08:01:15,202 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.33 vs. limit=2.0 2023-05-17 08:01:22,578 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=300125.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 08:01:29,238 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=300134.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 08:01:32,145 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.2415, 4.8943, 5.1143, 5.1296, 4.9571, 5.1029, 4.9939, 2.7568], device='cuda:0'), covar=tensor([0.0098, 0.0068, 0.0074, 0.0055, 0.0045, 0.0102, 0.0081, 0.0774], device='cuda:0'), in_proj_covar=tensor([0.0073, 0.0083, 0.0088, 0.0078, 0.0064, 0.0099, 0.0086, 0.0103], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-17 08:01:36,909 INFO [finetune.py:992] (0/2) Epoch 17, batch 4400, loss[loss=0.1433, simple_loss=0.2345, pruned_loss=0.02604, over 12099.00 frames. ], tot_loss[loss=0.1628, simple_loss=0.2525, pruned_loss=0.03652, over 2377167.73 frames. ], batch size: 33, lr: 3.32e-03, grad_scale: 16.0 2023-05-17 08:01:43,528 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=300154.0, num_to_drop=1, layers_to_drop={0} 2023-05-17 08:02:11,060 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.669e+02 2.546e+02 3.053e+02 3.701e+02 7.801e+02, threshold=6.107e+02, percent-clipped=1.0 2023-05-17 08:02:12,543 INFO [finetune.py:992] (0/2) Epoch 17, batch 4450, loss[loss=0.1421, simple_loss=0.2253, pruned_loss=0.02947, over 12360.00 frames. ], tot_loss[loss=0.163, simple_loss=0.2527, pruned_loss=0.03662, over 2377393.16 frames. ], batch size: 30, lr: 3.32e-03, grad_scale: 32.0 2023-05-17 08:02:12,721 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.0750, 4.8984, 4.8408, 4.9581, 4.6302, 5.0687, 5.0649, 5.2146], device='cuda:0'), covar=tensor([0.0310, 0.0176, 0.0222, 0.0348, 0.0733, 0.0322, 0.0134, 0.0184], device='cuda:0'), in_proj_covar=tensor([0.0206, 0.0208, 0.0200, 0.0254, 0.0250, 0.0228, 0.0182, 0.0240], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0005, 0.0004, 0.0004, 0.0003, 0.0005], device='cuda:0') 2023-05-17 08:02:32,116 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.5379, 2.6925, 3.2028, 4.4724, 2.4740, 4.4001, 4.6021, 4.6111], device='cuda:0'), covar=tensor([0.0160, 0.1177, 0.0498, 0.0149, 0.1282, 0.0240, 0.0125, 0.0098], device='cuda:0'), in_proj_covar=tensor([0.0124, 0.0206, 0.0185, 0.0122, 0.0191, 0.0181, 0.0177, 0.0127], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-17 08:02:48,862 INFO [finetune.py:992] (0/2) Epoch 17, batch 4500, loss[loss=0.1443, simple_loss=0.2432, pruned_loss=0.02266, over 12117.00 frames. ], tot_loss[loss=0.1622, simple_loss=0.2521, pruned_loss=0.03621, over 2378959.72 frames. ], batch size: 33, lr: 3.32e-03, grad_scale: 16.0 2023-05-17 08:03:09,990 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.7288, 3.0067, 3.8257, 4.7626, 4.0259, 4.7964, 4.1660, 3.7215], device='cuda:0'), covar=tensor([0.0047, 0.0370, 0.0142, 0.0048, 0.0142, 0.0063, 0.0105, 0.0300], device='cuda:0'), in_proj_covar=tensor([0.0092, 0.0125, 0.0106, 0.0082, 0.0107, 0.0118, 0.0103, 0.0141], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-17 08:03:10,875 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2023-05-17 08:03:13,101 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=3.24 vs. limit=5.0 2023-05-17 08:03:24,014 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.957e+02 2.601e+02 3.141e+02 3.922e+02 6.717e+02, threshold=6.282e+02, percent-clipped=1.0 2023-05-17 08:03:24,753 INFO [finetune.py:992] (0/2) Epoch 17, batch 4550, loss[loss=0.1579, simple_loss=0.2586, pruned_loss=0.02859, over 12370.00 frames. ], tot_loss[loss=0.162, simple_loss=0.2519, pruned_loss=0.03606, over 2369659.23 frames. ], batch size: 36, lr: 3.32e-03, grad_scale: 16.0 2023-05-17 08:03:40,972 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.3804, 4.9714, 5.3251, 4.6080, 4.9581, 4.6974, 5.4113, 4.9499], device='cuda:0'), covar=tensor([0.0288, 0.0374, 0.0294, 0.0283, 0.0438, 0.0314, 0.0205, 0.0363], device='cuda:0'), in_proj_covar=tensor([0.0277, 0.0282, 0.0306, 0.0277, 0.0278, 0.0275, 0.0251, 0.0223], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-17 08:04:00,673 INFO [finetune.py:992] (0/2) Epoch 17, batch 4600, loss[loss=0.1653, simple_loss=0.2486, pruned_loss=0.04104, over 12132.00 frames. ], tot_loss[loss=0.1623, simple_loss=0.2521, pruned_loss=0.03624, over 2378859.30 frames. ], batch size: 38, lr: 3.32e-03, grad_scale: 16.0 2023-05-17 08:04:04,518 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.5941, 2.7323, 3.2085, 4.4581, 2.4542, 4.3986, 4.5419, 4.6442], device='cuda:0'), covar=tensor([0.0134, 0.1208, 0.0535, 0.0141, 0.1348, 0.0247, 0.0150, 0.0091], device='cuda:0'), in_proj_covar=tensor([0.0124, 0.0206, 0.0185, 0.0122, 0.0191, 0.0181, 0.0178, 0.0127], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-17 08:04:36,141 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.93 vs. limit=2.0 2023-05-17 08:04:36,378 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.718e+02 2.656e+02 3.013e+02 3.569e+02 7.028e+02, threshold=6.027e+02, percent-clipped=3.0 2023-05-17 08:04:37,151 INFO [finetune.py:992] (0/2) Epoch 17, batch 4650, loss[loss=0.1818, simple_loss=0.274, pruned_loss=0.04474, over 12043.00 frames. ], tot_loss[loss=0.1626, simple_loss=0.2523, pruned_loss=0.03641, over 2378629.07 frames. ], batch size: 40, lr: 3.32e-03, grad_scale: 16.0 2023-05-17 08:04:59,023 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=300425.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 08:05:01,686 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=300429.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 08:05:12,812 INFO [finetune.py:992] (0/2) Epoch 17, batch 4700, loss[loss=0.1683, simple_loss=0.2441, pruned_loss=0.04625, over 11842.00 frames. ], tot_loss[loss=0.164, simple_loss=0.2537, pruned_loss=0.03717, over 2374185.61 frames. ], batch size: 26, lr: 3.32e-03, grad_scale: 16.0 2023-05-17 08:05:19,383 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=300454.0, num_to_drop=1, layers_to_drop={0} 2023-05-17 08:05:32,653 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=300473.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 08:05:38,391 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.2099, 6.1637, 5.7696, 5.7453, 6.2726, 5.5227, 5.6165, 5.7624], device='cuda:0'), covar=tensor([0.1525, 0.0896, 0.1098, 0.1624, 0.0874, 0.2187, 0.2162, 0.1052], device='cuda:0'), in_proj_covar=tensor([0.0373, 0.0514, 0.0417, 0.0462, 0.0479, 0.0452, 0.0419, 0.0402], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-17 08:05:47,514 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.635e+02 2.583e+02 3.059e+02 3.722e+02 6.746e+02, threshold=6.118e+02, percent-clipped=2.0 2023-05-17 08:05:48,237 INFO [finetune.py:992] (0/2) Epoch 17, batch 4750, loss[loss=0.1738, simple_loss=0.2623, pruned_loss=0.04262, over 12376.00 frames. ], tot_loss[loss=0.1644, simple_loss=0.254, pruned_loss=0.03738, over 2374857.35 frames. ], batch size: 38, lr: 3.32e-03, grad_scale: 16.0 2023-05-17 08:05:53,430 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=300502.0, num_to_drop=1, layers_to_drop={1} 2023-05-17 08:06:07,827 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.1275, 2.1686, 2.9590, 3.0135, 3.0139, 3.1809, 2.8366, 2.5198], device='cuda:0'), covar=tensor([0.0092, 0.0403, 0.0181, 0.0101, 0.0159, 0.0102, 0.0155, 0.0343], device='cuda:0'), in_proj_covar=tensor([0.0093, 0.0126, 0.0107, 0.0082, 0.0107, 0.0119, 0.0104, 0.0142], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-17 08:06:18,593 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.2832, 4.9901, 5.2861, 4.5578, 4.9941, 4.6715, 5.3112, 4.9160], device='cuda:0'), covar=tensor([0.0335, 0.0445, 0.0421, 0.0337, 0.0474, 0.0374, 0.0299, 0.0393], device='cuda:0'), in_proj_covar=tensor([0.0281, 0.0285, 0.0309, 0.0281, 0.0281, 0.0278, 0.0254, 0.0226], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-17 08:06:24,888 INFO [finetune.py:992] (0/2) Epoch 17, batch 4800, loss[loss=0.178, simple_loss=0.2746, pruned_loss=0.0407, over 12138.00 frames. ], tot_loss[loss=0.1633, simple_loss=0.2528, pruned_loss=0.03691, over 2368234.26 frames. ], batch size: 39, lr: 3.32e-03, grad_scale: 16.0 2023-05-17 08:06:34,266 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.6701, 2.6937, 4.2291, 4.4322, 2.8122, 2.5325, 2.8732, 2.1417], device='cuda:0'), covar=tensor([0.1760, 0.2975, 0.0547, 0.0470, 0.1343, 0.2668, 0.2806, 0.4393], device='cuda:0'), in_proj_covar=tensor([0.0309, 0.0395, 0.0278, 0.0305, 0.0281, 0.0322, 0.0402, 0.0384], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-17 08:07:00,330 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.865e+02 2.535e+02 3.170e+02 3.712e+02 6.173e+02, threshold=6.341e+02, percent-clipped=1.0 2023-05-17 08:07:01,079 INFO [finetune.py:992] (0/2) Epoch 17, batch 4850, loss[loss=0.1464, simple_loss=0.2286, pruned_loss=0.03212, over 12048.00 frames. ], tot_loss[loss=0.1631, simple_loss=0.2522, pruned_loss=0.03702, over 2362930.53 frames. ], batch size: 31, lr: 3.32e-03, grad_scale: 16.0 2023-05-17 08:07:17,235 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=300617.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 08:07:21,554 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.9895, 4.8072, 4.8046, 4.9367, 3.9534, 5.0468, 5.0658, 5.1878], device='cuda:0'), covar=tensor([0.0259, 0.0209, 0.0223, 0.0330, 0.1274, 0.0331, 0.0166, 0.0207], device='cuda:0'), in_proj_covar=tensor([0.0205, 0.0207, 0.0199, 0.0252, 0.0249, 0.0228, 0.0182, 0.0239], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0005, 0.0004, 0.0004, 0.0003, 0.0005], device='cuda:0') 2023-05-17 08:07:24,492 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.1803, 4.5859, 2.7768, 2.5846, 3.9295, 2.5978, 3.9133, 3.2022], device='cuda:0'), covar=tensor([0.0828, 0.0604, 0.1253, 0.1614, 0.0392, 0.1389, 0.0528, 0.0843], device='cuda:0'), in_proj_covar=tensor([0.0191, 0.0262, 0.0178, 0.0203, 0.0144, 0.0186, 0.0203, 0.0177], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-17 08:07:37,165 INFO [finetune.py:992] (0/2) Epoch 17, batch 4900, loss[loss=0.1656, simple_loss=0.2606, pruned_loss=0.0353, over 11128.00 frames. ], tot_loss[loss=0.163, simple_loss=0.2522, pruned_loss=0.03687, over 2355106.33 frames. ], batch size: 55, lr: 3.32e-03, grad_scale: 16.0 2023-05-17 08:07:48,878 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.2033, 2.7092, 3.7700, 3.1691, 3.6079, 3.3509, 2.7674, 3.6175], device='cuda:0'), covar=tensor([0.0158, 0.0373, 0.0165, 0.0262, 0.0162, 0.0197, 0.0370, 0.0166], device='cuda:0'), in_proj_covar=tensor([0.0187, 0.0213, 0.0199, 0.0196, 0.0228, 0.0174, 0.0205, 0.0198], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-17 08:08:00,414 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.9057, 3.5834, 5.2751, 2.8382, 2.9396, 3.9140, 3.4584, 3.9311], device='cuda:0'), covar=tensor([0.0443, 0.1018, 0.0347, 0.1193, 0.2070, 0.1464, 0.1266, 0.1128], device='cuda:0'), in_proj_covar=tensor([0.0242, 0.0240, 0.0260, 0.0187, 0.0242, 0.0299, 0.0228, 0.0270], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-17 08:08:01,821 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=300678.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 08:08:12,890 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.787e+02 2.566e+02 3.041e+02 3.803e+02 9.854e+02, threshold=6.082e+02, percent-clipped=2.0 2023-05-17 08:08:13,649 INFO [finetune.py:992] (0/2) Epoch 17, batch 4950, loss[loss=0.1721, simple_loss=0.2659, pruned_loss=0.0392, over 12043.00 frames. ], tot_loss[loss=0.1627, simple_loss=0.2516, pruned_loss=0.03686, over 2358844.13 frames. ], batch size: 40, lr: 3.32e-03, grad_scale: 16.0 2023-05-17 08:08:38,214 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=300729.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 08:08:49,239 INFO [finetune.py:992] (0/2) Epoch 17, batch 5000, loss[loss=0.1749, simple_loss=0.2633, pruned_loss=0.04321, over 12244.00 frames. ], tot_loss[loss=0.1638, simple_loss=0.2529, pruned_loss=0.03741, over 2350453.30 frames. ], batch size: 32, lr: 3.32e-03, grad_scale: 16.0 2023-05-17 08:09:10,593 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.4233, 5.2372, 5.3112, 5.3816, 5.0392, 5.0625, 4.7919, 5.3183], device='cuda:0'), covar=tensor([0.0694, 0.0605, 0.0956, 0.0582, 0.1805, 0.1323, 0.0578, 0.1085], device='cuda:0'), in_proj_covar=tensor([0.0563, 0.0720, 0.0641, 0.0659, 0.0871, 0.0777, 0.0579, 0.0499], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0005, 0.0005, 0.0003, 0.0003], device='cuda:0') 2023-05-17 08:09:11,941 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=300777.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 08:09:24,027 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.691e+02 2.799e+02 3.234e+02 3.956e+02 8.201e+02, threshold=6.469e+02, percent-clipped=3.0 2023-05-17 08:09:24,773 INFO [finetune.py:992] (0/2) Epoch 17, batch 5050, loss[loss=0.1551, simple_loss=0.2549, pruned_loss=0.02761, over 12092.00 frames. ], tot_loss[loss=0.1633, simple_loss=0.2526, pruned_loss=0.03698, over 2357539.69 frames. ], batch size: 32, lr: 3.32e-03, grad_scale: 16.0 2023-05-17 08:09:30,988 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=300802.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 08:09:33,159 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.3875, 3.5031, 3.2028, 3.0642, 2.8879, 2.7008, 3.4877, 2.2844], device='cuda:0'), covar=tensor([0.0441, 0.0164, 0.0203, 0.0222, 0.0386, 0.0352, 0.0155, 0.0514], device='cuda:0'), in_proj_covar=tensor([0.0200, 0.0165, 0.0172, 0.0195, 0.0207, 0.0203, 0.0179, 0.0205], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-17 08:09:57,685 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.3888, 4.8815, 4.1888, 5.0131, 4.5836, 3.0608, 4.2623, 3.0906], device='cuda:0'), covar=tensor([0.0784, 0.0583, 0.1329, 0.0455, 0.1061, 0.1644, 0.1089, 0.3357], device='cuda:0'), in_proj_covar=tensor([0.0315, 0.0385, 0.0367, 0.0334, 0.0377, 0.0279, 0.0353, 0.0371], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-17 08:10:02,317 INFO [finetune.py:992] (0/2) Epoch 17, batch 5100, loss[loss=0.1929, simple_loss=0.2886, pruned_loss=0.04856, over 11761.00 frames. ], tot_loss[loss=0.1629, simple_loss=0.2524, pruned_loss=0.03666, over 2361510.59 frames. ], batch size: 44, lr: 3.32e-03, grad_scale: 16.0 2023-05-17 08:10:15,281 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=300863.0, num_to_drop=1, layers_to_drop={0} 2023-05-17 08:10:37,312 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.054e+02 2.723e+02 3.013e+02 3.762e+02 5.575e+02, threshold=6.027e+02, percent-clipped=0.0 2023-05-17 08:10:37,912 INFO [finetune.py:992] (0/2) Epoch 17, batch 5150, loss[loss=0.2162, simple_loss=0.3016, pruned_loss=0.06542, over 11300.00 frames. ], tot_loss[loss=0.1626, simple_loss=0.2521, pruned_loss=0.03657, over 2366220.14 frames. ], batch size: 55, lr: 3.32e-03, grad_scale: 16.0 2023-05-17 08:10:39,490 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.9234, 4.7931, 4.7869, 4.9089, 3.7982, 5.0224, 4.9690, 5.1247], device='cuda:0'), covar=tensor([0.0257, 0.0222, 0.0234, 0.0396, 0.1352, 0.0387, 0.0199, 0.0251], device='cuda:0'), in_proj_covar=tensor([0.0206, 0.0208, 0.0201, 0.0255, 0.0252, 0.0229, 0.0184, 0.0241], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0005, 0.0004, 0.0004, 0.0003, 0.0005], device='cuda:0') 2023-05-17 08:11:07,628 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.2136, 4.6153, 2.8051, 2.5810, 3.9305, 2.4835, 3.8444, 3.0397], device='cuda:0'), covar=tensor([0.0824, 0.0534, 0.1373, 0.1679, 0.0367, 0.1541, 0.0630, 0.0964], device='cuda:0'), in_proj_covar=tensor([0.0191, 0.0261, 0.0178, 0.0202, 0.0142, 0.0185, 0.0202, 0.0177], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-17 08:11:13,874 INFO [finetune.py:992] (0/2) Epoch 17, batch 5200, loss[loss=0.2, simple_loss=0.2798, pruned_loss=0.0601, over 8182.00 frames. ], tot_loss[loss=0.1627, simple_loss=0.2524, pruned_loss=0.03654, over 2361298.66 frames. ], batch size: 98, lr: 3.32e-03, grad_scale: 16.0 2023-05-17 08:11:28,282 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.4073, 5.2280, 5.3179, 5.4125, 5.0112, 5.0753, 4.8159, 5.3087], device='cuda:0'), covar=tensor([0.0789, 0.0685, 0.0956, 0.0548, 0.2051, 0.1396, 0.0548, 0.1070], device='cuda:0'), in_proj_covar=tensor([0.0564, 0.0723, 0.0643, 0.0659, 0.0874, 0.0781, 0.0580, 0.0502], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0005, 0.0005, 0.0003, 0.0003], device='cuda:0') 2023-05-17 08:11:34,750 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=300973.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 08:11:40,752 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.70 vs. limit=2.0 2023-05-17 08:11:50,238 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.690e+02 2.582e+02 2.981e+02 3.629e+02 7.540e+02, threshold=5.961e+02, percent-clipped=3.0 2023-05-17 08:11:50,984 INFO [finetune.py:992] (0/2) Epoch 17, batch 5250, loss[loss=0.1835, simple_loss=0.2713, pruned_loss=0.0479, over 8134.00 frames. ], tot_loss[loss=0.1625, simple_loss=0.2523, pruned_loss=0.0364, over 2365233.27 frames. ], batch size: 99, lr: 3.32e-03, grad_scale: 16.0 2023-05-17 08:12:26,419 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.95 vs. limit=2.0 2023-05-17 08:12:26,590 INFO [finetune.py:992] (0/2) Epoch 17, batch 5300, loss[loss=0.1672, simple_loss=0.2465, pruned_loss=0.04396, over 12291.00 frames. ], tot_loss[loss=0.1634, simple_loss=0.2531, pruned_loss=0.03686, over 2360649.05 frames. ], batch size: 33, lr: 3.32e-03, grad_scale: 16.0 2023-05-17 08:13:01,963 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.012e+02 2.644e+02 3.162e+02 3.755e+02 1.139e+03, threshold=6.323e+02, percent-clipped=5.0 2023-05-17 08:13:02,684 INFO [finetune.py:992] (0/2) Epoch 17, batch 5350, loss[loss=0.1692, simple_loss=0.2558, pruned_loss=0.04136, over 12247.00 frames. ], tot_loss[loss=0.1634, simple_loss=0.2535, pruned_loss=0.03667, over 2369055.13 frames. ], batch size: 32, lr: 3.32e-03, grad_scale: 16.0 2023-05-17 08:13:10,903 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.8942, 2.2363, 3.4214, 2.8303, 3.2651, 3.0879, 2.4019, 3.3634], device='cuda:0'), covar=tensor([0.0153, 0.0450, 0.0173, 0.0284, 0.0207, 0.0192, 0.0401, 0.0156], device='cuda:0'), in_proj_covar=tensor([0.0188, 0.0214, 0.0201, 0.0198, 0.0230, 0.0176, 0.0207, 0.0199], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-17 08:13:35,595 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=301140.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 08:13:39,025 INFO [finetune.py:992] (0/2) Epoch 17, batch 5400, loss[loss=0.1579, simple_loss=0.2531, pruned_loss=0.03134, over 12351.00 frames. ], tot_loss[loss=0.1641, simple_loss=0.2543, pruned_loss=0.03698, over 2370828.81 frames. ], batch size: 36, lr: 3.32e-03, grad_scale: 16.0 2023-05-17 08:13:44,256 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.0139, 2.5005, 3.4741, 3.9956, 3.5692, 4.0778, 3.4400, 3.0378], device='cuda:0'), covar=tensor([0.0052, 0.0384, 0.0142, 0.0056, 0.0162, 0.0066, 0.0153, 0.0342], device='cuda:0'), in_proj_covar=tensor([0.0093, 0.0127, 0.0107, 0.0082, 0.0108, 0.0119, 0.0105, 0.0143], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-17 08:13:47,179 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.9710, 2.3551, 3.5428, 2.8937, 3.3590, 3.0562, 2.4926, 3.4700], device='cuda:0'), covar=tensor([0.0157, 0.0372, 0.0168, 0.0292, 0.0173, 0.0227, 0.0365, 0.0148], device='cuda:0'), in_proj_covar=tensor([0.0189, 0.0215, 0.0202, 0.0199, 0.0231, 0.0176, 0.0207, 0.0200], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-17 08:13:48,355 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=301158.0, num_to_drop=1, layers_to_drop={3} 2023-05-17 08:13:55,067 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.0148, 2.4918, 3.6110, 2.9794, 3.4138, 3.1620, 2.6243, 3.5252], device='cuda:0'), covar=tensor([0.0217, 0.0396, 0.0210, 0.0328, 0.0226, 0.0236, 0.0379, 0.0162], device='cuda:0'), in_proj_covar=tensor([0.0190, 0.0216, 0.0203, 0.0199, 0.0231, 0.0177, 0.0208, 0.0201], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-17 08:14:00,918 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=3.13 vs. limit=5.0 2023-05-17 08:14:13,685 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.794e+02 2.605e+02 3.084e+02 3.605e+02 6.767e+02, threshold=6.167e+02, percent-clipped=2.0 2023-05-17 08:14:14,361 INFO [finetune.py:992] (0/2) Epoch 17, batch 5450, loss[loss=0.1586, simple_loss=0.2527, pruned_loss=0.03222, over 12202.00 frames. ], tot_loss[loss=0.1643, simple_loss=0.2546, pruned_loss=0.03701, over 2373446.23 frames. ], batch size: 35, lr: 3.31e-03, grad_scale: 16.0 2023-05-17 08:14:19,091 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=301201.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 08:14:50,775 INFO [finetune.py:992] (0/2) Epoch 17, batch 5500, loss[loss=0.1747, simple_loss=0.2597, pruned_loss=0.04487, over 12118.00 frames. ], tot_loss[loss=0.1639, simple_loss=0.2542, pruned_loss=0.03684, over 2380817.70 frames. ], batch size: 39, lr: 3.31e-03, grad_scale: 16.0 2023-05-17 08:14:52,962 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.4539, 5.2757, 5.3778, 5.4301, 5.0508, 5.1269, 4.8061, 5.3563], device='cuda:0'), covar=tensor([0.0657, 0.0610, 0.0796, 0.0543, 0.1750, 0.1214, 0.0569, 0.1071], device='cuda:0'), in_proj_covar=tensor([0.0562, 0.0722, 0.0642, 0.0656, 0.0872, 0.0776, 0.0579, 0.0502], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0005, 0.0005, 0.0003, 0.0003], device='cuda:0') 2023-05-17 08:15:10,730 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=301273.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 08:15:24,647 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=3.33 vs. limit=5.0 2023-05-17 08:15:27,091 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.723e+02 2.504e+02 2.789e+02 3.471e+02 7.035e+02, threshold=5.578e+02, percent-clipped=1.0 2023-05-17 08:15:27,110 INFO [finetune.py:992] (0/2) Epoch 17, batch 5550, loss[loss=0.1685, simple_loss=0.26, pruned_loss=0.03848, over 11649.00 frames. ], tot_loss[loss=0.1637, simple_loss=0.2536, pruned_loss=0.0369, over 2385933.33 frames. ], batch size: 48, lr: 3.31e-03, grad_scale: 8.0 2023-05-17 08:15:45,946 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=301321.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 08:16:02,977 INFO [finetune.py:992] (0/2) Epoch 17, batch 5600, loss[loss=0.154, simple_loss=0.2389, pruned_loss=0.03453, over 12365.00 frames. ], tot_loss[loss=0.1636, simple_loss=0.2532, pruned_loss=0.03701, over 2383455.48 frames. ], batch size: 31, lr: 3.31e-03, grad_scale: 8.0 2023-05-17 08:16:38,974 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.613e+02 2.655e+02 3.111e+02 3.833e+02 7.587e+02, threshold=6.222e+02, percent-clipped=4.0 2023-05-17 08:16:38,993 INFO [finetune.py:992] (0/2) Epoch 17, batch 5650, loss[loss=0.1271, simple_loss=0.2099, pruned_loss=0.02218, over 12357.00 frames. ], tot_loss[loss=0.1641, simple_loss=0.2532, pruned_loss=0.03746, over 2375976.79 frames. ], batch size: 30, lr: 3.31e-03, grad_scale: 8.0 2023-05-17 08:16:45,467 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.37 vs. limit=2.0 2023-05-17 08:16:58,823 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.4272, 4.2122, 4.3366, 4.4629, 3.2065, 3.9937, 2.7848, 4.2434], device='cuda:0'), covar=tensor([0.1496, 0.0613, 0.0731, 0.0600, 0.1055, 0.0592, 0.1601, 0.1115], device='cuda:0'), in_proj_covar=tensor([0.0233, 0.0272, 0.0302, 0.0365, 0.0245, 0.0248, 0.0265, 0.0377], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-17 08:17:05,253 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=301430.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 08:17:14,221 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.93 vs. limit=2.0 2023-05-17 08:17:15,909 INFO [finetune.py:992] (0/2) Epoch 17, batch 5700, loss[loss=0.1701, simple_loss=0.2574, pruned_loss=0.04143, over 12205.00 frames. ], tot_loss[loss=0.1637, simple_loss=0.2528, pruned_loss=0.0373, over 2372513.84 frames. ], batch size: 35, lr: 3.31e-03, grad_scale: 8.0 2023-05-17 08:17:25,523 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=301458.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 08:17:49,046 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=301491.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 08:17:51,656 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.876e+02 2.682e+02 3.114e+02 3.562e+02 5.424e+02, threshold=6.228e+02, percent-clipped=0.0 2023-05-17 08:17:51,674 INFO [finetune.py:992] (0/2) Epoch 17, batch 5750, loss[loss=0.1656, simple_loss=0.2655, pruned_loss=0.03286, over 12199.00 frames. ], tot_loss[loss=0.1633, simple_loss=0.2529, pruned_loss=0.03687, over 2380743.15 frames. ], batch size: 35, lr: 3.31e-03, grad_scale: 8.0 2023-05-17 08:17:52,446 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=301496.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 08:17:59,732 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=301506.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 08:18:24,021 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.0601, 5.8075, 5.3711, 5.2971, 5.9211, 5.3310, 5.2919, 5.3368], device='cuda:0'), covar=tensor([0.1585, 0.0971, 0.1254, 0.2059, 0.0889, 0.2004, 0.1958, 0.1298], device='cuda:0'), in_proj_covar=tensor([0.0368, 0.0509, 0.0415, 0.0461, 0.0477, 0.0451, 0.0414, 0.0402], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-17 08:18:28,111 INFO [finetune.py:992] (0/2) Epoch 17, batch 5800, loss[loss=0.1597, simple_loss=0.2516, pruned_loss=0.03387, over 11650.00 frames. ], tot_loss[loss=0.1633, simple_loss=0.2529, pruned_loss=0.03682, over 2379684.05 frames. ], batch size: 48, lr: 3.31e-03, grad_scale: 8.0 2023-05-17 08:19:02,635 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.7267, 3.6868, 3.3448, 3.2441, 2.9659, 2.8653, 3.7801, 2.5193], device='cuda:0'), covar=tensor([0.0371, 0.0155, 0.0208, 0.0225, 0.0395, 0.0379, 0.0126, 0.0485], device='cuda:0'), in_proj_covar=tensor([0.0201, 0.0166, 0.0172, 0.0196, 0.0207, 0.0204, 0.0179, 0.0206], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-17 08:19:04,502 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.562e+02 2.491e+02 2.825e+02 3.428e+02 1.107e+03, threshold=5.650e+02, percent-clipped=1.0 2023-05-17 08:19:04,521 INFO [finetune.py:992] (0/2) Epoch 17, batch 5850, loss[loss=0.1524, simple_loss=0.2325, pruned_loss=0.03613, over 12244.00 frames. ], tot_loss[loss=0.1634, simple_loss=0.2528, pruned_loss=0.03701, over 2377523.70 frames. ], batch size: 28, lr: 3.31e-03, grad_scale: 8.0 2023-05-17 08:19:32,233 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.7550, 2.8912, 4.7639, 4.8445, 2.8286, 2.6425, 2.9990, 2.2532], device='cuda:0'), covar=tensor([0.1722, 0.3244, 0.0428, 0.0427, 0.1451, 0.2518, 0.3056, 0.4478], device='cuda:0'), in_proj_covar=tensor([0.0312, 0.0400, 0.0282, 0.0308, 0.0285, 0.0325, 0.0408, 0.0387], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-17 08:19:40,608 INFO [finetune.py:992] (0/2) Epoch 17, batch 5900, loss[loss=0.1391, simple_loss=0.2315, pruned_loss=0.02328, over 12260.00 frames. ], tot_loss[loss=0.1625, simple_loss=0.2521, pruned_loss=0.03646, over 2384155.23 frames. ], batch size: 32, lr: 3.31e-03, grad_scale: 8.0 2023-05-17 08:19:40,746 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.8881, 5.8893, 5.6764, 5.2553, 5.2208, 5.8576, 5.4441, 5.2650], device='cuda:0'), covar=tensor([0.0861, 0.1050, 0.0743, 0.1743, 0.0699, 0.0837, 0.1610, 0.1233], device='cuda:0'), in_proj_covar=tensor([0.0651, 0.0583, 0.0536, 0.0657, 0.0437, 0.0756, 0.0807, 0.0585], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002], device='cuda:0') 2023-05-17 08:19:46,391 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.9286, 5.9576, 5.7333, 5.2333, 5.1908, 5.8981, 5.5034, 5.3126], device='cuda:0'), covar=tensor([0.0837, 0.0940, 0.0689, 0.1662, 0.0708, 0.0746, 0.1553, 0.1161], device='cuda:0'), in_proj_covar=tensor([0.0653, 0.0585, 0.0538, 0.0659, 0.0439, 0.0757, 0.0809, 0.0587], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0004, 0.0004, 0.0003], device='cuda:0') 2023-05-17 08:20:07,872 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.7386, 3.0816, 3.7727, 4.7027, 4.0198, 4.7752, 3.9711, 3.4780], device='cuda:0'), covar=tensor([0.0030, 0.0356, 0.0156, 0.0041, 0.0117, 0.0056, 0.0140, 0.0316], device='cuda:0'), in_proj_covar=tensor([0.0093, 0.0128, 0.0109, 0.0083, 0.0109, 0.0120, 0.0106, 0.0144], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-17 08:20:16,723 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.853e+02 2.751e+02 3.117e+02 3.766e+02 7.477e+02, threshold=6.233e+02, percent-clipped=4.0 2023-05-17 08:20:16,742 INFO [finetune.py:992] (0/2) Epoch 17, batch 5950, loss[loss=0.1792, simple_loss=0.274, pruned_loss=0.04214, over 12149.00 frames. ], tot_loss[loss=0.1639, simple_loss=0.2535, pruned_loss=0.03713, over 2373399.04 frames. ], batch size: 36, lr: 3.31e-03, grad_scale: 8.0 2023-05-17 08:20:31,050 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.6802, 5.4135, 4.9874, 5.0206, 5.5180, 4.8434, 4.9504, 4.8758], device='cuda:0'), covar=tensor([0.1482, 0.1074, 0.1410, 0.1836, 0.1046, 0.2177, 0.1938, 0.1471], device='cuda:0'), in_proj_covar=tensor([0.0369, 0.0509, 0.0413, 0.0456, 0.0477, 0.0450, 0.0412, 0.0401], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-17 08:20:51,975 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=301743.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 08:20:53,220 INFO [finetune.py:992] (0/2) Epoch 17, batch 6000, loss[loss=0.1636, simple_loss=0.2544, pruned_loss=0.03636, over 12362.00 frames. ], tot_loss[loss=0.1643, simple_loss=0.2539, pruned_loss=0.03733, over 2377578.01 frames. ], batch size: 36, lr: 3.31e-03, grad_scale: 8.0 2023-05-17 08:20:53,221 INFO [finetune.py:1017] (0/2) Computing validation loss 2023-05-17 08:21:10,577 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.1392, 2.9911, 3.0076, 3.2644, 2.3960, 3.0500, 2.4788, 2.8925], device='cuda:0'), covar=tensor([0.1624, 0.0880, 0.0779, 0.0561, 0.1112, 0.0768, 0.1637, 0.0898], device='cuda:0'), in_proj_covar=tensor([0.0234, 0.0273, 0.0303, 0.0366, 0.0245, 0.0248, 0.0266, 0.0378], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-17 08:21:12,096 INFO [finetune.py:1026] (0/2) Epoch 17, validation: loss=0.3119, simple_loss=0.3879, pruned_loss=0.118, over 1020973.00 frames. 2023-05-17 08:21:12,096 INFO [finetune.py:1027] (0/2) Maximum memory allocated so far is 12827MB 2023-05-17 08:21:40,783 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.0991, 5.8936, 5.4564, 5.4574, 6.0064, 5.2391, 5.4136, 5.4703], device='cuda:0'), covar=tensor([0.1600, 0.0966, 0.1229, 0.1824, 0.0970, 0.2351, 0.1956, 0.1364], device='cuda:0'), in_proj_covar=tensor([0.0372, 0.0512, 0.0417, 0.0459, 0.0481, 0.0454, 0.0415, 0.0404], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-17 08:21:42,135 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=301786.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 08:21:48,305 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.904e+02 2.650e+02 3.139e+02 3.810e+02 6.475e+02, threshold=6.278e+02, percent-clipped=1.0 2023-05-17 08:21:48,324 INFO [finetune.py:992] (0/2) Epoch 17, batch 6050, loss[loss=0.1823, simple_loss=0.2784, pruned_loss=0.04312, over 12111.00 frames. ], tot_loss[loss=0.1643, simple_loss=0.2541, pruned_loss=0.0373, over 2377729.85 frames. ], batch size: 38, lr: 3.31e-03, grad_scale: 8.0 2023-05-17 08:21:49,109 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=301796.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 08:21:55,402 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=301804.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 08:22:02,305 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.6566, 2.9925, 3.7291, 4.5920, 3.9676, 4.6291, 3.8469, 3.4730], device='cuda:0'), covar=tensor([0.0038, 0.0398, 0.0169, 0.0056, 0.0134, 0.0081, 0.0175, 0.0375], device='cuda:0'), in_proj_covar=tensor([0.0093, 0.0128, 0.0109, 0.0083, 0.0109, 0.0120, 0.0107, 0.0145], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-17 08:22:23,751 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=301844.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 08:22:24,376 INFO [finetune.py:992] (0/2) Epoch 17, batch 6100, loss[loss=0.1878, simple_loss=0.2831, pruned_loss=0.04627, over 12345.00 frames. ], tot_loss[loss=0.165, simple_loss=0.2545, pruned_loss=0.03777, over 2370456.07 frames. ], batch size: 35, lr: 3.31e-03, grad_scale: 8.0 2023-05-17 08:22:39,710 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.1442, 2.7202, 3.7773, 3.1913, 3.6240, 3.3249, 2.7499, 3.6769], device='cuda:0'), covar=tensor([0.0162, 0.0349, 0.0152, 0.0248, 0.0153, 0.0195, 0.0355, 0.0147], device='cuda:0'), in_proj_covar=tensor([0.0189, 0.0215, 0.0202, 0.0197, 0.0230, 0.0176, 0.0207, 0.0200], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-17 08:22:44,804 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.1301, 2.1124, 2.7641, 3.1764, 2.2434, 3.2568, 3.0943, 3.3030], device='cuda:0'), covar=tensor([0.0229, 0.1202, 0.0485, 0.0213, 0.1114, 0.0377, 0.0372, 0.0166], device='cuda:0'), in_proj_covar=tensor([0.0124, 0.0205, 0.0184, 0.0123, 0.0190, 0.0181, 0.0180, 0.0127], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-17 08:23:00,276 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.900e+02 2.658e+02 2.989e+02 3.684e+02 9.440e+02, threshold=5.978e+02, percent-clipped=2.0 2023-05-17 08:23:00,295 INFO [finetune.py:992] (0/2) Epoch 17, batch 6150, loss[loss=0.1754, simple_loss=0.2647, pruned_loss=0.04299, over 12210.00 frames. ], tot_loss[loss=0.1642, simple_loss=0.2535, pruned_loss=0.03741, over 2370413.51 frames. ], batch size: 35, lr: 3.31e-03, grad_scale: 8.0 2023-05-17 08:23:03,304 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=301899.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 08:23:07,620 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=301905.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 08:23:36,862 INFO [finetune.py:992] (0/2) Epoch 17, batch 6200, loss[loss=0.1647, simple_loss=0.2606, pruned_loss=0.03438, over 12281.00 frames. ], tot_loss[loss=0.1643, simple_loss=0.2538, pruned_loss=0.03746, over 2373378.04 frames. ], batch size: 37, lr: 3.31e-03, grad_scale: 8.0 2023-05-17 08:23:48,289 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=301960.0, num_to_drop=1, layers_to_drop={3} 2023-05-17 08:23:52,500 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=301966.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 08:24:12,867 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.900e+02 2.627e+02 3.010e+02 3.487e+02 9.928e+02, threshold=6.019e+02, percent-clipped=2.0 2023-05-17 08:24:12,887 INFO [finetune.py:992] (0/2) Epoch 17, batch 6250, loss[loss=0.1413, simple_loss=0.233, pruned_loss=0.02485, over 12358.00 frames. ], tot_loss[loss=0.1637, simple_loss=0.2534, pruned_loss=0.03696, over 2378943.80 frames. ], batch size: 30, lr: 3.31e-03, grad_scale: 8.0 2023-05-17 08:24:16,467 INFO [checkpoint.py:75] (0/2) Saving checkpoint to pruned_transducer_stateless7/exp_giga_finetune/checkpoint-202000.pt 2023-05-17 08:24:26,062 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.2014, 5.1873, 4.9762, 4.5784, 4.7095, 5.1211, 4.8463, 4.6115], device='cuda:0'), covar=tensor([0.0914, 0.1075, 0.0851, 0.1646, 0.1319, 0.0945, 0.1654, 0.1245], device='cuda:0'), in_proj_covar=tensor([0.0658, 0.0588, 0.0539, 0.0667, 0.0443, 0.0757, 0.0813, 0.0588], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0004, 0.0004, 0.0003], device='cuda:0') 2023-05-17 08:24:31,274 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.9944, 4.4076, 3.9130, 4.7291, 4.1989, 2.7790, 4.0925, 2.9022], device='cuda:0'), covar=tensor([0.0964, 0.0778, 0.1565, 0.0579, 0.1336, 0.1876, 0.1085, 0.3417], device='cuda:0'), in_proj_covar=tensor([0.0319, 0.0389, 0.0369, 0.0339, 0.0382, 0.0284, 0.0358, 0.0376], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-17 08:24:51,784 INFO [finetune.py:992] (0/2) Epoch 17, batch 6300, loss[loss=0.1771, simple_loss=0.2774, pruned_loss=0.03841, over 12193.00 frames. ], tot_loss[loss=0.1635, simple_loss=0.2531, pruned_loss=0.03691, over 2378159.89 frames. ], batch size: 35, lr: 3.31e-03, grad_scale: 8.0 2023-05-17 08:25:06,171 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([1.9489, 4.1333, 4.2216, 4.2865, 2.6551, 4.1469, 2.7903, 4.1744], device='cuda:0'), covar=tensor([0.1837, 0.0665, 0.0735, 0.0595, 0.1311, 0.0548, 0.1756, 0.1022], device='cuda:0'), in_proj_covar=tensor([0.0233, 0.0272, 0.0301, 0.0364, 0.0244, 0.0247, 0.0265, 0.0375], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-17 08:25:21,807 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=302086.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 08:25:28,041 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.737e+02 2.542e+02 3.035e+02 3.884e+02 6.196e+02, threshold=6.069e+02, percent-clipped=2.0 2023-05-17 08:25:28,060 INFO [finetune.py:992] (0/2) Epoch 17, batch 6350, loss[loss=0.1442, simple_loss=0.2387, pruned_loss=0.02486, over 12275.00 frames. ], tot_loss[loss=0.1633, simple_loss=0.2526, pruned_loss=0.03697, over 2376314.68 frames. ], batch size: 37, lr: 3.31e-03, grad_scale: 8.0 2023-05-17 08:25:30,850 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=302099.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 08:25:56,215 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=302134.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 08:26:04,109 INFO [finetune.py:992] (0/2) Epoch 17, batch 6400, loss[loss=0.1554, simple_loss=0.2392, pruned_loss=0.03579, over 12200.00 frames. ], tot_loss[loss=0.1638, simple_loss=0.2531, pruned_loss=0.03721, over 2377452.69 frames. ], batch size: 29, lr: 3.31e-03, grad_scale: 8.0 2023-05-17 08:26:39,312 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.621e+02 2.608e+02 3.103e+02 3.901e+02 7.929e+02, threshold=6.206e+02, percent-clipped=1.0 2023-05-17 08:26:39,331 INFO [finetune.py:992] (0/2) Epoch 17, batch 6450, loss[loss=0.191, simple_loss=0.2789, pruned_loss=0.05152, over 11854.00 frames. ], tot_loss[loss=0.1638, simple_loss=0.2533, pruned_loss=0.03713, over 2380436.33 frames. ], batch size: 44, lr: 3.31e-03, grad_scale: 8.0 2023-05-17 08:26:43,380 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=302200.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 08:26:46,976 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.7504, 3.7566, 3.3732, 3.2725, 3.0675, 2.9349, 3.7935, 2.5359], device='cuda:0'), covar=tensor([0.0396, 0.0171, 0.0205, 0.0225, 0.0354, 0.0388, 0.0129, 0.0494], device='cuda:0'), in_proj_covar=tensor([0.0204, 0.0168, 0.0174, 0.0200, 0.0209, 0.0207, 0.0182, 0.0210], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-17 08:27:16,169 INFO [finetune.py:992] (0/2) Epoch 17, batch 6500, loss[loss=0.1441, simple_loss=0.2278, pruned_loss=0.03016, over 12359.00 frames. ], tot_loss[loss=0.1639, simple_loss=0.2535, pruned_loss=0.03715, over 2377198.31 frames. ], batch size: 30, lr: 3.31e-03, grad_scale: 8.0 2023-05-17 08:27:24,194 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=302255.0, num_to_drop=1, layers_to_drop={1} 2023-05-17 08:27:28,449 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=302261.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 08:27:28,555 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=302261.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 08:27:52,639 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.966e+02 2.722e+02 3.140e+02 3.968e+02 8.913e+02, threshold=6.279e+02, percent-clipped=2.0 2023-05-17 08:27:52,658 INFO [finetune.py:992] (0/2) Epoch 17, batch 6550, loss[loss=0.1521, simple_loss=0.2363, pruned_loss=0.03391, over 12033.00 frames. ], tot_loss[loss=0.1639, simple_loss=0.2541, pruned_loss=0.03691, over 2380344.05 frames. ], batch size: 31, lr: 3.31e-03, grad_scale: 8.0 2023-05-17 08:27:53,505 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=302296.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 08:27:58,528 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.66 vs. limit=2.0 2023-05-17 08:28:27,651 INFO [finetune.py:992] (0/2) Epoch 17, batch 6600, loss[loss=0.1612, simple_loss=0.2628, pruned_loss=0.02984, over 12289.00 frames. ], tot_loss[loss=0.1641, simple_loss=0.2542, pruned_loss=0.03698, over 2386218.21 frames. ], batch size: 37, lr: 3.31e-03, grad_scale: 8.0 2023-05-17 08:28:36,789 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=302357.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 08:28:38,304 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.2882, 4.1338, 4.2016, 4.5423, 2.9885, 4.0370, 2.7268, 4.0560], device='cuda:0'), covar=tensor([0.1839, 0.0729, 0.0935, 0.0531, 0.1296, 0.0640, 0.1921, 0.1278], device='cuda:0'), in_proj_covar=tensor([0.0233, 0.0271, 0.0301, 0.0363, 0.0244, 0.0247, 0.0264, 0.0374], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-17 08:29:03,989 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.795e+02 2.638e+02 3.054e+02 3.796e+02 1.353e+03, threshold=6.108e+02, percent-clipped=2.0 2023-05-17 08:29:04,008 INFO [finetune.py:992] (0/2) Epoch 17, batch 6650, loss[loss=0.1897, simple_loss=0.2827, pruned_loss=0.04833, over 12111.00 frames. ], tot_loss[loss=0.1646, simple_loss=0.2548, pruned_loss=0.0372, over 2387689.17 frames. ], batch size: 38, lr: 3.31e-03, grad_scale: 8.0 2023-05-17 08:29:07,683 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=302399.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 08:29:40,253 INFO [finetune.py:992] (0/2) Epoch 17, batch 6700, loss[loss=0.1816, simple_loss=0.2713, pruned_loss=0.04595, over 12364.00 frames. ], tot_loss[loss=0.1642, simple_loss=0.254, pruned_loss=0.03718, over 2381109.81 frames. ], batch size: 38, lr: 3.31e-03, grad_scale: 8.0 2023-05-17 08:29:41,732 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=302447.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 08:30:03,312 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=302477.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 08:30:16,289 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.118e+02 2.707e+02 3.206e+02 3.617e+02 8.520e+02, threshold=6.411e+02, percent-clipped=1.0 2023-05-17 08:30:16,308 INFO [finetune.py:992] (0/2) Epoch 17, batch 6750, loss[loss=0.1757, simple_loss=0.2682, pruned_loss=0.04164, over 10566.00 frames. ], tot_loss[loss=0.1645, simple_loss=0.2546, pruned_loss=0.03717, over 2387716.92 frames. ], batch size: 68, lr: 3.31e-03, grad_scale: 8.0 2023-05-17 08:30:46,970 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=302538.0, num_to_drop=1, layers_to_drop={3} 2023-05-17 08:30:52,451 INFO [finetune.py:992] (0/2) Epoch 17, batch 6800, loss[loss=0.1633, simple_loss=0.2616, pruned_loss=0.03252, over 12136.00 frames. ], tot_loss[loss=0.1653, simple_loss=0.2551, pruned_loss=0.03778, over 2370611.88 frames. ], batch size: 36, lr: 3.31e-03, grad_scale: 8.0 2023-05-17 08:30:59,015 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.2459, 4.3099, 4.2461, 4.5240, 3.1621, 3.9079, 2.7994, 4.2789], device='cuda:0'), covar=tensor([0.1645, 0.0608, 0.0848, 0.0665, 0.1151, 0.0686, 0.1689, 0.1218], device='cuda:0'), in_proj_covar=tensor([0.0235, 0.0271, 0.0302, 0.0365, 0.0245, 0.0248, 0.0265, 0.0375], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-17 08:30:59,606 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=302555.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 08:31:00,311 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=302556.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 08:31:03,943 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=302561.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 08:31:16,671 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.48 vs. limit=2.0 2023-05-17 08:31:28,298 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.834e+02 2.614e+02 3.177e+02 3.949e+02 9.293e+02, threshold=6.353e+02, percent-clipped=5.0 2023-05-17 08:31:28,316 INFO [finetune.py:992] (0/2) Epoch 17, batch 6850, loss[loss=0.1426, simple_loss=0.2234, pruned_loss=0.03091, over 11802.00 frames. ], tot_loss[loss=0.1644, simple_loss=0.254, pruned_loss=0.03742, over 2370827.99 frames. ], batch size: 26, lr: 3.31e-03, grad_scale: 8.0 2023-05-17 08:31:34,235 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=302603.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 08:31:38,536 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=302609.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 08:31:55,360 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=302632.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 08:32:05,132 INFO [finetune.py:992] (0/2) Epoch 17, batch 6900, loss[loss=0.1707, simple_loss=0.2726, pruned_loss=0.03435, over 12286.00 frames. ], tot_loss[loss=0.1654, simple_loss=0.2549, pruned_loss=0.03797, over 2360829.85 frames. ], batch size: 37, lr: 3.31e-03, grad_scale: 8.0 2023-05-17 08:32:10,109 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=302652.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 08:32:38,346 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.7534, 2.4255, 3.1947, 2.7579, 3.0778, 2.9380, 2.4241, 3.1463], device='cuda:0'), covar=tensor([0.0149, 0.0383, 0.0192, 0.0304, 0.0209, 0.0240, 0.0419, 0.0190], device='cuda:0'), in_proj_covar=tensor([0.0191, 0.0217, 0.0205, 0.0200, 0.0233, 0.0178, 0.0209, 0.0201], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-17 08:32:40,471 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=302693.0, num_to_drop=1, layers_to_drop={2} 2023-05-17 08:32:41,565 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.721e+02 2.663e+02 3.007e+02 3.511e+02 7.299e+02, threshold=6.014e+02, percent-clipped=1.0 2023-05-17 08:32:41,584 INFO [finetune.py:992] (0/2) Epoch 17, batch 6950, loss[loss=0.2421, simple_loss=0.3256, pruned_loss=0.07924, over 7830.00 frames. ], tot_loss[loss=0.1648, simple_loss=0.2546, pruned_loss=0.03746, over 2360405.05 frames. ], batch size: 98, lr: 3.30e-03, grad_scale: 8.0 2023-05-17 08:33:17,306 INFO [finetune.py:992] (0/2) Epoch 17, batch 7000, loss[loss=0.1712, simple_loss=0.2664, pruned_loss=0.03802, over 12307.00 frames. ], tot_loss[loss=0.165, simple_loss=0.2548, pruned_loss=0.03757, over 2362011.93 frames. ], batch size: 34, lr: 3.30e-03, grad_scale: 8.0 2023-05-17 08:33:39,110 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.35 vs. limit=2.0 2023-05-17 08:33:53,473 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.907e+02 2.681e+02 3.106e+02 3.603e+02 9.207e+02, threshold=6.212e+02, percent-clipped=1.0 2023-05-17 08:33:53,491 INFO [finetune.py:992] (0/2) Epoch 17, batch 7050, loss[loss=0.1785, simple_loss=0.2746, pruned_loss=0.04115, over 11908.00 frames. ], tot_loss[loss=0.1642, simple_loss=0.2539, pruned_loss=0.03729, over 2363831.27 frames. ], batch size: 44, lr: 3.30e-03, grad_scale: 8.0 2023-05-17 08:34:21,865 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=302833.0, num_to_drop=1, layers_to_drop={1} 2023-05-17 08:34:23,413 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.4869, 2.6472, 3.3181, 4.3993, 2.4918, 4.5054, 4.5270, 4.6203], device='cuda:0'), covar=tensor([0.0196, 0.1309, 0.0466, 0.0155, 0.1349, 0.0209, 0.0174, 0.0114], device='cuda:0'), in_proj_covar=tensor([0.0126, 0.0206, 0.0186, 0.0124, 0.0192, 0.0182, 0.0181, 0.0128], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-17 08:34:30,450 INFO [finetune.py:992] (0/2) Epoch 17, batch 7100, loss[loss=0.1543, simple_loss=0.2534, pruned_loss=0.02759, over 11870.00 frames. ], tot_loss[loss=0.1639, simple_loss=0.2534, pruned_loss=0.03715, over 2363057.54 frames. ], batch size: 44, lr: 3.30e-03, grad_scale: 8.0 2023-05-17 08:34:31,322 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.9271, 5.9402, 5.7551, 5.2614, 5.1349, 5.8812, 5.4532, 5.2888], device='cuda:0'), covar=tensor([0.0979, 0.0999, 0.0673, 0.1627, 0.0809, 0.0744, 0.1513, 0.1108], device='cuda:0'), in_proj_covar=tensor([0.0656, 0.0583, 0.0535, 0.0660, 0.0442, 0.0754, 0.0811, 0.0585], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002], device='cuda:0') 2023-05-17 08:34:38,721 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=302856.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 08:34:56,759 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=302881.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 08:35:06,439 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.809e+02 2.672e+02 3.103e+02 3.650e+02 5.427e+02, threshold=6.206e+02, percent-clipped=0.0 2023-05-17 08:35:06,457 INFO [finetune.py:992] (0/2) Epoch 17, batch 7150, loss[loss=0.1712, simple_loss=0.2579, pruned_loss=0.04227, over 12349.00 frames. ], tot_loss[loss=0.1641, simple_loss=0.2541, pruned_loss=0.03709, over 2365596.22 frames. ], batch size: 36, lr: 3.30e-03, grad_scale: 8.0 2023-05-17 08:35:13,169 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=302904.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 08:35:16,351 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.0769, 4.3817, 3.8429, 4.5208, 4.1440, 2.8368, 3.9943, 2.9103], device='cuda:0'), covar=tensor([0.0861, 0.0837, 0.1668, 0.0749, 0.1149, 0.1815, 0.1200, 0.3525], device='cuda:0'), in_proj_covar=tensor([0.0317, 0.0388, 0.0370, 0.0339, 0.0381, 0.0282, 0.0355, 0.0373], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-17 08:35:22,112 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.93 vs. limit=2.0 2023-05-17 08:35:41,177 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=302942.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 08:35:43,089 INFO [finetune.py:992] (0/2) Epoch 17, batch 7200, loss[loss=0.1719, simple_loss=0.2725, pruned_loss=0.0357, over 12170.00 frames. ], tot_loss[loss=0.1643, simple_loss=0.2538, pruned_loss=0.03737, over 2355766.44 frames. ], batch size: 36, lr: 3.30e-03, grad_scale: 8.0 2023-05-17 08:35:47,973 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=302952.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 08:36:10,873 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.3749, 5.2105, 5.3292, 5.3872, 5.0048, 5.0501, 4.7553, 5.3019], device='cuda:0'), covar=tensor([0.0736, 0.0531, 0.0766, 0.0500, 0.1788, 0.1375, 0.0565, 0.1027], device='cuda:0'), in_proj_covar=tensor([0.0564, 0.0723, 0.0639, 0.0660, 0.0879, 0.0781, 0.0587, 0.0502], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0005, 0.0005, 0.0004, 0.0003], device='cuda:0') 2023-05-17 08:36:14,296 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=302988.0, num_to_drop=1, layers_to_drop={2} 2023-05-17 08:36:17,204 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.4880, 3.4466, 3.1643, 3.0744, 2.8379, 2.6776, 3.4526, 2.3287], device='cuda:0'), covar=tensor([0.0462, 0.0194, 0.0204, 0.0224, 0.0446, 0.0434, 0.0191, 0.0570], device='cuda:0'), in_proj_covar=tensor([0.0202, 0.0166, 0.0171, 0.0197, 0.0206, 0.0205, 0.0180, 0.0209], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-17 08:36:19,191 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.750e+02 2.698e+02 3.189e+02 3.662e+02 7.824e+02, threshold=6.378e+02, percent-clipped=3.0 2023-05-17 08:36:19,210 INFO [finetune.py:992] (0/2) Epoch 17, batch 7250, loss[loss=0.1556, simple_loss=0.2599, pruned_loss=0.0256, over 12332.00 frames. ], tot_loss[loss=0.1648, simple_loss=0.2547, pruned_loss=0.03746, over 2363536.57 frames. ], batch size: 36, lr: 3.30e-03, grad_scale: 8.0 2023-05-17 08:36:20,048 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([6.1198, 6.1464, 5.8534, 5.4880, 5.3027, 6.0323, 5.5998, 5.3757], device='cuda:0'), covar=tensor([0.0696, 0.0831, 0.0640, 0.1627, 0.0678, 0.0813, 0.1787, 0.1156], device='cuda:0'), in_proj_covar=tensor([0.0658, 0.0583, 0.0538, 0.0663, 0.0444, 0.0755, 0.0815, 0.0586], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0004, 0.0004, 0.0002], device='cuda:0') 2023-05-17 08:36:22,989 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=303000.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 08:36:54,922 INFO [finetune.py:992] (0/2) Epoch 17, batch 7300, loss[loss=0.1643, simple_loss=0.2531, pruned_loss=0.03772, over 12345.00 frames. ], tot_loss[loss=0.165, simple_loss=0.2546, pruned_loss=0.03765, over 2363167.15 frames. ], batch size: 31, lr: 3.30e-03, grad_scale: 8.0 2023-05-17 08:37:00,242 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=303052.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 08:37:31,400 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.014e+02 2.740e+02 3.135e+02 3.845e+02 6.476e+02, threshold=6.269e+02, percent-clipped=1.0 2023-05-17 08:37:31,418 INFO [finetune.py:992] (0/2) Epoch 17, batch 7350, loss[loss=0.1767, simple_loss=0.2686, pruned_loss=0.04237, over 11683.00 frames. ], tot_loss[loss=0.1649, simple_loss=0.2545, pruned_loss=0.03762, over 2351482.93 frames. ], batch size: 48, lr: 3.30e-03, grad_scale: 8.0 2023-05-17 08:37:39,785 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.89 vs. limit=2.0 2023-05-17 08:37:44,445 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.7737, 3.2585, 5.1362, 2.8252, 2.6974, 3.8119, 3.1432, 3.8358], device='cuda:0'), covar=tensor([0.0479, 0.1307, 0.0320, 0.1216, 0.2204, 0.1603, 0.1511, 0.1229], device='cuda:0'), in_proj_covar=tensor([0.0240, 0.0239, 0.0259, 0.0186, 0.0240, 0.0299, 0.0228, 0.0271], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-17 08:37:45,081 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=303113.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 08:37:59,284 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=303133.0, num_to_drop=1, layers_to_drop={1} 2023-05-17 08:38:07,520 INFO [finetune.py:992] (0/2) Epoch 17, batch 7400, loss[loss=0.1512, simple_loss=0.2449, pruned_loss=0.02874, over 12028.00 frames. ], tot_loss[loss=0.1647, simple_loss=0.2544, pruned_loss=0.0375, over 2357913.57 frames. ], batch size: 31, lr: 3.30e-03, grad_scale: 8.0 2023-05-17 08:38:07,949 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.31 vs. limit=2.0 2023-05-17 08:38:17,064 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.2256, 4.3661, 2.7186, 2.5797, 3.8250, 2.5779, 3.7399, 2.9926], device='cuda:0'), covar=tensor([0.0781, 0.0664, 0.1247, 0.1604, 0.0318, 0.1448, 0.0578, 0.0992], device='cuda:0'), in_proj_covar=tensor([0.0194, 0.0266, 0.0180, 0.0205, 0.0145, 0.0188, 0.0206, 0.0180], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-17 08:38:33,308 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=303181.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 08:38:43,009 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.637e+02 2.618e+02 2.999e+02 3.673e+02 1.361e+03, threshold=5.997e+02, percent-clipped=1.0 2023-05-17 08:38:43,040 INFO [finetune.py:992] (0/2) Epoch 17, batch 7450, loss[loss=0.1473, simple_loss=0.2304, pruned_loss=0.03211, over 12261.00 frames. ], tot_loss[loss=0.1645, simple_loss=0.2543, pruned_loss=0.03729, over 2364084.68 frames. ], batch size: 32, lr: 3.30e-03, grad_scale: 8.0 2023-05-17 08:39:13,945 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=303237.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 08:39:19,438 INFO [finetune.py:992] (0/2) Epoch 17, batch 7500, loss[loss=0.2005, simple_loss=0.2928, pruned_loss=0.0541, over 12135.00 frames. ], tot_loss[loss=0.1653, simple_loss=0.2551, pruned_loss=0.03778, over 2357536.13 frames. ], batch size: 38, lr: 3.30e-03, grad_scale: 8.0 2023-05-17 08:39:22,528 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.0590, 3.9088, 4.0553, 3.7758, 3.9176, 3.7388, 4.0564, 3.6580], device='cuda:0'), covar=tensor([0.0362, 0.0408, 0.0336, 0.0256, 0.0389, 0.0339, 0.0290, 0.1396], device='cuda:0'), in_proj_covar=tensor([0.0281, 0.0287, 0.0312, 0.0282, 0.0281, 0.0280, 0.0257, 0.0231], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-17 08:39:29,544 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=303258.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 08:39:32,120 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.83 vs. limit=2.0 2023-05-17 08:39:51,144 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=303288.0, num_to_drop=1, layers_to_drop={1} 2023-05-17 08:39:55,782 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.675e+02 2.593e+02 3.048e+02 3.585e+02 8.586e+02, threshold=6.095e+02, percent-clipped=3.0 2023-05-17 08:39:55,801 INFO [finetune.py:992] (0/2) Epoch 17, batch 7550, loss[loss=0.1779, simple_loss=0.2701, pruned_loss=0.04283, over 12076.00 frames. ], tot_loss[loss=0.1651, simple_loss=0.2547, pruned_loss=0.03778, over 2363724.98 frames. ], batch size: 40, lr: 3.30e-03, grad_scale: 16.0 2023-05-17 08:39:58,068 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.9381, 2.4216, 3.5684, 2.9813, 3.5406, 3.1500, 2.4175, 3.5442], device='cuda:0'), covar=tensor([0.0242, 0.0531, 0.0204, 0.0359, 0.0180, 0.0261, 0.0593, 0.0185], device='cuda:0'), in_proj_covar=tensor([0.0191, 0.0216, 0.0205, 0.0198, 0.0232, 0.0177, 0.0208, 0.0201], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-17 08:40:13,044 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=303319.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 08:40:25,347 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=303336.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 08:40:31,491 INFO [finetune.py:992] (0/2) Epoch 17, batch 7600, loss[loss=0.1738, simple_loss=0.2628, pruned_loss=0.0424, over 12127.00 frames. ], tot_loss[loss=0.1647, simple_loss=0.2544, pruned_loss=0.03749, over 2372196.41 frames. ], batch size: 39, lr: 3.30e-03, grad_scale: 16.0 2023-05-17 08:40:58,938 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.8941, 3.3395, 5.2169, 2.7160, 2.7493, 3.9311, 3.2149, 3.8844], device='cuda:0'), covar=tensor([0.0363, 0.1137, 0.0353, 0.1245, 0.2072, 0.1627, 0.1380, 0.1243], device='cuda:0'), in_proj_covar=tensor([0.0240, 0.0238, 0.0258, 0.0186, 0.0239, 0.0297, 0.0227, 0.0271], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-17 08:41:06,987 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.7506, 2.8470, 4.6279, 4.8874, 3.0263, 2.5706, 3.0297, 2.1551], device='cuda:0'), covar=tensor([0.1652, 0.3102, 0.0483, 0.0382, 0.1260, 0.2668, 0.2745, 0.4290], device='cuda:0'), in_proj_covar=tensor([0.0310, 0.0397, 0.0283, 0.0307, 0.0284, 0.0325, 0.0405, 0.0385], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-17 08:41:08,083 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.627e+02 2.479e+02 2.824e+02 3.319e+02 6.002e+02, threshold=5.649e+02, percent-clipped=0.0 2023-05-17 08:41:08,101 INFO [finetune.py:992] (0/2) Epoch 17, batch 7650, loss[loss=0.1704, simple_loss=0.2598, pruned_loss=0.04051, over 11538.00 frames. ], tot_loss[loss=0.165, simple_loss=0.2549, pruned_loss=0.0375, over 2371020.77 frames. ], batch size: 48, lr: 3.30e-03, grad_scale: 16.0 2023-05-17 08:41:14,410 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=303403.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 08:41:17,929 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=303408.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 08:41:26,631 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.6230, 2.6018, 4.4144, 4.8085, 3.1907, 2.6284, 3.0191, 1.8347], device='cuda:0'), covar=tensor([0.1979, 0.3796, 0.0581, 0.0430, 0.1195, 0.2983, 0.3092, 0.5768], device='cuda:0'), in_proj_covar=tensor([0.0310, 0.0396, 0.0282, 0.0307, 0.0283, 0.0325, 0.0405, 0.0385], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-17 08:41:44,965 INFO [finetune.py:992] (0/2) Epoch 17, batch 7700, loss[loss=0.1377, simple_loss=0.2262, pruned_loss=0.02458, over 12094.00 frames. ], tot_loss[loss=0.1652, simple_loss=0.2555, pruned_loss=0.03748, over 2369983.43 frames. ], batch size: 32, lr: 3.30e-03, grad_scale: 16.0 2023-05-17 08:41:58,590 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=303464.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 08:42:20,227 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.575e+02 2.621e+02 3.064e+02 3.825e+02 7.913e+02, threshold=6.128e+02, percent-clipped=5.0 2023-05-17 08:42:20,245 INFO [finetune.py:992] (0/2) Epoch 17, batch 7750, loss[loss=0.1432, simple_loss=0.2223, pruned_loss=0.03201, over 12351.00 frames. ], tot_loss[loss=0.1642, simple_loss=0.2544, pruned_loss=0.03699, over 2374942.30 frames. ], batch size: 30, lr: 3.30e-03, grad_scale: 16.0 2023-05-17 08:42:36,840 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([6.0815, 6.1051, 5.8114, 5.3392, 5.2213, 6.0178, 5.6073, 5.4079], device='cuda:0'), covar=tensor([0.0730, 0.0860, 0.0692, 0.1688, 0.0741, 0.0738, 0.1421, 0.0948], device='cuda:0'), in_proj_covar=tensor([0.0660, 0.0588, 0.0541, 0.0667, 0.0445, 0.0759, 0.0819, 0.0592], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0004, 0.0004, 0.0003], device='cuda:0') 2023-05-17 08:42:50,691 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=303537.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 08:42:56,408 INFO [finetune.py:992] (0/2) Epoch 17, batch 7800, loss[loss=0.1646, simple_loss=0.2578, pruned_loss=0.03566, over 12352.00 frames. ], tot_loss[loss=0.1641, simple_loss=0.2543, pruned_loss=0.03692, over 2376695.82 frames. ], batch size: 36, lr: 3.30e-03, grad_scale: 16.0 2023-05-17 08:43:17,799 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=303574.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 08:43:20,663 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.2486, 3.5619, 3.6550, 4.0754, 2.9098, 3.5832, 2.4919, 3.4651], device='cuda:0'), covar=tensor([0.1640, 0.0900, 0.0925, 0.0696, 0.1208, 0.0746, 0.1892, 0.0928], device='cuda:0'), in_proj_covar=tensor([0.0233, 0.0271, 0.0301, 0.0364, 0.0245, 0.0249, 0.0264, 0.0376], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-17 08:43:25,444 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=303585.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 08:43:32,522 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.949e+02 2.681e+02 3.081e+02 3.940e+02 6.716e+02, threshold=6.161e+02, percent-clipped=3.0 2023-05-17 08:43:32,541 INFO [finetune.py:992] (0/2) Epoch 17, batch 7850, loss[loss=0.195, simple_loss=0.2822, pruned_loss=0.05391, over 12052.00 frames. ], tot_loss[loss=0.1642, simple_loss=0.2542, pruned_loss=0.03713, over 2371510.88 frames. ], batch size: 42, lr: 3.30e-03, grad_scale: 16.0 2023-05-17 08:43:35,405 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([6.1040, 6.1219, 5.8210, 5.3395, 5.3304, 6.0505, 5.6208, 5.4100], device='cuda:0'), covar=tensor([0.0733, 0.0854, 0.0619, 0.1697, 0.0679, 0.0694, 0.1545, 0.1016], device='cuda:0'), in_proj_covar=tensor([0.0663, 0.0593, 0.0542, 0.0670, 0.0446, 0.0763, 0.0823, 0.0595], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0004, 0.0004, 0.0003], device='cuda:0') 2023-05-17 08:43:35,538 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.4190, 4.0983, 4.1273, 4.4846, 3.0183, 4.0557, 2.7837, 4.0893], device='cuda:0'), covar=tensor([0.1587, 0.0734, 0.0894, 0.0615, 0.1236, 0.0636, 0.1714, 0.1087], device='cuda:0'), in_proj_covar=tensor([0.0233, 0.0271, 0.0301, 0.0365, 0.0245, 0.0249, 0.0264, 0.0377], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-17 08:43:46,375 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=303614.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 08:44:01,275 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=303635.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 08:44:08,051 INFO [finetune.py:992] (0/2) Epoch 17, batch 7900, loss[loss=0.163, simple_loss=0.2586, pruned_loss=0.03366, over 11843.00 frames. ], tot_loss[loss=0.1643, simple_loss=0.2544, pruned_loss=0.03709, over 2371337.60 frames. ], batch size: 44, lr: 3.30e-03, grad_scale: 16.0 2023-05-17 08:44:28,433 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=4.27 vs. limit=5.0 2023-05-17 08:44:44,107 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.093e+02 2.729e+02 3.278e+02 3.930e+02 6.379e+02, threshold=6.555e+02, percent-clipped=1.0 2023-05-17 08:44:44,126 INFO [finetune.py:992] (0/2) Epoch 17, batch 7950, loss[loss=0.1738, simple_loss=0.2679, pruned_loss=0.03989, over 12150.00 frames. ], tot_loss[loss=0.1643, simple_loss=0.2543, pruned_loss=0.03715, over 2373477.56 frames. ], batch size: 36, lr: 3.30e-03, grad_scale: 16.0 2023-05-17 08:44:54,271 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=303708.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 08:45:20,510 INFO [finetune.py:992] (0/2) Epoch 17, batch 8000, loss[loss=0.1554, simple_loss=0.2525, pruned_loss=0.02915, over 12194.00 frames. ], tot_loss[loss=0.1636, simple_loss=0.2534, pruned_loss=0.0369, over 2378077.33 frames. ], batch size: 35, lr: 3.30e-03, grad_scale: 16.0 2023-05-17 08:45:22,236 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=4.17 vs. limit=5.0 2023-05-17 08:45:28,451 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=303756.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 08:45:30,592 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=303759.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 08:45:55,804 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.977e+02 2.710e+02 3.151e+02 3.848e+02 7.315e+02, threshold=6.302e+02, percent-clipped=1.0 2023-05-17 08:45:55,823 INFO [finetune.py:992] (0/2) Epoch 17, batch 8050, loss[loss=0.1628, simple_loss=0.2639, pruned_loss=0.03079, over 12362.00 frames. ], tot_loss[loss=0.1652, simple_loss=0.255, pruned_loss=0.03768, over 2367075.09 frames. ], batch size: 36, lr: 3.30e-03, grad_scale: 16.0 2023-05-17 08:46:00,713 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.0921, 2.4241, 3.6177, 3.0189, 3.5136, 3.1535, 2.5451, 3.5586], device='cuda:0'), covar=tensor([0.0155, 0.0399, 0.0181, 0.0265, 0.0159, 0.0217, 0.0364, 0.0140], device='cuda:0'), in_proj_covar=tensor([0.0189, 0.0214, 0.0203, 0.0196, 0.0229, 0.0175, 0.0206, 0.0199], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-17 08:46:07,849 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=303810.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 08:46:33,001 INFO [finetune.py:992] (0/2) Epoch 17, batch 8100, loss[loss=0.1546, simple_loss=0.2422, pruned_loss=0.03347, over 12274.00 frames. ], tot_loss[loss=0.1652, simple_loss=0.2551, pruned_loss=0.03766, over 2368951.85 frames. ], batch size: 33, lr: 3.30e-03, grad_scale: 16.0 2023-05-17 08:46:41,013 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.3812, 4.8564, 2.9018, 2.4483, 4.3408, 2.5594, 4.0424, 3.2953], device='cuda:0'), covar=tensor([0.0725, 0.0483, 0.1217, 0.1831, 0.0246, 0.1529, 0.0457, 0.0912], device='cuda:0'), in_proj_covar=tensor([0.0196, 0.0269, 0.0183, 0.0207, 0.0147, 0.0189, 0.0208, 0.0183], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-17 08:46:51,345 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=303871.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 08:47:05,346 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.81 vs. limit=2.0 2023-05-17 08:47:08,398 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.599e+02 2.672e+02 3.103e+02 3.782e+02 1.395e+03, threshold=6.206e+02, percent-clipped=4.0 2023-05-17 08:47:08,420 INFO [finetune.py:992] (0/2) Epoch 17, batch 8150, loss[loss=0.1623, simple_loss=0.2553, pruned_loss=0.03464, over 12130.00 frames. ], tot_loss[loss=0.1652, simple_loss=0.255, pruned_loss=0.03764, over 2370694.07 frames. ], batch size: 39, lr: 3.30e-03, grad_scale: 16.0 2023-05-17 08:47:09,313 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=303896.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 08:47:22,005 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=303914.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 08:47:33,462 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=303930.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 08:47:44,046 INFO [finetune.py:992] (0/2) Epoch 17, batch 8200, loss[loss=0.1395, simple_loss=0.2202, pruned_loss=0.02942, over 12270.00 frames. ], tot_loss[loss=0.1647, simple_loss=0.2547, pruned_loss=0.03738, over 2376482.10 frames. ], batch size: 28, lr: 3.30e-03, grad_scale: 16.0 2023-05-17 08:47:46,209 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.2704, 6.0232, 5.5729, 5.6358, 6.1862, 5.5695, 5.5882, 5.6922], device='cuda:0'), covar=tensor([0.1519, 0.1026, 0.0861, 0.2117, 0.0861, 0.2062, 0.1826, 0.1058], device='cuda:0'), in_proj_covar=tensor([0.0374, 0.0518, 0.0419, 0.0466, 0.0482, 0.0458, 0.0419, 0.0407], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-17 08:47:53,599 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=303957.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 08:47:56,980 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=303962.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 08:48:04,808 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.0916, 4.7157, 4.8176, 5.0115, 4.7949, 5.0258, 4.8652, 2.6319], device='cuda:0'), covar=tensor([0.0116, 0.0069, 0.0100, 0.0052, 0.0046, 0.0099, 0.0078, 0.0848], device='cuda:0'), in_proj_covar=tensor([0.0072, 0.0083, 0.0088, 0.0076, 0.0063, 0.0099, 0.0086, 0.0102], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-17 08:48:05,442 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.9588, 4.6246, 4.2266, 4.2974, 4.7383, 4.1957, 4.2780, 4.1516], device='cuda:0'), covar=tensor([0.1782, 0.1187, 0.1519, 0.1911, 0.1142, 0.2118, 0.1963, 0.1535], device='cuda:0'), in_proj_covar=tensor([0.0375, 0.0519, 0.0421, 0.0468, 0.0484, 0.0460, 0.0420, 0.0408], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-17 08:48:05,554 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=303974.0, num_to_drop=1, layers_to_drop={0} 2023-05-17 08:48:20,262 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.983e+02 2.564e+02 2.946e+02 3.521e+02 6.941e+02, threshold=5.893e+02, percent-clipped=1.0 2023-05-17 08:48:20,281 INFO [finetune.py:992] (0/2) Epoch 17, batch 8250, loss[loss=0.1688, simple_loss=0.2702, pruned_loss=0.03371, over 11623.00 frames. ], tot_loss[loss=0.1646, simple_loss=0.2546, pruned_loss=0.03728, over 2382017.02 frames. ], batch size: 48, lr: 3.30e-03, grad_scale: 16.0 2023-05-17 08:48:24,845 INFO [checkpoint.py:75] (0/2) Saving checkpoint to pruned_transducer_stateless7/exp_giga_finetune/checkpoint-204000.pt 2023-05-17 08:48:31,522 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.0127, 5.9017, 5.5215, 5.4732, 6.0442, 5.2287, 5.3981, 5.4908], device='cuda:0'), covar=tensor([0.1607, 0.1010, 0.1038, 0.1896, 0.0914, 0.2443, 0.2108, 0.1242], device='cuda:0'), in_proj_covar=tensor([0.0374, 0.0518, 0.0420, 0.0467, 0.0482, 0.0458, 0.0419, 0.0408], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-17 08:48:53,030 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=304035.0, num_to_drop=1, layers_to_drop={3} 2023-05-17 08:49:00,026 INFO [finetune.py:992] (0/2) Epoch 17, batch 8300, loss[loss=0.1763, simple_loss=0.2653, pruned_loss=0.04368, over 12376.00 frames. ], tot_loss[loss=0.1643, simple_loss=0.2544, pruned_loss=0.03711, over 2381346.10 frames. ], batch size: 38, lr: 3.30e-03, grad_scale: 16.0 2023-05-17 08:49:10,200 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=304059.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 08:49:36,076 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.758e+02 2.634e+02 3.099e+02 3.547e+02 8.930e+02, threshold=6.198e+02, percent-clipped=2.0 2023-05-17 08:49:36,095 INFO [finetune.py:992] (0/2) Epoch 17, batch 8350, loss[loss=0.1786, simple_loss=0.2685, pruned_loss=0.04436, over 11397.00 frames. ], tot_loss[loss=0.1647, simple_loss=0.2546, pruned_loss=0.03746, over 2370090.01 frames. ], batch size: 55, lr: 3.30e-03, grad_scale: 16.0 2023-05-17 08:49:44,648 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=304107.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 08:50:12,019 INFO [finetune.py:992] (0/2) Epoch 17, batch 8400, loss[loss=0.1496, simple_loss=0.243, pruned_loss=0.02808, over 12314.00 frames. ], tot_loss[loss=0.1644, simple_loss=0.2543, pruned_loss=0.03727, over 2374890.02 frames. ], batch size: 34, lr: 3.30e-03, grad_scale: 16.0 2023-05-17 08:50:26,984 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=304166.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 08:50:44,840 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.1920, 5.0547, 5.0636, 5.0944, 4.7827, 5.1992, 5.2270, 5.4380], device='cuda:0'), covar=tensor([0.0212, 0.0159, 0.0157, 0.0330, 0.0666, 0.0291, 0.0135, 0.0159], device='cuda:0'), in_proj_covar=tensor([0.0208, 0.0210, 0.0202, 0.0259, 0.0253, 0.0231, 0.0186, 0.0243], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0005, 0.0004, 0.0004, 0.0003, 0.0005], device='cuda:0') 2023-05-17 08:50:47,444 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.079e+02 2.713e+02 3.189e+02 4.007e+02 8.492e+02, threshold=6.378e+02, percent-clipped=2.0 2023-05-17 08:50:47,462 INFO [finetune.py:992] (0/2) Epoch 17, batch 8450, loss[loss=0.1603, simple_loss=0.2463, pruned_loss=0.03714, over 12020.00 frames. ], tot_loss[loss=0.1655, simple_loss=0.2553, pruned_loss=0.0379, over 2361617.64 frames. ], batch size: 31, lr: 3.30e-03, grad_scale: 16.0 2023-05-17 08:51:01,270 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.0201, 4.5764, 4.6420, 4.8187, 4.6528, 4.8857, 4.7026, 2.4235], device='cuda:0'), covar=tensor([0.0127, 0.0107, 0.0146, 0.0084, 0.0068, 0.0130, 0.0119, 0.1052], device='cuda:0'), in_proj_covar=tensor([0.0073, 0.0083, 0.0089, 0.0077, 0.0064, 0.0099, 0.0086, 0.0103], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-17 08:51:12,601 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=304230.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 08:51:23,784 INFO [finetune.py:992] (0/2) Epoch 17, batch 8500, loss[loss=0.157, simple_loss=0.2459, pruned_loss=0.03408, over 12283.00 frames. ], tot_loss[loss=0.1645, simple_loss=0.2542, pruned_loss=0.03745, over 2363678.72 frames. ], batch size: 33, lr: 3.29e-03, grad_scale: 16.0 2023-05-17 08:51:24,889 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 2023-05-17 08:51:28,769 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=304252.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 08:51:47,825 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=304278.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 08:51:59,812 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.875e+02 2.720e+02 3.115e+02 3.611e+02 6.407e+02, threshold=6.230e+02, percent-clipped=1.0 2023-05-17 08:51:59,831 INFO [finetune.py:992] (0/2) Epoch 17, batch 8550, loss[loss=0.1676, simple_loss=0.2639, pruned_loss=0.0357, over 12200.00 frames. ], tot_loss[loss=0.1645, simple_loss=0.254, pruned_loss=0.03748, over 2370071.40 frames. ], batch size: 35, lr: 3.29e-03, grad_scale: 16.0 2023-05-17 08:52:24,669 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=304330.0, num_to_drop=1, layers_to_drop={2} 2023-05-17 08:52:29,533 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.5936, 3.5061, 3.1955, 3.0972, 2.9223, 2.6654, 3.4939, 2.3058], device='cuda:0'), covar=tensor([0.0411, 0.0164, 0.0224, 0.0235, 0.0401, 0.0457, 0.0167, 0.0540], device='cuda:0'), in_proj_covar=tensor([0.0202, 0.0167, 0.0173, 0.0197, 0.0208, 0.0208, 0.0182, 0.0210], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-17 08:52:34,913 INFO [finetune.py:992] (0/2) Epoch 17, batch 8600, loss[loss=0.2204, simple_loss=0.2981, pruned_loss=0.07137, over 7837.00 frames. ], tot_loss[loss=0.166, simple_loss=0.2556, pruned_loss=0.03819, over 2361507.80 frames. ], batch size: 98, lr: 3.29e-03, grad_scale: 16.0 2023-05-17 08:53:10,976 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=4.88 vs. limit=5.0 2023-05-17 08:53:11,173 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.027e+02 2.657e+02 3.103e+02 3.765e+02 6.079e+02, threshold=6.206e+02, percent-clipped=0.0 2023-05-17 08:53:11,191 INFO [finetune.py:992] (0/2) Epoch 17, batch 8650, loss[loss=0.1502, simple_loss=0.236, pruned_loss=0.03223, over 12261.00 frames. ], tot_loss[loss=0.1653, simple_loss=0.2549, pruned_loss=0.03789, over 2368626.97 frames. ], batch size: 32, lr: 3.29e-03, grad_scale: 16.0 2023-05-17 08:53:40,461 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.5807, 4.2715, 4.4777, 4.0289, 4.2933, 4.0284, 4.4781, 4.2672], device='cuda:0'), covar=tensor([0.0439, 0.0595, 0.0572, 0.0389, 0.0536, 0.0495, 0.0489, 0.0865], device='cuda:0'), in_proj_covar=tensor([0.0281, 0.0287, 0.0310, 0.0281, 0.0281, 0.0279, 0.0256, 0.0231], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-17 08:53:48,064 INFO [finetune.py:992] (0/2) Epoch 17, batch 8700, loss[loss=0.1513, simple_loss=0.2442, pruned_loss=0.02922, over 12288.00 frames. ], tot_loss[loss=0.1644, simple_loss=0.2542, pruned_loss=0.03724, over 2375845.75 frames. ], batch size: 33, lr: 3.29e-03, grad_scale: 16.0 2023-05-17 08:54:03,008 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=304466.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 08:54:13,229 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.6516, 4.4706, 4.4810, 4.5310, 4.2881, 4.6478, 4.6746, 4.8281], device='cuda:0'), covar=tensor([0.0280, 0.0203, 0.0213, 0.0407, 0.0696, 0.0357, 0.0173, 0.0191], device='cuda:0'), in_proj_covar=tensor([0.0209, 0.0210, 0.0202, 0.0260, 0.0253, 0.0232, 0.0186, 0.0245], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0005, 0.0004, 0.0004, 0.0003, 0.0005], device='cuda:0') 2023-05-17 08:54:23,652 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.888e+02 2.553e+02 2.955e+02 3.456e+02 6.024e+02, threshold=5.911e+02, percent-clipped=0.0 2023-05-17 08:54:23,675 INFO [finetune.py:992] (0/2) Epoch 17, batch 8750, loss[loss=0.1753, simple_loss=0.2616, pruned_loss=0.04455, over 12177.00 frames. ], tot_loss[loss=0.1635, simple_loss=0.2534, pruned_loss=0.0368, over 2379207.24 frames. ], batch size: 35, lr: 3.29e-03, grad_scale: 16.0 2023-05-17 08:54:37,292 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=304514.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 08:54:40,255 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=304518.0, num_to_drop=1, layers_to_drop={1} 2023-05-17 08:54:59,899 INFO [finetune.py:992] (0/2) Epoch 17, batch 8800, loss[loss=0.1926, simple_loss=0.2839, pruned_loss=0.05061, over 12312.00 frames. ], tot_loss[loss=0.1634, simple_loss=0.2532, pruned_loss=0.03679, over 2377795.38 frames. ], batch size: 34, lr: 3.29e-03, grad_scale: 16.0 2023-05-17 08:55:05,058 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=304552.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 08:55:25,113 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=304579.0, num_to_drop=1, layers_to_drop={2} 2023-05-17 08:55:36,329 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.848e+02 2.569e+02 2.981e+02 3.638e+02 6.773e+02, threshold=5.961e+02, percent-clipped=1.0 2023-05-17 08:55:36,358 INFO [finetune.py:992] (0/2) Epoch 17, batch 8850, loss[loss=0.1777, simple_loss=0.2636, pruned_loss=0.04584, over 12296.00 frames. ], tot_loss[loss=0.1636, simple_loss=0.2534, pruned_loss=0.03691, over 2365124.14 frames. ], batch size: 34, lr: 3.29e-03, grad_scale: 16.0 2023-05-17 08:55:40,250 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=304600.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 08:55:51,621 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=304616.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 08:55:55,826 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=304622.0, num_to_drop=1, layers_to_drop={0} 2023-05-17 08:56:01,734 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=304630.0, num_to_drop=1, layers_to_drop={2} 2023-05-17 08:56:09,090 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.90 vs. limit=2.0 2023-05-17 08:56:12,331 INFO [finetune.py:992] (0/2) Epoch 17, batch 8900, loss[loss=0.1639, simple_loss=0.2517, pruned_loss=0.03807, over 12021.00 frames. ], tot_loss[loss=0.1644, simple_loss=0.2543, pruned_loss=0.03721, over 2362344.53 frames. ], batch size: 40, lr: 3.29e-03, grad_scale: 16.0 2023-05-17 08:56:12,495 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=304645.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 08:56:18,598 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.8849, 3.8785, 3.9191, 4.0272, 3.7899, 3.8396, 3.6390, 3.9426], device='cuda:0'), covar=tensor([0.1520, 0.0723, 0.1472, 0.0795, 0.1708, 0.1233, 0.0657, 0.0915], device='cuda:0'), in_proj_covar=tensor([0.0563, 0.0729, 0.0641, 0.0657, 0.0872, 0.0777, 0.0584, 0.0499], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0005, 0.0005, 0.0003, 0.0003], device='cuda:0') 2023-05-17 08:56:29,405 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=304668.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 08:56:34,474 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=304675.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 08:56:34,513 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.5643, 3.3971, 3.1644, 3.0459, 2.7726, 2.6707, 3.3720, 2.1956], device='cuda:0'), covar=tensor([0.0392, 0.0161, 0.0194, 0.0214, 0.0438, 0.0361, 0.0151, 0.0585], device='cuda:0'), in_proj_covar=tensor([0.0202, 0.0166, 0.0172, 0.0198, 0.0207, 0.0206, 0.0181, 0.0210], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-17 08:56:35,943 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=304677.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 08:56:36,528 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=304678.0, num_to_drop=1, layers_to_drop={0} 2023-05-17 08:56:39,976 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=304683.0, num_to_drop=1, layers_to_drop={2} 2023-05-17 08:56:48,115 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.053e+02 2.720e+02 3.219e+02 3.830e+02 1.038e+03, threshold=6.438e+02, percent-clipped=4.0 2023-05-17 08:56:48,134 INFO [finetune.py:992] (0/2) Epoch 17, batch 8950, loss[loss=0.1633, simple_loss=0.2514, pruned_loss=0.03761, over 12049.00 frames. ], tot_loss[loss=0.1676, simple_loss=0.2571, pruned_loss=0.03904, over 2336748.89 frames. ], batch size: 40, lr: 3.29e-03, grad_scale: 16.0 2023-05-17 08:56:56,079 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=304706.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 08:57:12,818 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=304729.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 08:57:17,829 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=304736.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 08:57:23,612 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.80 vs. limit=2.0 2023-05-17 08:57:23,959 INFO [finetune.py:992] (0/2) Epoch 17, batch 9000, loss[loss=0.1724, simple_loss=0.2752, pruned_loss=0.03475, over 12264.00 frames. ], tot_loss[loss=0.1671, simple_loss=0.2566, pruned_loss=0.03874, over 2349034.16 frames. ], batch size: 32, lr: 3.29e-03, grad_scale: 16.0 2023-05-17 08:57:23,960 INFO [finetune.py:1017] (0/2) Computing validation loss 2023-05-17 08:57:30,541 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.8233, 5.7965, 5.7142, 5.2883, 5.1287, 5.7956, 5.3971, 5.5105], device='cuda:0'), covar=tensor([0.0793, 0.1107, 0.0619, 0.1763, 0.0564, 0.0769, 0.1641, 0.0877], device='cuda:0'), in_proj_covar=tensor([0.0653, 0.0585, 0.0538, 0.0663, 0.0441, 0.0752, 0.0814, 0.0590], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-05-17 08:57:33,384 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.3057, 4.1095, 4.2093, 4.1669, 3.8084, 4.3308, 4.3098, 4.3824], device='cuda:0'), covar=tensor([0.0216, 0.0240, 0.0226, 0.0489, 0.0680, 0.0426, 0.0188, 0.0301], device='cuda:0'), in_proj_covar=tensor([0.0207, 0.0208, 0.0200, 0.0257, 0.0251, 0.0231, 0.0185, 0.0242], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0005, 0.0004, 0.0004, 0.0003, 0.0005], device='cuda:0') 2023-05-17 08:57:41,376 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([1.8404, 2.1721, 2.1921, 2.0164, 2.0209, 2.0408, 1.8157, 1.6099], device='cuda:0'), covar=tensor([0.0382, 0.0205, 0.0189, 0.0231, 0.0351, 0.0281, 0.0300, 0.0490], device='cuda:0'), in_proj_covar=tensor([0.0203, 0.0168, 0.0173, 0.0199, 0.0209, 0.0208, 0.0182, 0.0211], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-17 08:57:43,041 INFO [finetune.py:1026] (0/2) Epoch 17, validation: loss=0.3201, simple_loss=0.3917, pruned_loss=0.1243, over 1020973.00 frames. 2023-05-17 08:57:43,042 INFO [finetune.py:1027] (0/2) Maximum memory allocated so far is 12827MB 2023-05-17 08:58:13,563 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.13 vs. limit=2.0 2023-05-17 08:58:19,157 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=304794.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 08:58:19,665 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.907e+02 2.754e+02 3.090e+02 3.592e+02 8.668e+02, threshold=6.181e+02, percent-clipped=1.0 2023-05-17 08:58:19,683 INFO [finetune.py:992] (0/2) Epoch 17, batch 9050, loss[loss=0.1842, simple_loss=0.2733, pruned_loss=0.04759, over 12108.00 frames. ], tot_loss[loss=0.1658, simple_loss=0.2553, pruned_loss=0.0382, over 2349810.02 frames. ], batch size: 42, lr: 3.29e-03, grad_scale: 16.0 2023-05-17 08:58:24,664 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.38 vs. limit=2.0 2023-05-17 08:58:56,323 INFO [finetune.py:992] (0/2) Epoch 17, batch 9100, loss[loss=0.1399, simple_loss=0.2262, pruned_loss=0.02687, over 12349.00 frames. ], tot_loss[loss=0.1649, simple_loss=0.2545, pruned_loss=0.0376, over 2358385.70 frames. ], batch size: 30, lr: 3.29e-03, grad_scale: 16.0 2023-05-17 08:59:03,653 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=304855.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 08:59:17,228 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=304874.0, num_to_drop=1, layers_to_drop={0} 2023-05-17 08:59:20,468 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.65 vs. limit=2.0 2023-05-17 08:59:32,220 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.591e+02 2.375e+02 2.795e+02 3.469e+02 7.319e+02, threshold=5.590e+02, percent-clipped=2.0 2023-05-17 08:59:32,239 INFO [finetune.py:992] (0/2) Epoch 17, batch 9150, loss[loss=0.1478, simple_loss=0.2441, pruned_loss=0.02577, over 12359.00 frames. ], tot_loss[loss=0.165, simple_loss=0.2548, pruned_loss=0.03756, over 2363309.50 frames. ], batch size: 35, lr: 3.29e-03, grad_scale: 16.0 2023-05-17 08:59:33,329 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=2.62 vs. limit=5.0 2023-05-17 09:00:08,397 INFO [finetune.py:992] (0/2) Epoch 17, batch 9200, loss[loss=0.1314, simple_loss=0.2246, pruned_loss=0.01913, over 11998.00 frames. ], tot_loss[loss=0.1639, simple_loss=0.2537, pruned_loss=0.03705, over 2365788.89 frames. ], batch size: 28, lr: 3.29e-03, grad_scale: 16.0 2023-05-17 09:00:11,744 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.96 vs. limit=2.0 2023-05-17 09:00:28,213 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=304972.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 09:00:32,316 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=304978.0, num_to_drop=1, layers_to_drop={1} 2023-05-17 09:00:39,677 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.90 vs. limit=2.0 2023-05-17 09:00:44,330 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.959e+02 2.586e+02 3.113e+02 3.849e+02 6.015e+02, threshold=6.226e+02, percent-clipped=1.0 2023-05-17 09:00:44,350 INFO [finetune.py:992] (0/2) Epoch 17, batch 9250, loss[loss=0.1772, simple_loss=0.2758, pruned_loss=0.03936, over 11875.00 frames. ], tot_loss[loss=0.1638, simple_loss=0.2537, pruned_loss=0.03694, over 2360251.27 frames. ], batch size: 44, lr: 3.29e-03, grad_scale: 16.0 2023-05-17 09:00:48,920 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=305001.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 09:00:52,654 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.4416, 2.6368, 3.1936, 4.2810, 2.3562, 4.3180, 4.3848, 4.5334], device='cuda:0'), covar=tensor([0.0126, 0.1345, 0.0499, 0.0160, 0.1416, 0.0277, 0.0155, 0.0096], device='cuda:0'), in_proj_covar=tensor([0.0126, 0.0209, 0.0189, 0.0126, 0.0195, 0.0186, 0.0183, 0.0130], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002], device='cuda:0') 2023-05-17 09:01:05,367 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=305024.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 09:01:10,347 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=305031.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 09:01:20,154 INFO [finetune.py:992] (0/2) Epoch 17, batch 9300, loss[loss=0.1571, simple_loss=0.2439, pruned_loss=0.03509, over 12126.00 frames. ], tot_loss[loss=0.1636, simple_loss=0.2535, pruned_loss=0.03678, over 2359858.81 frames. ], batch size: 30, lr: 3.29e-03, grad_scale: 16.0 2023-05-17 09:01:42,031 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.3175, 4.6345, 2.8846, 2.6426, 3.9504, 2.6442, 3.9539, 3.2100], device='cuda:0'), covar=tensor([0.0718, 0.0425, 0.1133, 0.1512, 0.0312, 0.1325, 0.0511, 0.0815], device='cuda:0'), in_proj_covar=tensor([0.0192, 0.0261, 0.0178, 0.0202, 0.0144, 0.0184, 0.0204, 0.0178], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-17 09:01:55,874 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.622e+02 2.651e+02 3.225e+02 3.688e+02 6.047e+02, threshold=6.451e+02, percent-clipped=0.0 2023-05-17 09:01:55,893 INFO [finetune.py:992] (0/2) Epoch 17, batch 9350, loss[loss=0.1471, simple_loss=0.2366, pruned_loss=0.02879, over 12287.00 frames. ], tot_loss[loss=0.164, simple_loss=0.2543, pruned_loss=0.03683, over 2369077.61 frames. ], batch size: 33, lr: 3.29e-03, grad_scale: 16.0 2023-05-17 09:02:06,362 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.5484, 4.1502, 4.4076, 4.3812, 4.3286, 4.4020, 4.2795, 2.3319], device='cuda:0'), covar=tensor([0.0215, 0.0134, 0.0150, 0.0098, 0.0075, 0.0200, 0.0186, 0.1132], device='cuda:0'), in_proj_covar=tensor([0.0072, 0.0082, 0.0087, 0.0076, 0.0063, 0.0098, 0.0086, 0.0102], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-17 09:02:08,523 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.4045, 5.1379, 5.3220, 5.4105, 4.9435, 5.0643, 4.7560, 5.3631], device='cuda:0'), covar=tensor([0.0690, 0.0712, 0.0960, 0.0524, 0.2214, 0.1275, 0.0627, 0.0993], device='cuda:0'), in_proj_covar=tensor([0.0563, 0.0727, 0.0644, 0.0657, 0.0878, 0.0774, 0.0584, 0.0500], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0005, 0.0005, 0.0003, 0.0003], device='cuda:0') 2023-05-17 09:02:26,732 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.89 vs. limit=2.0 2023-05-17 09:02:32,074 INFO [finetune.py:992] (0/2) Epoch 17, batch 9400, loss[loss=0.1342, simple_loss=0.224, pruned_loss=0.02226, over 12142.00 frames. ], tot_loss[loss=0.1651, simple_loss=0.255, pruned_loss=0.03758, over 2366203.62 frames. ], batch size: 30, lr: 3.29e-03, grad_scale: 16.0 2023-05-17 09:02:35,635 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=305150.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 09:02:52,822 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=305174.0, num_to_drop=1, layers_to_drop={2} 2023-05-17 09:03:04,560 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.3750, 4.6992, 2.9035, 2.6035, 4.0404, 2.7244, 3.9583, 3.2035], device='cuda:0'), covar=tensor([0.0730, 0.0503, 0.1208, 0.1684, 0.0361, 0.1275, 0.0534, 0.0884], device='cuda:0'), in_proj_covar=tensor([0.0192, 0.0262, 0.0178, 0.0203, 0.0144, 0.0184, 0.0204, 0.0178], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-17 09:03:07,804 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.776e+02 2.616e+02 3.003e+02 3.663e+02 5.501e+02, threshold=6.006e+02, percent-clipped=0.0 2023-05-17 09:03:07,822 INFO [finetune.py:992] (0/2) Epoch 17, batch 9450, loss[loss=0.1856, simple_loss=0.2731, pruned_loss=0.04904, over 12038.00 frames. ], tot_loss[loss=0.1641, simple_loss=0.2542, pruned_loss=0.03702, over 2374043.35 frames. ], batch size: 40, lr: 3.29e-03, grad_scale: 16.0 2023-05-17 09:03:28,265 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=305222.0, num_to_drop=1, layers_to_drop={0} 2023-05-17 09:03:33,859 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.4233, 2.7004, 3.6475, 4.3696, 3.6744, 4.4329, 3.7085, 3.1989], device='cuda:0'), covar=tensor([0.0042, 0.0394, 0.0156, 0.0055, 0.0168, 0.0073, 0.0152, 0.0395], device='cuda:0'), in_proj_covar=tensor([0.0089, 0.0122, 0.0104, 0.0081, 0.0105, 0.0116, 0.0102, 0.0138], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-17 09:03:43,082 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=305243.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 09:03:44,266 INFO [finetune.py:992] (0/2) Epoch 17, batch 9500, loss[loss=0.1611, simple_loss=0.2476, pruned_loss=0.03728, over 12353.00 frames. ], tot_loss[loss=0.1642, simple_loss=0.2544, pruned_loss=0.03705, over 2369490.72 frames. ], batch size: 36, lr: 3.29e-03, grad_scale: 16.0 2023-05-17 09:03:51,752 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2023-05-17 09:04:04,238 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=305272.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 09:04:08,505 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=305278.0, num_to_drop=1, layers_to_drop={1} 2023-05-17 09:04:20,292 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.849e+02 2.458e+02 2.931e+02 3.477e+02 6.918e+02, threshold=5.861e+02, percent-clipped=2.0 2023-05-17 09:04:20,311 INFO [finetune.py:992] (0/2) Epoch 17, batch 9550, loss[loss=0.1662, simple_loss=0.2534, pruned_loss=0.03947, over 11213.00 frames. ], tot_loss[loss=0.1631, simple_loss=0.253, pruned_loss=0.0366, over 2378307.98 frames. ], batch size: 55, lr: 3.29e-03, grad_scale: 32.0 2023-05-17 09:04:24,913 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=305301.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 09:04:27,141 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=305304.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 09:04:38,127 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=305320.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 09:04:39,678 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=305322.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 09:04:41,061 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=305324.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 09:04:42,415 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=305326.0, num_to_drop=1, layers_to_drop={0} 2023-05-17 09:04:45,874 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=305331.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 09:04:48,779 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.3176, 4.7100, 2.9295, 2.8360, 4.0918, 2.7243, 3.9363, 3.2647], device='cuda:0'), covar=tensor([0.0704, 0.0444, 0.1108, 0.1359, 0.0320, 0.1195, 0.0513, 0.0765], device='cuda:0'), in_proj_covar=tensor([0.0192, 0.0262, 0.0178, 0.0203, 0.0144, 0.0184, 0.0204, 0.0178], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-17 09:04:56,392 INFO [finetune.py:992] (0/2) Epoch 17, batch 9600, loss[loss=0.1896, simple_loss=0.2729, pruned_loss=0.05318, over 11769.00 frames. ], tot_loss[loss=0.1642, simple_loss=0.2538, pruned_loss=0.03725, over 2369784.12 frames. ], batch size: 44, lr: 3.29e-03, grad_scale: 32.0 2023-05-17 09:04:59,208 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=305349.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 09:05:09,453 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.7844, 3.6179, 3.1820, 3.1997, 2.7712, 2.7539, 3.6648, 2.4231], device='cuda:0'), covar=tensor([0.0402, 0.0174, 0.0268, 0.0259, 0.0582, 0.0453, 0.0182, 0.0540], device='cuda:0'), in_proj_covar=tensor([0.0204, 0.0169, 0.0175, 0.0200, 0.0211, 0.0208, 0.0183, 0.0212], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-17 09:05:15,794 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=305372.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 09:05:20,906 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=305379.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 09:05:23,946 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=305383.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 09:05:26,128 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.4117, 4.6509, 3.9849, 4.9350, 4.4685, 2.9566, 4.1630, 3.1789], device='cuda:0'), covar=tensor([0.0803, 0.0794, 0.1692, 0.0550, 0.1397, 0.1745, 0.1167, 0.3209], device='cuda:0'), in_proj_covar=tensor([0.0320, 0.0390, 0.0373, 0.0343, 0.0386, 0.0283, 0.0360, 0.0376], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-17 09:05:28,823 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.4462, 4.8364, 2.9420, 2.7777, 4.2009, 2.8188, 4.0468, 3.3102], device='cuda:0'), covar=tensor([0.0713, 0.0430, 0.1288, 0.1610, 0.0277, 0.1273, 0.0522, 0.0847], device='cuda:0'), in_proj_covar=tensor([0.0191, 0.0261, 0.0178, 0.0203, 0.0144, 0.0184, 0.0203, 0.0178], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-17 09:05:32,259 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.723e+02 2.584e+02 3.111e+02 3.897e+02 1.039e+03, threshold=6.221e+02, percent-clipped=5.0 2023-05-17 09:05:32,277 INFO [finetune.py:992] (0/2) Epoch 17, batch 9650, loss[loss=0.143, simple_loss=0.2304, pruned_loss=0.02778, over 12127.00 frames. ], tot_loss[loss=0.1631, simple_loss=0.2527, pruned_loss=0.03679, over 2371620.13 frames. ], batch size: 30, lr: 3.29e-03, grad_scale: 32.0 2023-05-17 09:06:09,066 INFO [finetune.py:992] (0/2) Epoch 17, batch 9700, loss[loss=0.1592, simple_loss=0.2595, pruned_loss=0.02943, over 12162.00 frames. ], tot_loss[loss=0.1632, simple_loss=0.2529, pruned_loss=0.03678, over 2381465.56 frames. ], batch size: 36, lr: 3.29e-03, grad_scale: 32.0 2023-05-17 09:06:12,050 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.8918, 4.7228, 4.8512, 4.9017, 4.5513, 4.6292, 4.3778, 4.8260], device='cuda:0'), covar=tensor([0.0761, 0.0662, 0.0882, 0.0724, 0.1921, 0.1226, 0.0595, 0.1099], device='cuda:0'), in_proj_covar=tensor([0.0557, 0.0721, 0.0639, 0.0653, 0.0869, 0.0766, 0.0581, 0.0499], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0005, 0.0004, 0.0003, 0.0003], device='cuda:0') 2023-05-17 09:06:12,786 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=305450.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 09:06:30,998 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.6396, 3.5271, 3.2452, 3.1892, 2.9233, 2.7592, 3.5548, 2.2590], device='cuda:0'), covar=tensor([0.0410, 0.0146, 0.0209, 0.0219, 0.0383, 0.0400, 0.0150, 0.0563], device='cuda:0'), in_proj_covar=tensor([0.0202, 0.0167, 0.0173, 0.0198, 0.0208, 0.0206, 0.0181, 0.0210], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-17 09:06:44,923 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.688e+02 2.649e+02 3.133e+02 3.869e+02 5.650e+02, threshold=6.266e+02, percent-clipped=0.0 2023-05-17 09:06:44,942 INFO [finetune.py:992] (0/2) Epoch 17, batch 9750, loss[loss=0.1608, simple_loss=0.2543, pruned_loss=0.03368, over 12141.00 frames. ], tot_loss[loss=0.1638, simple_loss=0.2539, pruned_loss=0.03686, over 2372832.27 frames. ], batch size: 38, lr: 3.29e-03, grad_scale: 32.0 2023-05-17 09:06:47,095 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=305498.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 09:07:20,609 INFO [finetune.py:992] (0/2) Epoch 17, batch 9800, loss[loss=0.1449, simple_loss=0.2298, pruned_loss=0.02996, over 12204.00 frames. ], tot_loss[loss=0.1623, simple_loss=0.2524, pruned_loss=0.03616, over 2379764.17 frames. ], batch size: 29, lr: 3.29e-03, grad_scale: 32.0 2023-05-17 09:07:35,363 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.79 vs. limit=2.0 2023-05-17 09:07:57,098 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.890e+02 2.590e+02 2.906e+02 3.453e+02 6.733e+02, threshold=5.813e+02, percent-clipped=1.0 2023-05-17 09:07:57,118 INFO [finetune.py:992] (0/2) Epoch 17, batch 9850, loss[loss=0.171, simple_loss=0.2802, pruned_loss=0.03088, over 12347.00 frames. ], tot_loss[loss=0.1619, simple_loss=0.252, pruned_loss=0.03592, over 2378097.26 frames. ], batch size: 35, lr: 3.29e-03, grad_scale: 32.0 2023-05-17 09:07:59,933 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=305599.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 09:08:17,270 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=305623.0, num_to_drop=1, layers_to_drop={0} 2023-05-17 09:08:18,759 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.8767, 3.0149, 4.6976, 4.8111, 2.9924, 2.7200, 3.1025, 2.3301], device='cuda:0'), covar=tensor([0.1617, 0.2957, 0.0423, 0.0455, 0.1327, 0.2577, 0.2604, 0.3813], device='cuda:0'), in_proj_covar=tensor([0.0310, 0.0397, 0.0283, 0.0308, 0.0284, 0.0325, 0.0407, 0.0385], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-17 09:08:33,426 INFO [finetune.py:992] (0/2) Epoch 17, batch 9900, loss[loss=0.1694, simple_loss=0.2688, pruned_loss=0.03496, over 12174.00 frames. ], tot_loss[loss=0.1618, simple_loss=0.2517, pruned_loss=0.03594, over 2385262.16 frames. ], batch size: 35, lr: 3.29e-03, grad_scale: 32.0 2023-05-17 09:08:38,462 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=305652.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 09:08:39,742 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.6003, 4.4810, 4.5721, 4.6487, 4.2997, 4.3782, 4.1682, 4.5342], device='cuda:0'), covar=tensor([0.0813, 0.0664, 0.1005, 0.0595, 0.1886, 0.1300, 0.0619, 0.1091], device='cuda:0'), in_proj_covar=tensor([0.0561, 0.0725, 0.0644, 0.0656, 0.0873, 0.0772, 0.0584, 0.0502], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0005, 0.0005, 0.0003, 0.0003], device='cuda:0') 2023-05-17 09:08:57,054 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=305678.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 09:08:59,794 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.62 vs. limit=2.0 2023-05-17 09:09:01,714 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=305684.0, num_to_drop=1, layers_to_drop={3} 2023-05-17 09:09:09,279 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.517e+02 2.607e+02 3.012e+02 3.644e+02 9.148e+02, threshold=6.023e+02, percent-clipped=2.0 2023-05-17 09:09:09,297 INFO [finetune.py:992] (0/2) Epoch 17, batch 9950, loss[loss=0.1841, simple_loss=0.2722, pruned_loss=0.04802, over 12129.00 frames. ], tot_loss[loss=0.1618, simple_loss=0.2515, pruned_loss=0.03605, over 2385644.74 frames. ], batch size: 38, lr: 3.29e-03, grad_scale: 32.0 2023-05-17 09:09:09,720 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-05-17 09:09:22,924 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=305713.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 09:09:30,606 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.79 vs. limit=2.0 2023-05-17 09:09:45,750 INFO [finetune.py:992] (0/2) Epoch 17, batch 10000, loss[loss=0.125, simple_loss=0.2098, pruned_loss=0.02007, over 12132.00 frames. ], tot_loss[loss=0.1615, simple_loss=0.2513, pruned_loss=0.03587, over 2387491.08 frames. ], batch size: 30, lr: 3.29e-03, grad_scale: 32.0 2023-05-17 09:10:09,366 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.0963, 5.9461, 5.5481, 5.4546, 6.0719, 5.3944, 5.4181, 5.5291], device='cuda:0'), covar=tensor([0.1604, 0.1078, 0.1075, 0.2133, 0.1027, 0.2142, 0.2024, 0.1214], device='cuda:0'), in_proj_covar=tensor([0.0369, 0.0514, 0.0418, 0.0464, 0.0480, 0.0452, 0.0418, 0.0401], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-17 09:10:22,156 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.074e+02 2.635e+02 3.117e+02 3.744e+02 7.189e+02, threshold=6.234e+02, percent-clipped=2.0 2023-05-17 09:10:22,174 INFO [finetune.py:992] (0/2) Epoch 17, batch 10050, loss[loss=0.1423, simple_loss=0.229, pruned_loss=0.02784, over 12190.00 frames. ], tot_loss[loss=0.1621, simple_loss=0.2518, pruned_loss=0.0362, over 2388061.45 frames. ], batch size: 29, lr: 3.28e-03, grad_scale: 32.0 2023-05-17 09:10:35,894 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.39 vs. limit=2.0 2023-05-17 09:10:36,366 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.0414, 4.9668, 4.8811, 4.9674, 3.9915, 5.1150, 5.1111, 5.1925], device='cuda:0'), covar=tensor([0.0331, 0.0190, 0.0197, 0.0359, 0.1287, 0.0363, 0.0181, 0.0247], device='cuda:0'), in_proj_covar=tensor([0.0208, 0.0211, 0.0201, 0.0260, 0.0255, 0.0233, 0.0186, 0.0245], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0005, 0.0004, 0.0005, 0.0003, 0.0005], device='cuda:0') 2023-05-17 09:10:59,049 INFO [finetune.py:992] (0/2) Epoch 17, batch 10100, loss[loss=0.1711, simple_loss=0.2739, pruned_loss=0.03422, over 12142.00 frames. ], tot_loss[loss=0.1629, simple_loss=0.2528, pruned_loss=0.03657, over 2385166.39 frames. ], batch size: 34, lr: 3.28e-03, grad_scale: 16.0 2023-05-17 09:11:04,235 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=305852.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 09:11:08,089 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 2023-05-17 09:11:23,203 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.0434, 2.3762, 3.5917, 3.0483, 3.5013, 3.2161, 2.5451, 3.5045], device='cuda:0'), covar=tensor([0.0162, 0.0452, 0.0205, 0.0273, 0.0157, 0.0207, 0.0433, 0.0153], device='cuda:0'), in_proj_covar=tensor([0.0189, 0.0216, 0.0205, 0.0197, 0.0231, 0.0177, 0.0207, 0.0201], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-17 09:11:23,453 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=4.58 vs. limit=5.0 2023-05-17 09:11:34,766 INFO [finetune.py:992] (0/2) Epoch 17, batch 10150, loss[loss=0.1359, simple_loss=0.2224, pruned_loss=0.02471, over 12023.00 frames. ], tot_loss[loss=0.1625, simple_loss=0.2522, pruned_loss=0.03637, over 2387431.42 frames. ], batch size: 28, lr: 3.28e-03, grad_scale: 16.0 2023-05-17 09:11:35,446 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.972e+02 2.729e+02 3.119e+02 3.676e+02 7.359e+02, threshold=6.237e+02, percent-clipped=2.0 2023-05-17 09:11:37,675 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=305899.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 09:11:42,014 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.6163, 3.6210, 3.2585, 3.2203, 2.9012, 2.7723, 3.6445, 2.5159], device='cuda:0'), covar=tensor([0.0435, 0.0189, 0.0264, 0.0266, 0.0520, 0.0458, 0.0148, 0.0535], device='cuda:0'), in_proj_covar=tensor([0.0203, 0.0167, 0.0174, 0.0198, 0.0209, 0.0206, 0.0181, 0.0210], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-17 09:11:47,796 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=305913.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 09:12:10,962 INFO [finetune.py:992] (0/2) Epoch 17, batch 10200, loss[loss=0.1676, simple_loss=0.2619, pruned_loss=0.03668, over 12024.00 frames. ], tot_loss[loss=0.1628, simple_loss=0.2525, pruned_loss=0.03653, over 2382518.47 frames. ], batch size: 40, lr: 3.28e-03, grad_scale: 16.0 2023-05-17 09:12:12,365 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=305947.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 09:12:34,121 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-05-17 09:12:34,664 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.8848, 4.6425, 4.6015, 4.7800, 4.5653, 4.9029, 4.7325, 2.5412], device='cuda:0'), covar=tensor([0.0128, 0.0069, 0.0122, 0.0064, 0.0058, 0.0101, 0.0088, 0.0832], device='cuda:0'), in_proj_covar=tensor([0.0073, 0.0082, 0.0087, 0.0076, 0.0063, 0.0097, 0.0086, 0.0101], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-17 09:12:34,668 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=305978.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 09:12:35,331 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=305979.0, num_to_drop=1, layers_to_drop={2} 2023-05-17 09:12:47,288 INFO [finetune.py:992] (0/2) Epoch 17, batch 10250, loss[loss=0.137, simple_loss=0.2217, pruned_loss=0.02617, over 12016.00 frames. ], tot_loss[loss=0.1629, simple_loss=0.2528, pruned_loss=0.03654, over 2382574.22 frames. ], batch size: 28, lr: 3.28e-03, grad_scale: 16.0 2023-05-17 09:12:47,940 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.000e+02 2.699e+02 3.265e+02 4.214e+02 9.898e+02, threshold=6.531e+02, percent-clipped=5.0 2023-05-17 09:12:51,114 INFO [checkpoint.py:75] (0/2) Saving checkpoint to pruned_transducer_stateless7/exp_giga_finetune/checkpoint-206000.pt 2023-05-17 09:13:00,002 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=306008.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 09:13:12,837 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=306026.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 09:13:26,457 INFO [finetune.py:992] (0/2) Epoch 17, batch 10300, loss[loss=0.15, simple_loss=0.2314, pruned_loss=0.03429, over 12138.00 frames. ], tot_loss[loss=0.1634, simple_loss=0.253, pruned_loss=0.03689, over 2365090.80 frames. ], batch size: 30, lr: 3.28e-03, grad_scale: 16.0 2023-05-17 09:13:36,761 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.6730, 3.3593, 5.1779, 2.6351, 2.8657, 3.6810, 3.2389, 3.7410], device='cuda:0'), covar=tensor([0.0433, 0.1181, 0.0286, 0.1241, 0.1984, 0.1599, 0.1387, 0.1282], device='cuda:0'), in_proj_covar=tensor([0.0242, 0.0239, 0.0262, 0.0186, 0.0241, 0.0300, 0.0229, 0.0273], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-17 09:13:41,002 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.6855, 3.3565, 5.2246, 2.5931, 2.8978, 3.7323, 3.2733, 3.7827], device='cuda:0'), covar=tensor([0.0395, 0.1201, 0.0230, 0.1234, 0.1901, 0.1627, 0.1319, 0.1242], device='cuda:0'), in_proj_covar=tensor([0.0242, 0.0239, 0.0262, 0.0186, 0.0241, 0.0300, 0.0229, 0.0273], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-17 09:13:43,725 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.1395, 5.9872, 5.5824, 5.5338, 6.0875, 5.3341, 5.4323, 5.5512], device='cuda:0'), covar=tensor([0.1543, 0.0925, 0.1059, 0.1940, 0.0992, 0.2178, 0.1916, 0.1116], device='cuda:0'), in_proj_covar=tensor([0.0368, 0.0513, 0.0415, 0.0461, 0.0478, 0.0451, 0.0417, 0.0398], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-17 09:14:02,906 INFO [finetune.py:992] (0/2) Epoch 17, batch 10350, loss[loss=0.1552, simple_loss=0.2428, pruned_loss=0.03375, over 12170.00 frames. ], tot_loss[loss=0.1635, simple_loss=0.253, pruned_loss=0.03705, over 2359987.96 frames. ], batch size: 31, lr: 3.28e-03, grad_scale: 16.0 2023-05-17 09:14:03,613 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.765e+02 2.658e+02 3.080e+02 3.475e+02 5.774e+02, threshold=6.161e+02, percent-clipped=0.0 2023-05-17 09:14:38,947 INFO [finetune.py:992] (0/2) Epoch 17, batch 10400, loss[loss=0.148, simple_loss=0.2274, pruned_loss=0.03429, over 12005.00 frames. ], tot_loss[loss=0.1625, simple_loss=0.2516, pruned_loss=0.03669, over 2367256.39 frames. ], batch size: 28, lr: 3.28e-03, grad_scale: 16.0 2023-05-17 09:14:58,371 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=2.43 vs. limit=5.0 2023-05-17 09:15:13,707 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=306194.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 09:15:14,235 INFO [finetune.py:992] (0/2) Epoch 17, batch 10450, loss[loss=0.1556, simple_loss=0.2471, pruned_loss=0.03206, over 12277.00 frames. ], tot_loss[loss=0.1628, simple_loss=0.2519, pruned_loss=0.03683, over 2366667.86 frames. ], batch size: 28, lr: 3.28e-03, grad_scale: 16.0 2023-05-17 09:15:14,936 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.945e+02 2.638e+02 3.099e+02 3.984e+02 6.407e+02, threshold=6.199e+02, percent-clipped=2.0 2023-05-17 09:15:15,194 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.0661, 2.4504, 3.6573, 3.0876, 3.4764, 3.2315, 2.6679, 3.5370], device='cuda:0'), covar=tensor([0.0151, 0.0415, 0.0166, 0.0294, 0.0183, 0.0213, 0.0384, 0.0166], device='cuda:0'), in_proj_covar=tensor([0.0190, 0.0218, 0.0205, 0.0198, 0.0232, 0.0177, 0.0208, 0.0202], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-17 09:15:24,512 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=306208.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 09:15:28,559 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.45 vs. limit=2.0 2023-05-17 09:15:50,238 INFO [finetune.py:992] (0/2) Epoch 17, batch 10500, loss[loss=0.1666, simple_loss=0.2542, pruned_loss=0.0395, over 12376.00 frames. ], tot_loss[loss=0.1636, simple_loss=0.2529, pruned_loss=0.03717, over 2366561.42 frames. ], batch size: 38, lr: 3.28e-03, grad_scale: 16.0 2023-05-17 09:15:57,792 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=306255.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 09:16:15,173 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=306279.0, num_to_drop=1, layers_to_drop={1} 2023-05-17 09:16:26,426 INFO [finetune.py:992] (0/2) Epoch 17, batch 10550, loss[loss=0.1506, simple_loss=0.2479, pruned_loss=0.02666, over 12300.00 frames. ], tot_loss[loss=0.1634, simple_loss=0.2534, pruned_loss=0.03673, over 2371906.41 frames. ], batch size: 34, lr: 3.28e-03, grad_scale: 16.0 2023-05-17 09:16:27,137 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.576e+02 2.688e+02 3.207e+02 3.614e+02 7.710e+02, threshold=6.413e+02, percent-clipped=2.0 2023-05-17 09:16:35,825 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=306308.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 09:16:44,481 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.4950, 2.7124, 3.5767, 4.4441, 3.8237, 4.4580, 3.7969, 3.2359], device='cuda:0'), covar=tensor([0.0047, 0.0380, 0.0154, 0.0048, 0.0131, 0.0090, 0.0136, 0.0356], device='cuda:0'), in_proj_covar=tensor([0.0091, 0.0125, 0.0107, 0.0082, 0.0107, 0.0119, 0.0105, 0.0141], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-17 09:16:49,254 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=306327.0, num_to_drop=1, layers_to_drop={0} 2023-05-17 09:16:50,998 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.42 vs. limit=2.0 2023-05-17 09:16:57,647 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 2023-05-17 09:17:02,125 INFO [finetune.py:992] (0/2) Epoch 17, batch 10600, loss[loss=0.2116, simple_loss=0.2935, pruned_loss=0.06486, over 12065.00 frames. ], tot_loss[loss=0.1634, simple_loss=0.2531, pruned_loss=0.03686, over 2377352.67 frames. ], batch size: 40, lr: 3.28e-03, grad_scale: 16.0 2023-05-17 09:17:10,795 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=306356.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 09:17:30,654 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=306384.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 09:17:38,227 INFO [finetune.py:992] (0/2) Epoch 17, batch 10650, loss[loss=0.1309, simple_loss=0.2152, pruned_loss=0.02328, over 11725.00 frames. ], tot_loss[loss=0.1634, simple_loss=0.2529, pruned_loss=0.03695, over 2376090.08 frames. ], batch size: 26, lr: 3.28e-03, grad_scale: 16.0 2023-05-17 09:17:38,923 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.819e+02 2.440e+02 2.991e+02 3.762e+02 6.396e+02, threshold=5.981e+02, percent-clipped=0.0 2023-05-17 09:18:14,685 INFO [finetune.py:992] (0/2) Epoch 17, batch 10700, loss[loss=0.1851, simple_loss=0.275, pruned_loss=0.04758, over 12107.00 frames. ], tot_loss[loss=0.1633, simple_loss=0.2531, pruned_loss=0.03678, over 2377858.03 frames. ], batch size: 38, lr: 3.28e-03, grad_scale: 16.0 2023-05-17 09:18:14,918 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=306445.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 09:18:20,431 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([6.1740, 6.1143, 5.9419, 5.4544, 5.3229, 6.0864, 5.6997, 5.4443], device='cuda:0'), covar=tensor([0.0708, 0.1039, 0.0630, 0.1603, 0.0733, 0.0740, 0.1444, 0.1081], device='cuda:0'), in_proj_covar=tensor([0.0663, 0.0588, 0.0539, 0.0664, 0.0442, 0.0761, 0.0816, 0.0593], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0004, 0.0004, 0.0003], device='cuda:0') 2023-05-17 09:18:39,789 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2023-05-17 09:18:49,907 INFO [finetune.py:992] (0/2) Epoch 17, batch 10750, loss[loss=0.1616, simple_loss=0.2557, pruned_loss=0.03371, over 12194.00 frames. ], tot_loss[loss=0.1634, simple_loss=0.2533, pruned_loss=0.03677, over 2372011.86 frames. ], batch size: 35, lr: 3.28e-03, grad_scale: 16.0 2023-05-17 09:18:50,615 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.791e+02 2.600e+02 3.061e+02 3.731e+02 9.933e+02, threshold=6.123e+02, percent-clipped=1.0 2023-05-17 09:18:59,971 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=306508.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 09:19:25,916 INFO [finetune.py:992] (0/2) Epoch 17, batch 10800, loss[loss=0.1799, simple_loss=0.2713, pruned_loss=0.0443, over 12127.00 frames. ], tot_loss[loss=0.1639, simple_loss=0.2538, pruned_loss=0.03695, over 2369424.66 frames. ], batch size: 39, lr: 3.28e-03, grad_scale: 16.0 2023-05-17 09:19:29,529 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=306550.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 09:19:33,860 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=306556.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 09:20:02,312 INFO [finetune.py:992] (0/2) Epoch 17, batch 10850, loss[loss=0.1564, simple_loss=0.2549, pruned_loss=0.02891, over 12197.00 frames. ], tot_loss[loss=0.1634, simple_loss=0.2534, pruned_loss=0.03667, over 2371855.25 frames. ], batch size: 35, lr: 3.28e-03, grad_scale: 16.0 2023-05-17 09:20:03,002 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.932e+02 2.667e+02 3.157e+02 3.680e+02 8.758e+02, threshold=6.314e+02, percent-clipped=1.0 2023-05-17 09:20:15,133 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.3574, 6.1649, 5.7427, 5.7497, 6.2728, 5.4760, 5.7363, 5.8264], device='cuda:0'), covar=tensor([0.1428, 0.0836, 0.0921, 0.1685, 0.0819, 0.2418, 0.1733, 0.1087], device='cuda:0'), in_proj_covar=tensor([0.0374, 0.0522, 0.0421, 0.0466, 0.0481, 0.0456, 0.0418, 0.0404], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-17 09:20:15,466 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.12 vs. limit=2.0 2023-05-17 09:20:23,413 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.5825, 2.7941, 3.3094, 4.4030, 2.5736, 4.4243, 4.5685, 4.6266], device='cuda:0'), covar=tensor([0.0123, 0.1207, 0.0482, 0.0157, 0.1311, 0.0250, 0.0137, 0.0098], device='cuda:0'), in_proj_covar=tensor([0.0126, 0.0208, 0.0188, 0.0126, 0.0193, 0.0185, 0.0183, 0.0130], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-17 09:20:40,233 INFO [finetune.py:992] (0/2) Epoch 17, batch 10900, loss[loss=0.2037, simple_loss=0.3045, pruned_loss=0.05144, over 12118.00 frames. ], tot_loss[loss=0.1634, simple_loss=0.2537, pruned_loss=0.03655, over 2366346.29 frames. ], batch size: 38, lr: 3.28e-03, grad_scale: 16.0 2023-05-17 09:21:10,656 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.3431, 5.1222, 5.2499, 5.3228, 4.8400, 4.9844, 4.7251, 5.2437], device='cuda:0'), covar=tensor([0.0670, 0.0653, 0.0981, 0.0622, 0.2146, 0.1367, 0.0624, 0.1001], device='cuda:0'), in_proj_covar=tensor([0.0567, 0.0734, 0.0652, 0.0663, 0.0888, 0.0782, 0.0589, 0.0505], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0005, 0.0004, 0.0004, 0.0005, 0.0005, 0.0004, 0.0003], device='cuda:0') 2023-05-17 09:21:14,269 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=306692.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 09:21:16,261 INFO [finetune.py:992] (0/2) Epoch 17, batch 10950, loss[loss=0.2205, simple_loss=0.3129, pruned_loss=0.06404, over 11157.00 frames. ], tot_loss[loss=0.1637, simple_loss=0.254, pruned_loss=0.03675, over 2369471.54 frames. ], batch size: 55, lr: 3.28e-03, grad_scale: 16.0 2023-05-17 09:21:16,960 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.710e+02 2.498e+02 2.941e+02 3.485e+02 6.942e+02, threshold=5.883e+02, percent-clipped=2.0 2023-05-17 09:21:48,833 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=306740.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 09:21:52,290 INFO [finetune.py:992] (0/2) Epoch 17, batch 11000, loss[loss=0.1417, simple_loss=0.2288, pruned_loss=0.02729, over 12173.00 frames. ], tot_loss[loss=0.1652, simple_loss=0.2556, pruned_loss=0.03736, over 2364353.34 frames. ], batch size: 29, lr: 3.28e-03, grad_scale: 16.0 2023-05-17 09:21:55,271 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.7989, 4.5378, 4.6103, 4.7094, 4.4349, 4.7940, 4.6414, 2.7366], device='cuda:0'), covar=tensor([0.0100, 0.0087, 0.0104, 0.0067, 0.0066, 0.0096, 0.0102, 0.0794], device='cuda:0'), in_proj_covar=tensor([0.0074, 0.0083, 0.0088, 0.0078, 0.0063, 0.0099, 0.0087, 0.0102], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-17 09:21:58,205 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=306753.0, num_to_drop=1, layers_to_drop={2} 2023-05-17 09:22:15,827 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.6054, 4.3083, 4.5493, 4.0811, 4.3547, 4.1505, 4.5678, 4.2811], device='cuda:0'), covar=tensor([0.0388, 0.0448, 0.0399, 0.0318, 0.0432, 0.0403, 0.0315, 0.0598], device='cuda:0'), in_proj_covar=tensor([0.0279, 0.0284, 0.0308, 0.0280, 0.0278, 0.0276, 0.0253, 0.0227], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-17 09:22:28,093 INFO [finetune.py:992] (0/2) Epoch 17, batch 11050, loss[loss=0.1479, simple_loss=0.2478, pruned_loss=0.02406, over 12309.00 frames. ], tot_loss[loss=0.1678, simple_loss=0.2581, pruned_loss=0.03876, over 2334803.45 frames. ], batch size: 34, lr: 3.28e-03, grad_scale: 16.0 2023-05-17 09:22:28,811 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.776e+02 2.828e+02 3.526e+02 4.495e+02 1.369e+03, threshold=7.052e+02, percent-clipped=10.0 2023-05-17 09:22:48,474 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.3886, 3.5841, 3.3022, 3.7191, 3.4747, 2.6130, 3.2126, 2.7928], device='cuda:0'), covar=tensor([0.0966, 0.1074, 0.1558, 0.0889, 0.1209, 0.1819, 0.1447, 0.3229], device='cuda:0'), in_proj_covar=tensor([0.0315, 0.0386, 0.0369, 0.0339, 0.0380, 0.0280, 0.0355, 0.0374], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-17 09:23:03,929 INFO [finetune.py:992] (0/2) Epoch 17, batch 11100, loss[loss=0.2139, simple_loss=0.3035, pruned_loss=0.06212, over 11202.00 frames. ], tot_loss[loss=0.1723, simple_loss=0.262, pruned_loss=0.04132, over 2294645.65 frames. ], batch size: 55, lr: 3.28e-03, grad_scale: 16.0 2023-05-17 09:23:07,627 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=306850.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 09:23:32,802 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.3753, 4.2745, 4.2683, 3.7735, 4.1340, 4.4341, 3.8287, 3.8947], device='cuda:0'), covar=tensor([0.1540, 0.1878, 0.1371, 0.2845, 0.2471, 0.1484, 0.2845, 0.3122], device='cuda:0'), in_proj_covar=tensor([0.0662, 0.0591, 0.0540, 0.0668, 0.0443, 0.0759, 0.0816, 0.0595], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0004, 0.0004, 0.0003], device='cuda:0') 2023-05-17 09:23:38,824 INFO [finetune.py:992] (0/2) Epoch 17, batch 11150, loss[loss=0.3344, simple_loss=0.3893, pruned_loss=0.1397, over 6256.00 frames. ], tot_loss[loss=0.18, simple_loss=0.2688, pruned_loss=0.04562, over 2224790.91 frames. ], batch size: 98, lr: 3.28e-03, grad_scale: 16.0 2023-05-17 09:23:39,462 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.968e+02 3.233e+02 3.980e+02 4.807e+02 1.094e+03, threshold=7.960e+02, percent-clipped=4.0 2023-05-17 09:23:40,803 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=306898.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 09:24:14,242 INFO [finetune.py:992] (0/2) Epoch 17, batch 11200, loss[loss=0.2983, simple_loss=0.356, pruned_loss=0.1203, over 7883.00 frames. ], tot_loss[loss=0.1859, simple_loss=0.2739, pruned_loss=0.04891, over 2172728.91 frames. ], batch size: 100, lr: 3.28e-03, grad_scale: 16.0 2023-05-17 09:24:49,642 INFO [finetune.py:992] (0/2) Epoch 17, batch 11250, loss[loss=0.2114, simple_loss=0.31, pruned_loss=0.05644, over 10244.00 frames. ], tot_loss[loss=0.1944, simple_loss=0.2814, pruned_loss=0.05373, over 2099544.40 frames. ], batch size: 69, lr: 3.28e-03, grad_scale: 16.0 2023-05-17 09:24:50,188 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.250e+02 3.461e+02 4.288e+02 5.110e+02 1.210e+03, threshold=8.577e+02, percent-clipped=4.0 2023-05-17 09:25:21,045 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=307040.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 09:25:24,337 INFO [finetune.py:992] (0/2) Epoch 17, batch 11300, loss[loss=0.1824, simple_loss=0.277, pruned_loss=0.04393, over 12346.00 frames. ], tot_loss[loss=0.1999, simple_loss=0.2864, pruned_loss=0.05669, over 2063158.36 frames. ], batch size: 36, lr: 3.28e-03, grad_scale: 16.0 2023-05-17 09:25:26,589 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=307048.0, num_to_drop=1, layers_to_drop={0} 2023-05-17 09:25:47,306 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.8470, 2.2489, 2.8343, 2.7765, 2.9266, 2.9054, 2.7983, 2.3590], device='cuda:0'), covar=tensor([0.0093, 0.0346, 0.0176, 0.0101, 0.0120, 0.0115, 0.0145, 0.0374], device='cuda:0'), in_proj_covar=tensor([0.0091, 0.0124, 0.0106, 0.0081, 0.0106, 0.0119, 0.0104, 0.0141], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-17 09:25:55,071 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=307088.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 09:25:59,693 INFO [finetune.py:992] (0/2) Epoch 17, batch 11350, loss[loss=0.2446, simple_loss=0.3229, pruned_loss=0.08321, over 6834.00 frames. ], tot_loss[loss=0.2062, simple_loss=0.2918, pruned_loss=0.0603, over 1997119.46 frames. ], batch size: 98, lr: 3.28e-03, grad_scale: 16.0 2023-05-17 09:26:00,342 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.955e+02 3.468e+02 4.102e+02 4.903e+02 8.634e+02, threshold=8.204e+02, percent-clipped=1.0 2023-05-17 09:26:34,221 INFO [finetune.py:992] (0/2) Epoch 17, batch 11400, loss[loss=0.2368, simple_loss=0.3139, pruned_loss=0.07991, over 7177.00 frames. ], tot_loss[loss=0.2108, simple_loss=0.2957, pruned_loss=0.06296, over 1940651.80 frames. ], batch size: 101, lr: 3.28e-03, grad_scale: 16.0 2023-05-17 09:27:02,099 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.4467, 4.0590, 4.2175, 4.3341, 4.2492, 4.3849, 4.3396, 2.4978], device='cuda:0'), covar=tensor([0.0074, 0.0107, 0.0119, 0.0067, 0.0059, 0.0118, 0.0074, 0.0957], device='cuda:0'), in_proj_covar=tensor([0.0072, 0.0082, 0.0088, 0.0077, 0.0063, 0.0098, 0.0086, 0.0101], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-17 09:27:09,343 INFO [finetune.py:992] (0/2) Epoch 17, batch 11450, loss[loss=0.2651, simple_loss=0.3307, pruned_loss=0.09973, over 6868.00 frames. ], tot_loss[loss=0.2149, simple_loss=0.2987, pruned_loss=0.06555, over 1899489.46 frames. ], batch size: 99, lr: 3.28e-03, grad_scale: 16.0 2023-05-17 09:27:09,987 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.192e+02 3.229e+02 3.849e+02 4.739e+02 1.180e+03, threshold=7.698e+02, percent-clipped=2.0 2023-05-17 09:27:42,460 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.2675, 3.4639, 3.1929, 3.4989, 3.3654, 2.5747, 3.1185, 2.8474], device='cuda:0'), covar=tensor([0.0905, 0.1041, 0.1626, 0.0895, 0.1529, 0.1784, 0.1478, 0.2824], device='cuda:0'), in_proj_covar=tensor([0.0305, 0.0372, 0.0355, 0.0325, 0.0366, 0.0271, 0.0342, 0.0362], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-17 09:27:43,512 INFO [finetune.py:992] (0/2) Epoch 17, batch 11500, loss[loss=0.2129, simple_loss=0.293, pruned_loss=0.06636, over 11177.00 frames. ], tot_loss[loss=0.2189, simple_loss=0.3016, pruned_loss=0.06812, over 1852775.16 frames. ], batch size: 55, lr: 3.28e-03, grad_scale: 16.0 2023-05-17 09:28:04,810 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.9543, 2.3024, 2.8715, 2.8818, 2.9946, 3.0044, 2.9421, 2.4311], device='cuda:0'), covar=tensor([0.0088, 0.0363, 0.0187, 0.0095, 0.0107, 0.0104, 0.0128, 0.0381], device='cuda:0'), in_proj_covar=tensor([0.0090, 0.0123, 0.0106, 0.0080, 0.0105, 0.0118, 0.0103, 0.0140], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-17 09:28:19,071 INFO [finetune.py:992] (0/2) Epoch 17, batch 11550, loss[loss=0.1999, simple_loss=0.29, pruned_loss=0.05485, over 10455.00 frames. ], tot_loss[loss=0.2223, simple_loss=0.304, pruned_loss=0.07028, over 1811980.73 frames. ], batch size: 69, lr: 3.28e-03, grad_scale: 8.0 2023-05-17 09:28:20,291 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.952e+02 3.645e+02 4.042e+02 4.908e+02 1.002e+03, threshold=8.085e+02, percent-clipped=3.0 2023-05-17 09:28:35,175 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.8526, 4.4407, 4.1314, 4.1722, 4.4928, 4.0284, 4.1054, 4.0228], device='cuda:0'), covar=tensor([0.1740, 0.1130, 0.1212, 0.1726, 0.1033, 0.1987, 0.1734, 0.1264], device='cuda:0'), in_proj_covar=tensor([0.0359, 0.0500, 0.0406, 0.0450, 0.0462, 0.0437, 0.0404, 0.0388], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-17 09:28:51,397 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=307342.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 09:28:53,866 INFO [finetune.py:992] (0/2) Epoch 17, batch 11600, loss[loss=0.2309, simple_loss=0.3124, pruned_loss=0.07465, over 6925.00 frames. ], tot_loss[loss=0.2234, simple_loss=0.3046, pruned_loss=0.07113, over 1783735.37 frames. ], batch size: 100, lr: 3.27e-03, grad_scale: 8.0 2023-05-17 09:28:54,711 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.0241, 4.8917, 4.8442, 4.9289, 4.5447, 5.0727, 5.0905, 5.1624], device='cuda:0'), covar=tensor([0.0195, 0.0159, 0.0171, 0.0345, 0.0695, 0.0240, 0.0124, 0.0185], device='cuda:0'), in_proj_covar=tensor([0.0191, 0.0191, 0.0185, 0.0240, 0.0233, 0.0213, 0.0170, 0.0227], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:0') 2023-05-17 09:28:56,090 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=307348.0, num_to_drop=1, layers_to_drop={1} 2023-05-17 09:29:29,965 INFO [finetune.py:992] (0/2) Epoch 17, batch 11650, loss[loss=0.2086, simple_loss=0.2937, pruned_loss=0.06177, over 12121.00 frames. ], tot_loss[loss=0.2233, simple_loss=0.3039, pruned_loss=0.07129, over 1772431.12 frames. ], batch size: 38, lr: 3.27e-03, grad_scale: 8.0 2023-05-17 09:29:30,625 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.40 vs. limit=2.0 2023-05-17 09:29:30,867 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=307396.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 09:29:31,449 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.341e+02 3.378e+02 3.794e+02 4.382e+02 6.754e+02, threshold=7.588e+02, percent-clipped=0.0 2023-05-17 09:29:36,497 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=307403.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 09:30:05,914 INFO [finetune.py:992] (0/2) Epoch 17, batch 11700, loss[loss=0.281, simple_loss=0.3462, pruned_loss=0.1079, over 6667.00 frames. ], tot_loss[loss=0.2241, simple_loss=0.304, pruned_loss=0.07207, over 1746539.85 frames. ], batch size: 98, lr: 3.27e-03, grad_scale: 8.0 2023-05-17 09:30:12,297 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.15 vs. limit=2.0 2023-05-17 09:30:40,540 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.0650, 1.9583, 2.1728, 2.0644, 2.1549, 2.2730, 1.8736, 2.1937], device='cuda:0'), covar=tensor([0.0108, 0.0256, 0.0158, 0.0186, 0.0148, 0.0157, 0.0275, 0.0132], device='cuda:0'), in_proj_covar=tensor([0.0180, 0.0207, 0.0192, 0.0188, 0.0219, 0.0168, 0.0199, 0.0190], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-17 09:30:40,964 INFO [finetune.py:992] (0/2) Epoch 17, batch 11750, loss[loss=0.2499, simple_loss=0.2999, pruned_loss=0.09992, over 6366.00 frames. ], tot_loss[loss=0.2245, simple_loss=0.304, pruned_loss=0.07253, over 1724811.17 frames. ], batch size: 99, lr: 3.27e-03, grad_scale: 8.0 2023-05-17 09:30:42,311 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.161e+02 3.492e+02 4.034e+02 4.769e+02 1.195e+03, threshold=8.068e+02, percent-clipped=4.0 2023-05-17 09:30:44,547 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.2724, 2.3738, 3.1341, 4.2432, 2.3215, 4.3429, 4.1264, 4.2944], device='cuda:0'), covar=tensor([0.0124, 0.1415, 0.0525, 0.0160, 0.1466, 0.0168, 0.0268, 0.0129], device='cuda:0'), in_proj_covar=tensor([0.0121, 0.0202, 0.0181, 0.0120, 0.0186, 0.0177, 0.0175, 0.0125], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-17 09:30:47,406 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.93 vs. limit=2.0 2023-05-17 09:31:15,030 INFO [finetune.py:992] (0/2) Epoch 17, batch 11800, loss[loss=0.2448, simple_loss=0.3103, pruned_loss=0.08969, over 6488.00 frames. ], tot_loss[loss=0.2263, simple_loss=0.3056, pruned_loss=0.07354, over 1723694.97 frames. ], batch size: 101, lr: 3.27e-03, grad_scale: 8.0 2023-05-17 09:31:31,217 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.7553, 3.0172, 2.4871, 2.2383, 2.8050, 2.3786, 2.9712, 2.5926], device='cuda:0'), covar=tensor([0.0593, 0.0502, 0.0882, 0.1291, 0.0239, 0.1075, 0.0430, 0.0756], device='cuda:0'), in_proj_covar=tensor([0.0185, 0.0249, 0.0173, 0.0196, 0.0139, 0.0180, 0.0195, 0.0172], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-17 09:31:35,876 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.8129, 3.7899, 3.7875, 3.8831, 3.6835, 3.7124, 3.5909, 3.7832], device='cuda:0'), covar=tensor([0.1097, 0.0651, 0.1279, 0.0676, 0.1386, 0.1182, 0.0568, 0.1035], device='cuda:0'), in_proj_covar=tensor([0.0524, 0.0678, 0.0604, 0.0611, 0.0810, 0.0719, 0.0542, 0.0470], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0003], device='cuda:0') 2023-05-17 09:31:50,565 INFO [finetune.py:992] (0/2) Epoch 17, batch 11850, loss[loss=0.234, simple_loss=0.3047, pruned_loss=0.08162, over 6742.00 frames. ], tot_loss[loss=0.2283, simple_loss=0.3072, pruned_loss=0.07476, over 1693695.36 frames. ], batch size: 98, lr: 3.27e-03, grad_scale: 8.0 2023-05-17 09:31:51,996 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.549e+02 3.416e+02 4.242e+02 5.268e+02 1.514e+03, threshold=8.485e+02, percent-clipped=2.0 2023-05-17 09:31:52,867 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=307598.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 09:31:57,127 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=307604.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 09:32:08,408 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=307620.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 09:32:15,957 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=2.77 vs. limit=5.0 2023-05-17 09:32:25,775 INFO [finetune.py:992] (0/2) Epoch 17, batch 11900, loss[loss=0.1846, simple_loss=0.2803, pruned_loss=0.04451, over 11222.00 frames. ], tot_loss[loss=0.2282, simple_loss=0.3078, pruned_loss=0.07433, over 1675882.45 frames. ], batch size: 55, lr: 3.27e-03, grad_scale: 8.0 2023-05-17 09:32:35,312 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=307659.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 09:32:39,364 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=307665.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 09:32:50,797 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=307681.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 09:32:53,557 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.0939, 2.0046, 2.1984, 2.0884, 2.1989, 2.3336, 1.8582, 2.2602], device='cuda:0'), covar=tensor([0.0127, 0.0322, 0.0189, 0.0212, 0.0165, 0.0188, 0.0341, 0.0147], device='cuda:0'), in_proj_covar=tensor([0.0175, 0.0202, 0.0188, 0.0184, 0.0214, 0.0164, 0.0195, 0.0185], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-17 09:33:00,032 INFO [finetune.py:992] (0/2) Epoch 17, batch 11950, loss[loss=0.1772, simple_loss=0.2673, pruned_loss=0.04354, over 6772.00 frames. ], tot_loss[loss=0.2238, simple_loss=0.3045, pruned_loss=0.07153, over 1676445.44 frames. ], batch size: 98, lr: 3.27e-03, grad_scale: 8.0 2023-05-17 09:33:01,320 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.326e+02 3.144e+02 3.799e+02 4.450e+02 7.453e+02, threshold=7.599e+02, percent-clipped=0.0 2023-05-17 09:33:02,185 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=307698.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 09:33:17,336 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.9033, 2.3350, 3.3742, 2.8705, 3.1756, 3.1046, 2.3853, 3.2623], device='cuda:0'), covar=tensor([0.0149, 0.0445, 0.0154, 0.0324, 0.0178, 0.0207, 0.0460, 0.0166], device='cuda:0'), in_proj_covar=tensor([0.0176, 0.0203, 0.0188, 0.0184, 0.0214, 0.0164, 0.0196, 0.0185], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-17 09:33:35,909 INFO [finetune.py:992] (0/2) Epoch 17, batch 12000, loss[loss=0.2164, simple_loss=0.293, pruned_loss=0.06992, over 6971.00 frames. ], tot_loss[loss=0.2176, simple_loss=0.2999, pruned_loss=0.06759, over 1682460.44 frames. ], batch size: 98, lr: 3.27e-03, grad_scale: 8.0 2023-05-17 09:33:35,910 INFO [finetune.py:1017] (0/2) Computing validation loss 2023-05-17 09:33:43,569 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.2559, 5.1673, 5.2558, 5.2929, 4.9165, 4.9040, 4.8685, 5.0703], device='cuda:0'), covar=tensor([0.0672, 0.0526, 0.0775, 0.0495, 0.1911, 0.1438, 0.0529, 0.1304], device='cuda:0'), in_proj_covar=tensor([0.0523, 0.0674, 0.0600, 0.0610, 0.0807, 0.0717, 0.0538, 0.0469], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0003], device='cuda:0') 2023-05-17 09:33:53,997 INFO [finetune.py:1026] (0/2) Epoch 17, validation: loss=0.2896, simple_loss=0.3623, pruned_loss=0.1084, over 1020973.00 frames. 2023-05-17 09:33:53,998 INFO [finetune.py:1027] (0/2) Maximum memory allocated so far is 12827MB 2023-05-17 09:34:07,017 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.81 vs. limit=2.0 2023-05-17 09:34:07,622 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.0535, 1.9719, 2.2406, 2.1029, 2.1741, 2.3122, 1.9068, 2.2304], device='cuda:0'), covar=tensor([0.0133, 0.0361, 0.0177, 0.0228, 0.0163, 0.0189, 0.0349, 0.0152], device='cuda:0'), in_proj_covar=tensor([0.0175, 0.0203, 0.0187, 0.0184, 0.0214, 0.0164, 0.0195, 0.0185], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-17 09:34:28,284 INFO [finetune.py:992] (0/2) Epoch 17, batch 12050, loss[loss=0.2041, simple_loss=0.2881, pruned_loss=0.06006, over 7056.00 frames. ], tot_loss[loss=0.2128, simple_loss=0.2962, pruned_loss=0.0647, over 1685458.98 frames. ], batch size: 97, lr: 3.27e-03, grad_scale: 8.0 2023-05-17 09:34:29,603 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.966e+02 2.957e+02 3.433e+02 3.945e+02 6.254e+02, threshold=6.866e+02, percent-clipped=0.0 2023-05-17 09:35:01,338 INFO [finetune.py:992] (0/2) Epoch 17, batch 12100, loss[loss=0.2446, simple_loss=0.3103, pruned_loss=0.08946, over 6582.00 frames. ], tot_loss[loss=0.2117, simple_loss=0.2951, pruned_loss=0.06418, over 1667736.54 frames. ], batch size: 99, lr: 3.27e-03, grad_scale: 8.0 2023-05-17 09:35:20,153 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.4011, 3.1861, 3.0946, 3.2498, 2.6982, 3.2269, 2.5909, 2.6954], device='cuda:0'), covar=tensor([0.1551, 0.0870, 0.0818, 0.0537, 0.1019, 0.0755, 0.1683, 0.0587], device='cuda:0'), in_proj_covar=tensor([0.0228, 0.0266, 0.0293, 0.0349, 0.0240, 0.0241, 0.0259, 0.0362], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-17 09:35:33,744 INFO [finetune.py:992] (0/2) Epoch 17, batch 12150, loss[loss=0.2494, simple_loss=0.319, pruned_loss=0.08989, over 6984.00 frames. ], tot_loss[loss=0.2122, simple_loss=0.2957, pruned_loss=0.06437, over 1671698.44 frames. ], batch size: 98, lr: 3.27e-03, grad_scale: 8.0 2023-05-17 09:35:35,029 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.236e+02 3.185e+02 3.632e+02 4.072e+02 1.178e+03, threshold=7.264e+02, percent-clipped=3.0 2023-05-17 09:36:01,245 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.6890, 3.5092, 3.4815, 3.6344, 3.5872, 3.7417, 3.5916, 2.6641], device='cuda:0'), covar=tensor([0.0109, 0.0098, 0.0169, 0.0090, 0.0080, 0.0118, 0.0095, 0.0778], device='cuda:0'), in_proj_covar=tensor([0.0070, 0.0079, 0.0085, 0.0074, 0.0061, 0.0094, 0.0082, 0.0099], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-17 09:36:05,406 INFO [finetune.py:992] (0/2) Epoch 17, batch 12200, loss[loss=0.2409, simple_loss=0.304, pruned_loss=0.08893, over 6672.00 frames. ], tot_loss[loss=0.2134, simple_loss=0.2967, pruned_loss=0.06508, over 1668016.32 frames. ], batch size: 98, lr: 3.27e-03, grad_scale: 8.0 2023-05-17 09:36:10,980 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=307954.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 09:36:14,660 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=307960.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 09:36:24,610 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=307976.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 09:36:27,973 INFO [checkpoint.py:75] (0/2) Saving checkpoint to pruned_transducer_stateless7/exp_giga_finetune/epoch-17.pt 2023-05-17 09:36:50,395 INFO [finetune.py:992] (0/2) Epoch 18, batch 0, loss[loss=0.1662, simple_loss=0.2553, pruned_loss=0.03852, over 12294.00 frames. ], tot_loss[loss=0.1662, simple_loss=0.2553, pruned_loss=0.03852, over 12294.00 frames. ], batch size: 34, lr: 3.27e-03, grad_scale: 8.0 2023-05-17 09:36:50,395 INFO [finetune.py:1017] (0/2) Computing validation loss 2023-05-17 09:37:07,740 INFO [finetune.py:1026] (0/2) Epoch 18, validation: loss=0.289, simple_loss=0.361, pruned_loss=0.1085, over 1020973.00 frames. 2023-05-17 09:37:07,741 INFO [finetune.py:1027] (0/2) Maximum memory allocated so far is 12827MB 2023-05-17 09:37:19,681 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.214e+02 3.181e+02 3.643e+02 4.435e+02 7.153e+02, threshold=7.285e+02, percent-clipped=0.0 2023-05-17 09:37:20,540 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=307998.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 09:37:22,081 INFO [checkpoint.py:75] (0/2) Saving checkpoint to pruned_transducer_stateless7/exp_giga_finetune/checkpoint-208000.pt 2023-05-17 09:37:46,445 INFO [finetune.py:992] (0/2) Epoch 18, batch 50, loss[loss=0.1484, simple_loss=0.2371, pruned_loss=0.02981, over 12329.00 frames. ], tot_loss[loss=0.1712, simple_loss=0.2623, pruned_loss=0.04004, over 534452.84 frames. ], batch size: 31, lr: 3.27e-03, grad_scale: 8.0 2023-05-17 09:37:57,580 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=308046.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 09:38:22,608 INFO [finetune.py:992] (0/2) Epoch 18, batch 100, loss[loss=0.1639, simple_loss=0.254, pruned_loss=0.03687, over 12346.00 frames. ], tot_loss[loss=0.1725, simple_loss=0.2639, pruned_loss=0.04054, over 951676.20 frames. ], batch size: 31, lr: 3.27e-03, grad_scale: 8.0 2023-05-17 09:38:35,363 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.554e+02 2.661e+02 3.037e+02 3.590e+02 6.074e+02, threshold=6.073e+02, percent-clipped=0.0 2023-05-17 09:38:58,949 INFO [finetune.py:992] (0/2) Epoch 18, batch 150, loss[loss=0.188, simple_loss=0.2823, pruned_loss=0.04685, over 11580.00 frames. ], tot_loss[loss=0.1727, simple_loss=0.264, pruned_loss=0.0407, over 1257593.71 frames. ], batch size: 48, lr: 3.27e-03, grad_scale: 4.0 2023-05-17 09:39:34,692 INFO [finetune.py:992] (0/2) Epoch 18, batch 200, loss[loss=0.1653, simple_loss=0.2539, pruned_loss=0.03839, over 12151.00 frames. ], tot_loss[loss=0.1701, simple_loss=0.2612, pruned_loss=0.03945, over 1508532.93 frames. ], batch size: 29, lr: 3.27e-03, grad_scale: 4.0 2023-05-17 09:39:47,605 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.776e+02 2.638e+02 3.218e+02 3.807e+02 6.334e+02, threshold=6.435e+02, percent-clipped=2.0 2023-05-17 09:39:48,083 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.69 vs. limit=2.0 2023-05-17 09:39:51,015 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.3184, 5.1465, 5.2493, 5.2995, 4.9097, 4.9553, 4.7074, 5.1941], device='cuda:0'), covar=tensor([0.0726, 0.0652, 0.0889, 0.0615, 0.2136, 0.1504, 0.0590, 0.1189], device='cuda:0'), in_proj_covar=tensor([0.0527, 0.0682, 0.0609, 0.0615, 0.0815, 0.0729, 0.0545, 0.0475], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0004, 0.0004, 0.0005, 0.0004, 0.0003, 0.0003], device='cuda:0') 2023-05-17 09:39:51,092 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=308202.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 09:40:11,467 INFO [finetune.py:992] (0/2) Epoch 18, batch 250, loss[loss=0.1524, simple_loss=0.2389, pruned_loss=0.03298, over 12303.00 frames. ], tot_loss[loss=0.1678, simple_loss=0.2587, pruned_loss=0.0384, over 1708577.90 frames. ], batch size: 33, lr: 3.27e-03, grad_scale: 4.0 2023-05-17 09:40:20,555 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.77 vs. limit=2.0 2023-05-17 09:40:22,566 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2023-05-17 09:40:28,772 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=308254.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 09:40:33,066 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=308260.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 09:40:35,324 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=308263.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 09:40:44,443 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=308276.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 09:40:47,193 INFO [finetune.py:992] (0/2) Epoch 18, batch 300, loss[loss=0.1665, simple_loss=0.2674, pruned_loss=0.03275, over 12123.00 frames. ], tot_loss[loss=0.1671, simple_loss=0.2579, pruned_loss=0.0382, over 1855675.04 frames. ], batch size: 38, lr: 3.27e-03, grad_scale: 4.0 2023-05-17 09:41:00,538 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.864e+02 2.598e+02 3.095e+02 3.618e+02 6.538e+02, threshold=6.191e+02, percent-clipped=1.0 2023-05-17 09:41:03,441 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=308302.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 09:41:07,613 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=308308.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 09:41:08,425 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=308309.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 09:41:09,209 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.5771, 2.9739, 3.6202, 4.5053, 3.8054, 4.4727, 3.7447, 3.1345], device='cuda:0'), covar=tensor([0.0033, 0.0336, 0.0170, 0.0041, 0.0136, 0.0086, 0.0144, 0.0399], device='cuda:0'), in_proj_covar=tensor([0.0091, 0.0125, 0.0106, 0.0081, 0.0105, 0.0119, 0.0103, 0.0141], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-17 09:41:19,206 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=308324.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 09:41:23,345 INFO [finetune.py:992] (0/2) Epoch 18, batch 350, loss[loss=0.1432, simple_loss=0.2305, pruned_loss=0.02791, over 12286.00 frames. ], tot_loss[loss=0.167, simple_loss=0.2577, pruned_loss=0.03816, over 1963994.18 frames. ], batch size: 33, lr: 3.27e-03, grad_scale: 4.0 2023-05-17 09:41:24,105 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([6.0716, 6.0618, 5.7672, 5.3649, 5.1852, 5.9495, 5.6357, 5.3600], device='cuda:0'), covar=tensor([0.0712, 0.0871, 0.0704, 0.1541, 0.0739, 0.0676, 0.1424, 0.1008], device='cuda:0'), in_proj_covar=tensor([0.0640, 0.0568, 0.0516, 0.0640, 0.0426, 0.0724, 0.0779, 0.0571], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002], device='cuda:0') 2023-05-17 09:41:52,811 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=308370.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 09:41:59,827 INFO [finetune.py:992] (0/2) Epoch 18, batch 400, loss[loss=0.1394, simple_loss=0.2203, pruned_loss=0.02928, over 12283.00 frames. ], tot_loss[loss=0.1649, simple_loss=0.2549, pruned_loss=0.0374, over 2065552.99 frames. ], batch size: 28, lr: 3.27e-03, grad_scale: 8.0 2023-05-17 09:42:01,831 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.41 vs. limit=2.0 2023-05-17 09:42:12,391 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.787e+02 2.569e+02 3.167e+02 3.621e+02 6.825e+02, threshold=6.334e+02, percent-clipped=2.0 2023-05-17 09:42:31,214 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.2928, 4.6982, 4.0539, 4.9391, 4.5598, 2.9023, 4.1399, 3.0098], device='cuda:0'), covar=tensor([0.0836, 0.0798, 0.1470, 0.0543, 0.1043, 0.1864, 0.1160, 0.3556], device='cuda:0'), in_proj_covar=tensor([0.0310, 0.0379, 0.0360, 0.0329, 0.0371, 0.0277, 0.0348, 0.0369], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-17 09:42:35,239 INFO [finetune.py:992] (0/2) Epoch 18, batch 450, loss[loss=0.1661, simple_loss=0.2608, pruned_loss=0.03567, over 12373.00 frames. ], tot_loss[loss=0.1656, simple_loss=0.2559, pruned_loss=0.03766, over 2131878.31 frames. ], batch size: 35, lr: 3.27e-03, grad_scale: 8.0 2023-05-17 09:42:38,186 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.0234, 3.0367, 4.4253, 2.3672, 2.6056, 3.2491, 2.9857, 3.3852], device='cuda:0'), covar=tensor([0.0536, 0.1365, 0.0414, 0.1451, 0.2185, 0.1774, 0.1451, 0.1385], device='cuda:0'), in_proj_covar=tensor([0.0233, 0.0234, 0.0251, 0.0182, 0.0237, 0.0290, 0.0224, 0.0264], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-17 09:43:07,249 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([2.9992, 2.4382, 3.5750, 3.0269, 3.4140, 3.2034, 2.6430, 3.5226], device='cuda:0'), covar=tensor([0.0166, 0.0427, 0.0199, 0.0277, 0.0181, 0.0207, 0.0395, 0.0162], device='cuda:0'), in_proj_covar=tensor([0.0181, 0.0208, 0.0193, 0.0189, 0.0220, 0.0169, 0.0200, 0.0190], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-17 09:43:11,321 INFO [finetune.py:992] (0/2) Epoch 18, batch 500, loss[loss=0.188, simple_loss=0.2792, pruned_loss=0.04841, over 12270.00 frames. ], tot_loss[loss=0.1659, simple_loss=0.256, pruned_loss=0.03788, over 2185459.06 frames. ], batch size: 37, lr: 3.26e-03, grad_scale: 8.0 2023-05-17 09:43:24,086 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.574e+02 2.639e+02 3.231e+02 3.895e+02 6.138e+02, threshold=6.461e+02, percent-clipped=0.0 2023-05-17 09:43:35,223 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=4.68 vs. limit=5.0 2023-05-17 09:43:47,327 INFO [finetune.py:992] (0/2) Epoch 18, batch 550, loss[loss=0.2317, simple_loss=0.3045, pruned_loss=0.0794, over 7946.00 frames. ], tot_loss[loss=0.1659, simple_loss=0.2559, pruned_loss=0.03797, over 2221477.53 frames. ], batch size: 98, lr: 3.26e-03, grad_scale: 8.0 2023-05-17 09:44:03,553 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=308552.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 09:44:07,494 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=308558.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 09:44:21,660 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.2525, 5.0766, 5.2021, 5.2424, 4.8410, 4.9244, 4.6599, 5.1855], device='cuda:0'), covar=tensor([0.0847, 0.0615, 0.1013, 0.0686, 0.2089, 0.1419, 0.0619, 0.1028], device='cuda:0'), in_proj_covar=tensor([0.0542, 0.0699, 0.0625, 0.0630, 0.0839, 0.0747, 0.0561, 0.0484], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0005, 0.0004, 0.0003, 0.0003], device='cuda:0') 2023-05-17 09:44:22,999 INFO [finetune.py:992] (0/2) Epoch 18, batch 600, loss[loss=0.1728, simple_loss=0.2613, pruned_loss=0.04215, over 12112.00 frames. ], tot_loss[loss=0.1655, simple_loss=0.2558, pruned_loss=0.03758, over 2256326.34 frames. ], batch size: 38, lr: 3.26e-03, grad_scale: 8.0 2023-05-17 09:44:36,548 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.927e+02 2.569e+02 2.931e+02 3.751e+02 8.067e+02, threshold=5.861e+02, percent-clipped=1.0 2023-05-17 09:44:47,765 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=308613.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 09:44:55,973 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.92 vs. limit=2.0 2023-05-17 09:44:59,650 INFO [finetune.py:992] (0/2) Epoch 18, batch 650, loss[loss=0.1394, simple_loss=0.2251, pruned_loss=0.02683, over 12273.00 frames. ], tot_loss[loss=0.1644, simple_loss=0.2549, pruned_loss=0.03697, over 2284708.18 frames. ], batch size: 28, lr: 3.26e-03, grad_scale: 8.0 2023-05-17 09:45:24,973 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=308665.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 09:45:26,166 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.75 vs. limit=2.0 2023-05-17 09:45:26,833 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.33 vs. limit=2.0 2023-05-17 09:45:27,797 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=308669.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 09:45:32,931 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=192, metric=1.93 vs. limit=2.0 2023-05-17 09:45:35,319 INFO [finetune.py:992] (0/2) Epoch 18, batch 700, loss[loss=0.1569, simple_loss=0.2546, pruned_loss=0.0296, over 12120.00 frames. ], tot_loss[loss=0.164, simple_loss=0.2546, pruned_loss=0.0367, over 2307173.21 frames. ], batch size: 39, lr: 3.26e-03, grad_scale: 8.0 2023-05-17 09:45:48,062 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.889e+02 2.567e+02 3.046e+02 3.784e+02 7.394e+02, threshold=6.092e+02, percent-clipped=4.0 2023-05-17 09:45:57,122 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=4.55 vs. limit=5.0 2023-05-17 09:46:06,396 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=4.81 vs. limit=5.0 2023-05-17 09:46:10,693 INFO [finetune.py:992] (0/2) Epoch 18, batch 750, loss[loss=0.18, simple_loss=0.2698, pruned_loss=0.04508, over 12072.00 frames. ], tot_loss[loss=0.1644, simple_loss=0.2546, pruned_loss=0.03704, over 2325645.58 frames. ], batch size: 42, lr: 3.26e-03, grad_scale: 8.0 2023-05-17 09:46:10,899 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=308730.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 09:46:12,323 INFO [zipformer.py:625] (0/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=308732.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 09:46:46,606 INFO [finetune.py:992] (0/2) Epoch 18, batch 800, loss[loss=0.1583, simple_loss=0.2472, pruned_loss=0.03467, over 12124.00 frames. ], tot_loss[loss=0.1647, simple_loss=0.2547, pruned_loss=0.03731, over 2340727.10 frames. ], batch size: 33, lr: 3.26e-03, grad_scale: 8.0 2023-05-17 09:46:56,291 INFO [zipformer.py:625] (0/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=308793.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 09:46:59,534 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.100e+02 2.666e+02 2.982e+02 3.536e+02 6.354e+02, threshold=5.964e+02, percent-clipped=1.0 2023-05-17 09:47:09,167 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.9367, 4.6079, 4.6698, 4.7832, 4.6682, 4.9470, 4.6580, 2.4012], device='cuda:0'), covar=tensor([0.0131, 0.0101, 0.0129, 0.0080, 0.0066, 0.0114, 0.0140, 0.1029], device='cuda:0'), in_proj_covar=tensor([0.0071, 0.0079, 0.0085, 0.0075, 0.0061, 0.0095, 0.0083, 0.0100], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-17 09:47:23,174 INFO [finetune.py:992] (0/2) Epoch 18, batch 850, loss[loss=0.1779, simple_loss=0.2675, pruned_loss=0.04414, over 12284.00 frames. ], tot_loss[loss=0.1647, simple_loss=0.2546, pruned_loss=0.03736, over 2353993.48 frames. ], batch size: 37, lr: 3.26e-03, grad_scale: 8.0 2023-05-17 09:47:34,150 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.4067, 2.7489, 3.5641, 4.3153, 3.7406, 4.3584, 3.7345, 3.2235], device='cuda:0'), covar=tensor([0.0039, 0.0389, 0.0166, 0.0054, 0.0142, 0.0075, 0.0144, 0.0330], device='cuda:0'), in_proj_covar=tensor([0.0092, 0.0125, 0.0107, 0.0081, 0.0106, 0.0119, 0.0104, 0.0141], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-17 09:47:39,342 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2023-05-17 09:47:43,559 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=308858.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 09:47:58,893 INFO [finetune.py:992] (0/2) Epoch 18, batch 900, loss[loss=0.1436, simple_loss=0.2437, pruned_loss=0.02173, over 12296.00 frames. ], tot_loss[loss=0.1641, simple_loss=0.2544, pruned_loss=0.03689, over 2362542.40 frames. ], batch size: 33, lr: 3.26e-03, grad_scale: 8.0 2023-05-17 09:48:08,626 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([4.2858, 4.6492, 4.0845, 4.9533, 4.4745, 2.7203, 4.2008, 3.1357], device='cuda:0'), covar=tensor([0.0904, 0.0775, 0.1554, 0.0557, 0.1269, 0.2011, 0.1111, 0.3374], device='cuda:0'), in_proj_covar=tensor([0.0314, 0.0384, 0.0364, 0.0336, 0.0376, 0.0280, 0.0352, 0.0374], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-17 09:48:11,877 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.820e+02 2.597e+02 3.126e+02 3.640e+02 8.251e+02, threshold=6.251e+02, percent-clipped=4.0 2023-05-17 09:48:18,277 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=308906.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 09:48:19,679 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=308908.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 09:48:35,350 INFO [finetune.py:992] (0/2) Epoch 18, batch 950, loss[loss=0.1675, simple_loss=0.2654, pruned_loss=0.03481, over 12345.00 frames. ], tot_loss[loss=0.1639, simple_loss=0.2542, pruned_loss=0.03676, over 2370919.59 frames. ], batch size: 36, lr: 3.26e-03, grad_scale: 8.0 2023-05-17 09:49:01,216 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=308965.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 09:49:06,494 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.36 vs. limit=2.0 2023-05-17 09:49:11,620 INFO [finetune.py:992] (0/2) Epoch 18, batch 1000, loss[loss=0.1432, simple_loss=0.2344, pruned_loss=0.02599, over 12114.00 frames. ], tot_loss[loss=0.1646, simple_loss=0.2548, pruned_loss=0.03719, over 2353586.55 frames. ], batch size: 30, lr: 3.26e-03, grad_scale: 8.0 2023-05-17 09:49:24,360 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.688e+02 2.709e+02 3.115e+02 3.712e+02 7.348e+02, threshold=6.229e+02, percent-clipped=2.0 2023-05-17 09:49:35,662 INFO [zipformer.py:625] (0/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=309013.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 09:49:44,123 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=309025.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 09:49:47,612 INFO [finetune.py:992] (0/2) Epoch 18, batch 1050, loss[loss=0.1362, simple_loss=0.2237, pruned_loss=0.02433, over 12134.00 frames. ], tot_loss[loss=0.1637, simple_loss=0.2537, pruned_loss=0.03688, over 2363660.98 frames. ], batch size: 30, lr: 3.26e-03, grad_scale: 8.0 2023-05-17 09:49:59,888 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([3.3131, 2.6417, 3.7711, 3.2557, 3.5908, 3.4545, 2.8051, 3.7544], device='cuda:0'), covar=tensor([0.0133, 0.0423, 0.0169, 0.0264, 0.0178, 0.0170, 0.0418, 0.0135], device='cuda:0'), in_proj_covar=tensor([0.0184, 0.0211, 0.0197, 0.0192, 0.0224, 0.0172, 0.0204, 0.0194], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-17 09:50:23,751 INFO [finetune.py:992] (0/2) Epoch 18, batch 1100, loss[loss=0.1587, simple_loss=0.2606, pruned_loss=0.0284, over 12189.00 frames. ], tot_loss[loss=0.1634, simple_loss=0.253, pruned_loss=0.03688, over 2364081.28 frames. ], batch size: 35, lr: 3.26e-03, grad_scale: 8.0 2023-05-17 09:50:29,563 INFO [zipformer.py:625] (0/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=309088.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 09:50:36,376 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 2.166e+02 2.763e+02 3.161e+02 3.707e+02 8.705e+02, threshold=6.323e+02, percent-clipped=2.0 2023-05-17 09:50:49,051 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.40 vs. limit=2.0 2023-05-17 09:50:59,906 INFO [finetune.py:992] (0/2) Epoch 18, batch 1150, loss[loss=0.1584, simple_loss=0.2537, pruned_loss=0.03152, over 12144.00 frames. ], tot_loss[loss=0.1635, simple_loss=0.2532, pruned_loss=0.03692, over 2372702.28 frames. ], batch size: 36, lr: 3.26e-03, grad_scale: 8.0 2023-05-17 09:51:13,075 INFO [zipformer.py:1454] (0/2) attn_weights_entropy = tensor([5.1011, 4.6929, 5.0896, 4.4579, 4.7603, 4.5975, 5.1401, 4.7759], device='cuda:0'), covar=tensor([0.0314, 0.0445, 0.0327, 0.0297, 0.0454, 0.0337, 0.0205, 0.0399], device='cuda:0'), in_proj_covar=tensor([0.0272, 0.0279, 0.0301, 0.0275, 0.0272, 0.0271, 0.0245, 0.0222], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-17 09:51:35,938 INFO [finetune.py:992] (0/2) Epoch 18, batch 1200, loss[loss=0.1584, simple_loss=0.2446, pruned_loss=0.0361, over 11820.00 frames. ], tot_loss[loss=0.1641, simple_loss=0.2539, pruned_loss=0.03713, over 2370545.77 frames. ], batch size: 26, lr: 3.26e-03, grad_scale: 8.0 2023-05-17 09:51:49,216 INFO [scaling.py:679] (0/2) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-05-17 09:51:49,450 INFO [optim.py:368] (0/2) Clipping_scale=2.0, grad-norm quartiles 1.785e+02 2.620e+02 3.079e+02 3.784e+02 9.052e+02, threshold=6.158e+02, percent-clipped=1.0 2023-05-17 09:51:57,077 INFO [zipformer.py:625] (0/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=309208.0, num_to_drop=0, layers_to_drop=set() 2023-05-17 09:52:12,677 INFO [finetune.py:992] (0/2) Epoch 18, batch 1250, loss[loss=0.171, simple_loss=0.2568, pruned_loss=0.04261, over 12092.00 frames. ], tot_loss[loss=0.164, simple_loss=0.2539, pruned_loss=0.03704, over 2373508.44 frames. ], batch size: 42, lr: 3.26e-03, grad_scale: 8.0 2023-05-17 09:52:14,478 INFO [scaling.py:679] (0/2) Whitening: num_groups=1, num_channels=384, metric=2.27 vs. limit=5.0 2023-05-17 09:52:14,994 INFO [finetune.py:1294] (0/2) Saving batch to pruned_transducer_stateless7/exp_giga_finetune/batch-a689ee27-eec1-83b6-15a8-f48f39643825.pt 2023-05-17 09:52:15,037 INFO [finetune.py:1300] (0/2) features shape: torch.Size([40, 1225, 80]) 2023-05-17 09:52:15,041 INFO [finetune.py:1304] (0/2) num tokens: 2704