2024-08-06 08:06:14,317 INFO [trainer.py:870] (3/8) Training started
2024-08-06 08:06:14,318 INFO [trainer.py:889] (3/8) Device: cuda:3
2024-08-06 08:06:14,318 INFO [trainer.py:890] (3/8) {'best_train_loss': inf, 'best_valid_loss': inf, 'best_train_epoch': -1, 'best_valid_epoch': -1, 'batch_idx_train': 0, 'log_interval': 100, 'reset_interval': 200, 'valid_interval': 2000, 'env_info': {'k2-version': '1.24.3', 'k2-build-type': 'Release', 'k2-with-cuda': True, 'k2-git-sha1': '279b0c87015a615b81b147251814d737a548f397', 'k2-git-date': 'Wed May 24 22:24:09 2023', 'lhotse-version': '1.26.0', 'torch-version': '2.0.1+cu118', 'torch-cuda-available': True, 'torch-cuda-version': '11.8', 'python-version': '3.10', 'icefall-git-branch': None, 'icefall-git-sha1': None, 'icefall-git-date': None, 'icefall-path': '/workspace/icefall_llm', 'k2-path': '/usr/local/lib/python3.10/dist-packages/k2/__init__.py', 'lhotse-path': '/usr/local/lib/python3.10/dist-packages/lhotse/__init__.py', 'hostname': '6867463', 'IP address': '0.104.202.7'}, 'world_size': 8, 'master_port': 12354, 'tensorboard': True, 'num_epochs': 20, 'start_epoch': 1, 'start_batch': 0, 'exp_dir': PosixPath('exp/valle'), 'optimizer_name': 'ScaledAdam', 'scheduler_name': 'Eden', 'base_lr': 0.03, 'warmup_steps': 200, 'seed': 42, 'inf_check': False, 'save_every_n': 20000, 'keep_last_k': 20, 'average_period': 0, 'accumulate_grad_steps': 1, 'dtype': 'bfloat16', 'filter_min_duration': 0.5, 'filter_max_duration': 14.0, 'train_stage': 1, 'visualize': False, 'oom_check': False, 'model_name': 'valle', 'decoder_dim': 1024, 'nhead': 16, 'num_decoder_layers': 12, 'scale_factor': 1.0, 'norm_first': True, 'add_prenet': False, 'prefix_mode': 1, 'share_embedding': True, 'prepend_bos': False, 'num_quantizers': 8, 'scaling_xformers': False, 'manifest_dir': PosixPath('data/tokenized'), 'max_duration': 320, 'bucketing_sampler': True, 'num_buckets': 6, 'concatenate_cuts': False, 'duration_factor': 1.0, 'gap': 0.1, 'on_the_fly_feats': False, 'shuffle': True, 'buffer_size': 40000, 'shuffle_buffer_size': 100000, 'drop_last': False, 'return_cuts': True, 'num_workers': 8, 'enable_spec_aug': False, 'spec_aug_time_warp_factor': 80, 'input_strategy': 'PrecomputedFeatures', 'dataset': 'libritts', 'text_tokens': 'data/tokenized/unique_text_tokens.k2symbols', 'sampling_rate': 24000}
2024-08-06 08:06:14,318 INFO [trainer.py:892] (3/8) About to create model
2024-08-06 08:06:15,086 INFO [trainer.py:899] (3/8) Number of model parameters: 367386628
2024-08-06 08:06:16,728 INFO [trainer.py:914] (3/8) Using DDP
2024-08-06 08:06:19,151 INFO [datamodule.py:427] (3/8) About to get train cuts
2024-08-06 08:06:19,153 INFO [datamodule.py:434] (3/8) About to get dev cuts
2024-08-06 08:06:19,154 INFO [datamodule.py:292] (3/8) Disable SpecAugment
2024-08-06 08:06:19,154 INFO [datamodule.py:294] (3/8) About to create train dataset
2024-08-06 08:06:19,155 INFO [datamodule.py:323] (3/8) Using DynamicBucketingSampler
2024-08-06 08:06:19,758 INFO [datamodule.py:344] (3/8) About to create train dataloader
2024-08-06 08:06:19,758 INFO [datamodule.py:367] (3/8) About to create dev dataset
2024-08-06 08:06:20,081 INFO [datamodule.py:388] (3/8) About to create dev dataloader
2024-08-06 08:08:02,124 INFO [trainer.py:765] (3/8) Epoch 1, batch 100, train_loss[loss=4.321, ArTop10Accuracy=0.494, over 14253.00 frames. ], tot_loss[loss=5.055, ArTop10Accuracy=0.3723, over 4750.99 frames. ], batch size: 62, lr: 2.25e-02
2024-08-06 08:09:28,832 INFO [trainer.py:765] (3/8) Epoch 1, batch 200, train_loss[loss=4.028, ArTop10Accuracy=0.5454, over 13785.00 frames. ], tot_loss[loss=4.489, ArTop10Accuracy=0.4677, over 7752.98 frames. ], batch size: 34, lr: 3.00e-02
2024-08-06 08:10:52,433 INFO [trainer.py:765] (3/8) Epoch 1, batch 300, train_loss[loss=3.866, ArTop10Accuracy=0.5741, over 14454.00 frames. ], tot_loss[loss=4.214, ArTop10Accuracy=0.5139, over 9356.97 frames. ], batch size: 44, lr: 3.00e-02
2024-08-06 08:12:12,702 INFO [trainer.py:765] (3/8) Epoch 1, batch 400, train_loss[loss=3.711, ArTop10Accuracy=0.6045, over 10398.00 frames. ], tot_loss[loss=4.026, ArTop10Accuracy=0.5457, over 10273.01 frames. ], batch size: 14, lr: 3.00e-02
2024-08-06 08:13:40,054 INFO [trainer.py:765] (3/8) Epoch 1, batch 500, train_loss[loss=3.637, ArTop10Accuracy=0.6122, over 12705.00 frames. ], tot_loss[loss=3.879, ArTop10Accuracy=0.5712, over 10836.94 frames. ], batch size: 23, lr: 2.99e-02
2024-08-06 08:15:00,247 INFO [trainer.py:765] (3/8) Epoch 1, batch 600, train_loss[loss=3.602, ArTop10Accuracy=0.6212, over 11367.00 frames. ], tot_loss[loss=3.767, ArTop10Accuracy=0.5912, over 11362.36 frames. ], batch size: 18, lr: 2.99e-02
2024-08-06 08:16:26,429 INFO [trainer.py:765] (3/8) Epoch 1, batch 700, train_loss[loss=3.538, ArTop10Accuracy=0.633, over 10170.00 frames. ], tot_loss[loss=3.684, ArTop10Accuracy=0.6062, over 11494.64 frames. ], batch size: 12, lr: 2.99e-02
2024-08-06 08:17:43,022 INFO [trainer.py:765] (3/8) Epoch 1, batch 800, train_loss[loss=3.434, ArTop10Accuracy=0.6542, over 10245.00 frames. ], tot_loss[loss=3.624, ArTop10Accuracy=0.617, over 11656.86 frames. ], batch size: 12, lr: 2.98e-02
2024-08-06 08:18:56,155 INFO [trainer.py:765] (3/8) Epoch 1, batch 900, train_loss[loss=3.534, ArTop10Accuracy=0.6336, over 12966.00 frames. ], tot_loss[loss=3.566, ArTop10Accuracy=0.6278, over 11705.99 frames. ], batch size: 27, lr: 2.98e-02
2024-08-06 08:20:12,867 INFO [trainer.py:765] (3/8) Epoch 1, batch 1000, train_loss[loss=3.492, ArTop10Accuracy=0.6432, over 12921.00 frames. ], tot_loss[loss=3.524, ArTop10Accuracy=0.6352, over 11908.83 frames. ], batch size: 27, lr: 2.97e-02
2024-08-06 08:20:13,547 INFO [optim.py:386] (3/8) Clipping_scale=2.0, grad-norm quartiles 9.300e+01 1.871e+02 2.675e+02 4.030e+02 9.119e+03, threshold=5.351e+02, percent-clipped=0.0
2024-08-06 08:21:29,161 INFO [trainer.py:765] (3/8) Epoch 1, batch 1100, train_loss[loss=3.45, ArTop10Accuracy=0.6492, over 13404.00 frames. ], tot_loss[loss=3.486, ArTop10Accuracy=0.6422, over 11968.58 frames. ], batch size: 34, lr: 2.96e-02
2024-08-06 08:22:45,417 INFO [trainer.py:765] (3/8) Epoch 1, batch 1200, train_loss[loss=3.485, ArTop10Accuracy=0.6411, over 12429.00 frames. ], tot_loss[loss=3.457, ArTop10Accuracy=0.6477, over 11881.54 frames. ], batch size: 101, lr: 2.96e-02
2024-08-06 08:23:45,270 INFO [trainer.py:650] (3/8) Reaches end of dataloader.
2024-08-06 08:25:36,244 INFO [trainer.py:765] (3/8) Epoch 2, batch 100, train_loss[loss=3.434, ArTop10Accuracy=0.6529, over 14478.00 frames. ], tot_loss[loss=3.425, ArTop10Accuracy=0.6525, over 4759.18 frames. ], batch size: 62, lr: 2.90e-02
2024-08-06 08:26:58,962 INFO [trainer.py:765] (3/8) Epoch 2, batch 200, train_loss[loss=3.295, ArTop10Accuracy=0.6847, over 13569.00 frames. ], tot_loss[loss=3.39, ArTop10Accuracy=0.6593, over 7756.33 frames. ], batch size: 34, lr: 2.89e-02
2024-08-06 08:28:25,540 INFO [trainer.py:765] (3/8) Epoch 2, batch 300, train_loss[loss=3.367, ArTop10Accuracy=0.6621, over 14055.00 frames. ], tot_loss[loss=3.367, ArTop10Accuracy=0.6633, over 9383.94 frames. ], batch size: 44, lr: 2.89e-02
2024-08-06 08:29:48,643 INFO [trainer.py:765] (3/8) Epoch 2, batch 400, train_loss[loss=3.229, ArTop10Accuracy=0.6936, over 10959.00 frames. ], tot_loss[loss=3.356, ArTop10Accuracy=0.6656, over 10302.25 frames. ], batch size: 15, lr: 2.88e-02
2024-08-06 08:31:22,908 INFO [trainer.py:765] (3/8) Epoch 2, batch 500, train_loss[loss=3.37, ArTop10Accuracy=0.6632, over 12750.00 frames. ], tot_loss[loss=3.343, ArTop10Accuracy=0.668, over 10853.14 frames. ], batch size: 23, lr: 2.87e-02
2024-08-06 08:32:45,694 INFO [trainer.py:765] (3/8) Epoch 2, batch 600, train_loss[loss=3.274, ArTop10Accuracy=0.684, over 11472.00 frames. ], tot_loss[loss=3.332, ArTop10Accuracy=0.6701, over 11378.49 frames. ], batch size: 18, lr: 2.86e-02
2024-08-06 08:34:13,589 INFO [trainer.py:765] (3/8) Epoch 2, batch 700, train_loss[loss=3.395, ArTop10Accuracy=0.6511, over 9318.00 frames. ], tot_loss[loss=3.328, ArTop10Accuracy=0.671, over 11516.79 frames. ], batch size: 11, lr: 2.85e-02
2024-08-06 08:34:31,181 INFO [trainer.py:803] (3/8) Computing validation loss
2024-08-06 08:34:40,888 INFO [trainer.py:811] (3/8) Epoch 2, validation: loss=3.277, ArTop10Accuracy=0.6803, over 1827537.00 frames. 
2024-08-06 08:34:40,889 INFO [trainer.py:814] (3/8) Maximum memory allocated so far is 29540MB
2024-08-06 08:34:41,706 INFO [optim.py:386] (3/8) Clipping_scale=2.0, grad-norm quartiles 7.953e+01 1.592e+02 2.200e+02 3.344e+02 2.949e+03, threshold=4.400e+02, percent-clipped=8.6
2024-08-06 08:35:39,883 INFO [trainer.py:765] (3/8) Epoch 2, batch 800, train_loss[loss=3.373, ArTop10Accuracy=0.6659, over 9150.00 frames. ], tot_loss[loss=3.325, ArTop10Accuracy=0.6717, over 11642.86 frames. ], batch size: 11, lr: 2.84e-02
2024-08-06 08:36:56,377 INFO [trainer.py:765] (3/8) Epoch 2, batch 900, train_loss[loss=3.308, ArTop10Accuracy=0.6785, over 12996.00 frames. ], tot_loss[loss=3.31, ArTop10Accuracy=0.6747, over 11701.06 frames. ], batch size: 27, lr: 2.83e-02
2024-08-06 08:38:10,517 INFO [trainer.py:765] (3/8) Epoch 2, batch 1000, train_loss[loss=3.3, ArTop10Accuracy=0.6792, over 12924.00 frames. ], tot_loss[loss=3.303, ArTop10Accuracy=0.6759, over 11883.62 frames. ], batch size: 27, lr: 2.82e-02
2024-08-06 08:39:25,066 INFO [trainer.py:765] (3/8) Epoch 2, batch 1100, train_loss[loss=3.255, ArTop10Accuracy=0.6854, over 13749.00 frames. ], tot_loss[loss=3.294, ArTop10Accuracy=0.6775, over 11961.15 frames. ], batch size: 34, lr: 2.81e-02
2024-08-06 08:40:38,226 INFO [trainer.py:765] (3/8) Epoch 2, batch 1200, train_loss[loss=3.319, ArTop10Accuracy=0.6732, over 12072.00 frames. ], tot_loss[loss=3.284, ArTop10Accuracy=0.6796, over 11891.67 frames. ], batch size: 101, lr: 2.80e-02
2024-08-06 08:41:38,664 INFO [trainer.py:650] (3/8) Reaches end of dataloader.
2024-08-06 08:43:36,656 INFO [trainer.py:765] (3/8) Epoch 3, batch 100, train_loss[loss=3.244, ArTop10Accuracy=0.6856, over 14499.00 frames. ], tot_loss[loss=3.254, ArTop10Accuracy=0.6846, over 4767.19 frames. ], batch size: 62, lr: 2.67e-02
2024-08-06 08:45:10,506 INFO [trainer.py:765] (3/8) Epoch 3, batch 200, train_loss[loss=3.139, ArTop10Accuracy=0.7055, over 13473.00 frames. ], tot_loss[loss=3.23, ArTop10Accuracy=0.6888, over 7744.68 frames. ], batch size: 34, lr: 2.66e-02
2024-08-06 08:46:29,264 INFO [trainer.py:765] (3/8) Epoch 3, batch 300, train_loss[loss=3.263, ArTop10Accuracy=0.6859, over 14445.00 frames. ], tot_loss[loss=3.209, ArTop10Accuracy=0.6931, over 9375.03 frames. ], batch size: 45, lr: 2.64e-02
2024-08-06 08:48:04,225 INFO [trainer.py:765] (3/8) Epoch 3, batch 400, train_loss[loss=3.121, ArTop10Accuracy=0.7107, over 10851.00 frames. ], tot_loss[loss=3.191, ArTop10Accuracy=0.6967, over 10285.27 frames. ], batch size: 15, lr: 2.63e-02
2024-08-06 08:48:40,887 INFO [optim.py:386] (3/8) Clipping_scale=2.0, grad-norm quartiles 9.282e+01 1.561e+02 1.981e+02 2.686e+02 1.768e+03, threshold=3.962e+02, percent-clipped=7.6
2024-08-06 08:49:25,547 INFO [trainer.py:765] (3/8) Epoch 3, batch 500, train_loss[loss=3.155, ArTop10Accuracy=0.709, over 12711.00 frames. ], tot_loss[loss=3.174, ArTop10Accuracy=0.7001, over 10844.23 frames. ], batch size: 23, lr: 2.62e-02
2024-08-06 08:51:00,483 INFO [trainer.py:765] (3/8) Epoch 3, batch 600, train_loss[loss=3.06, ArTop10Accuracy=0.725, over 11397.00 frames. ], tot_loss[loss=3.157, ArTop10Accuracy=0.7033, over 11347.80 frames. ], batch size: 18, lr: 2.61e-02
2024-08-06 08:52:31,624 INFO [trainer.py:765] (3/8) Epoch 3, batch 700, train_loss[loss=3.042, ArTop10Accuracy=0.7264, over 10038.00 frames. ], tot_loss[loss=3.145, ArTop10Accuracy=0.7056, over 11508.63 frames. ], batch size: 12, lr: 2.60e-02
2024-08-06 08:53:57,394 INFO [trainer.py:765] (3/8) Epoch 3, batch 800, train_loss[loss=3.221, ArTop10Accuracy=0.6856, over 10071.00 frames. ], tot_loss[loss=3.138, ArTop10Accuracy=0.7071, over 11643.06 frames. ], batch size: 12, lr: 2.59e-02
2024-08-06 08:55:15,124 INFO [trainer.py:765] (3/8) Epoch 3, batch 900, train_loss[loss=3.071, ArTop10Accuracy=0.7171, over 12846.00 frames. ], tot_loss[loss=3.117, ArTop10Accuracy=0.711, over 11684.97 frames. ], batch size: 27, lr: 2.57e-02
2024-08-06 08:56:31,563 INFO [trainer.py:765] (3/8) Epoch 3, batch 1000, train_loss[loss=3.188, ArTop10Accuracy=0.6936, over 12897.00 frames. ], tot_loss[loss=3.112, ArTop10Accuracy=0.7118, over 11865.36 frames. ], batch size: 27, lr: 2.56e-02
2024-08-06 08:57:46,512 INFO [trainer.py:765] (3/8) Epoch 3, batch 1100, train_loss[loss=3.058, ArTop10Accuracy=0.7222, over 13737.00 frames. ], tot_loss[loss=3.105, ArTop10Accuracy=0.7131, over 11941.48 frames. ], batch size: 34, lr: 2.55e-02
2024-08-06 08:59:01,405 INFO [trainer.py:765] (3/8) Epoch 3, batch 1200, train_loss[loss=3.175, ArTop10Accuracy=0.7042, over 12597.00 frames. ], tot_loss[loss=3.095, ArTop10Accuracy=0.7152, over 11835.39 frames. ], batch size: 101, lr: 2.54e-02
2024-08-06 09:00:02,031 INFO [trainer.py:650] (3/8) Reaches end of dataloader.
2024-08-06 09:01:50,747 INFO [trainer.py:765] (3/8) Epoch 4, batch 100, train_loss[loss=3.104, ArTop10Accuracy=0.7088, over 14130.00 frames. ], tot_loss[loss=3.069, ArTop10Accuracy=0.7194, over 4744.71 frames. ], batch size: 62, lr: 2.38e-02
2024-08-06 09:02:52,864 INFO [trainer.py:803] (3/8) Computing validation loss
2024-08-06 09:03:02,384 INFO [trainer.py:811] (3/8) Epoch 4, validation: loss=2.997, ArTop10Accuracy=0.7338, over 1827537.00 frames. 
2024-08-06 09:03:02,385 INFO [trainer.py:814] (3/8) Maximum memory allocated so far is 29540MB
2024-08-06 09:03:03,370 INFO [optim.py:386] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.072e+02 1.499e+02 1.782e+02 2.273e+02 1.100e+03, threshold=3.565e+02, percent-clipped=4.7
2024-08-06 09:03:29,279 INFO [trainer.py:765] (3/8) Epoch 4, batch 200, train_loss[loss=3.02, ArTop10Accuracy=0.7308, over 13695.00 frames. ], tot_loss[loss=3.044, ArTop10Accuracy=0.7244, over 7751.43 frames. ], batch size: 34, lr: 2.37e-02
2024-08-06 09:05:01,738 INFO [trainer.py:765] (3/8) Epoch 4, batch 300, train_loss[loss=3.065, ArTop10Accuracy=0.7227, over 14499.00 frames. ], tot_loss[loss=3.04, ArTop10Accuracy=0.7256, over 9375.57 frames. ], batch size: 45, lr: 2.36e-02
2024-08-06 09:06:28,156 INFO [trainer.py:765] (3/8) Epoch 4, batch 400, train_loss[loss=2.988, ArTop10Accuracy=0.7383, over 10809.00 frames. ], tot_loss[loss=3.03, ArTop10Accuracy=0.7275, over 10281.95 frames. ], batch size: 15, lr: 2.34e-02
2024-08-06 09:08:01,930 INFO [trainer.py:765] (3/8) Epoch 4, batch 500, train_loss[loss=2.973, ArTop10Accuracy=0.7446, over 12387.00 frames. ], tot_loss[loss=3.022, ArTop10Accuracy=0.729, over 10821.67 frames. ], batch size: 22, lr: 2.33e-02
2024-08-06 09:09:28,547 INFO [trainer.py:765] (3/8) Epoch 4, batch 600, train_loss[loss=2.865, ArTop10Accuracy=0.7627, over 11403.00 frames. ], tot_loss[loss=3.02, ArTop10Accuracy=0.7295, over 11351.68 frames. ], batch size: 18, lr: 2.32e-02
2024-08-06 09:10:59,872 INFO [trainer.py:765] (3/8) Epoch 4, batch 700, train_loss[loss=2.839, ArTop10Accuracy=0.765, over 9333.00 frames. ], tot_loss[loss=3.022, ArTop10Accuracy=0.7291, over 11490.84 frames. ], batch size: 11, lr: 2.31e-02
2024-08-06 09:12:17,519 INFO [trainer.py:765] (3/8) Epoch 4, batch 800, train_loss[loss=2.979, ArTop10Accuracy=0.7389, over 9318.00 frames. ], tot_loss[loss=3.021, ArTop10Accuracy=0.7293, over 11615.33 frames. ], batch size: 11, lr: 2.30e-02
2024-08-06 09:13:33,219 INFO [trainer.py:765] (3/8) Epoch 4, batch 900, train_loss[loss=2.958, ArTop10Accuracy=0.7471, over 12633.00 frames. ], tot_loss[loss=3.011, ArTop10Accuracy=0.7311, over 11664.01 frames. ], batch size: 27, lr: 2.29e-02
2024-08-06 09:14:47,526 INFO [trainer.py:765] (3/8) Epoch 4, batch 1000, train_loss[loss=2.934, ArTop10Accuracy=0.7402, over 12672.00 frames. ], tot_loss[loss=3.013, ArTop10Accuracy=0.7307, over 11855.93 frames. ], batch size: 27, lr: 2.28e-02
2024-08-06 09:16:02,988 INFO [trainer.py:765] (3/8) Epoch 4, batch 1100, train_loss[loss=3.039, ArTop10Accuracy=0.7209, over 13689.00 frames. ], tot_loss[loss=3.015, ArTop10Accuracy=0.7302, over 11936.22 frames. ], batch size: 34, lr: 2.26e-02
2024-08-06 09:16:53,297 INFO [optim.py:386] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.100e+02 1.440e+02 1.636e+02 1.968e+02 7.702e+02, threshold=3.273e+02, percent-clipped=1.3
2024-08-06 09:17:18,350 INFO [trainer.py:765] (3/8) Epoch 4, batch 1200, train_loss[loss=3.08, ArTop10Accuracy=0.7212, over 13128.00 frames. ], tot_loss[loss=3.012, ArTop10Accuracy=0.7306, over 11871.12 frames. ], batch size: 103, lr: 2.25e-02
2024-08-06 09:18:17,420 INFO [trainer.py:650] (3/8) Reaches end of dataloader.
2024-08-06 09:20:17,177 INFO [trainer.py:765] (3/8) Epoch 5, batch 100, train_loss[loss=3.008, ArTop10Accuracy=0.7247, over 14334.00 frames. ], tot_loss[loss=2.988, ArTop10Accuracy=0.7348, over 4769.18 frames. ], batch size: 63, lr: 2.10e-02
2024-08-06 09:21:52,302 INFO [trainer.py:765] (3/8) Epoch 5, batch 200, train_loss[loss=3.035, ArTop10Accuracy=0.7238, over 13557.00 frames. ], tot_loss[loss=2.981, ArTop10Accuracy=0.7363, over 7749.95 frames. ], batch size: 34, lr: 2.09e-02
2024-08-06 09:23:19,247 INFO [trainer.py:765] (3/8) Epoch 5, batch 300, train_loss[loss=3.01, ArTop10Accuracy=0.7287, over 14130.00 frames. ], tot_loss[loss=2.969, ArTop10Accuracy=0.7386, over 9362.27 frames. ], batch size: 44, lr: 2.08e-02
2024-08-06 09:24:53,543 INFO [trainer.py:765] (3/8) Epoch 5, batch 400, train_loss[loss=2.831, ArTop10Accuracy=0.7688, over 10419.00 frames. ], tot_loss[loss=2.967, ArTop10Accuracy=0.7389, over 10263.43 frames. ], batch size: 14, lr: 2.07e-02
2024-08-06 09:26:19,424 INFO [trainer.py:765] (3/8) Epoch 5, batch 500, train_loss[loss=2.948, ArTop10Accuracy=0.7401, over 12189.00 frames. ], tot_loss[loss=2.964, ArTop10Accuracy=0.7394, over 10831.03 frames. ], batch size: 22, lr: 2.06e-02
2024-08-06 09:27:49,543 INFO [trainer.py:765] (3/8) Epoch 5, batch 600, train_loss[loss=3.025, ArTop10Accuracy=0.7193, over 11202.00 frames. ], tot_loss[loss=2.963, ArTop10Accuracy=0.7395, over 11339.96 frames. ], batch size: 18, lr: 2.05e-02
2024-08-06 09:29:21,676 INFO [trainer.py:765] (3/8) Epoch 5, batch 700, train_loss[loss=2.829, ArTop10Accuracy=0.7669, over 9447.00 frames. ], tot_loss[loss=2.965, ArTop10Accuracy=0.7396, over 11490.41 frames. ], batch size: 11, lr: 2.04e-02
2024-08-06 09:30:44,699 INFO [trainer.py:765] (3/8) Epoch 5, batch 800, train_loss[loss=2.886, ArTop10Accuracy=0.7585, over 9441.00 frames. ], tot_loss[loss=2.969, ArTop10Accuracy=0.7388, over 11618.93 frames. ], batch size: 11, lr: 2.03e-02
2024-08-06 09:31:51,245 INFO [trainer.py:803] (3/8) Computing validation loss
2024-08-06 09:32:00,762 INFO [trainer.py:811] (3/8) Epoch 5, validation: loss=2.926, ArTop10Accuracy=0.7466, over 1827537.00 frames. 
2024-08-06 09:32:00,763 INFO [trainer.py:814] (3/8) Maximum memory allocated so far is 29540MB
2024-08-06 09:32:01,714 INFO [optim.py:386] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.060e+02 1.349e+02 1.525e+02 1.806e+02 1.007e+03, threshold=3.049e+02, percent-clipped=2.3
2024-08-06 09:32:10,560 INFO [trainer.py:765] (3/8) Epoch 5, batch 900, train_loss[loss=2.864, ArTop10Accuracy=0.758, over 12900.00 frames. ], tot_loss[loss=2.958, ArTop10Accuracy=0.741, over 11669.20 frames. ], batch size: 27, lr: 2.02e-02
2024-08-06 09:33:27,329 INFO [trainer.py:765] (3/8) Epoch 5, batch 1000, train_loss[loss=2.95, ArTop10Accuracy=0.7414, over 12960.00 frames. ], tot_loss[loss=2.961, ArTop10Accuracy=0.7401, over 11887.77 frames. ], batch size: 27, lr: 2.01e-02
2024-08-06 09:34:42,307 INFO [trainer.py:765] (3/8) Epoch 5, batch 1100, train_loss[loss=2.925, ArTop10Accuracy=0.7515, over 13506.00 frames. ], tot_loss[loss=2.967, ArTop10Accuracy=0.7391, over 11948.88 frames. ], batch size: 34, lr: 2.00e-02
2024-08-06 09:35:56,338 INFO [trainer.py:765] (3/8) Epoch 5, batch 1200, train_loss[loss=2.999, ArTop10Accuracy=0.7334, over 12924.00 frames. ], tot_loss[loss=2.967, ArTop10Accuracy=0.7391, over 11862.59 frames. ], batch size: 101, lr: 1.99e-02
2024-08-06 09:36:55,670 INFO [trainer.py:650] (3/8) Reaches end of dataloader.
2024-08-06 09:38:52,670 INFO [trainer.py:765] (3/8) Epoch 6, batch 100, train_loss[loss=3, ArTop10Accuracy=0.7349, over 14295.00 frames. ], tot_loss[loss=2.947, ArTop10Accuracy=0.7423, over 4739.69 frames. ], batch size: 62, lr: 1.85e-02
2024-08-06 09:40:19,840 INFO [trainer.py:765] (3/8) Epoch 6, batch 200, train_loss[loss=2.892, ArTop10Accuracy=0.7548, over 13599.00 frames. ], tot_loss[loss=2.937, ArTop10Accuracy=0.7444, over 7746.65 frames. ], batch size: 34, lr: 1.84e-02
2024-08-06 09:41:52,970 INFO [trainer.py:765] (3/8) Epoch 6, batch 300, train_loss[loss=3.024, ArTop10Accuracy=0.7262, over 14010.00 frames. ], tot_loss[loss=2.93, ArTop10Accuracy=0.746, over 9367.16 frames. ], batch size: 44, lr: 1.83e-02
2024-08-06 09:43:17,833 INFO [trainer.py:765] (3/8) Epoch 6, batch 400, train_loss[loss=2.828, ArTop10Accuracy=0.7645, over 10341.00 frames. ], tot_loss[loss=2.926, ArTop10Accuracy=0.7468, over 10282.94 frames. ], batch size: 14, lr: 1.83e-02
2024-08-06 09:44:54,133 INFO [trainer.py:765] (3/8) Epoch 6, batch 500, train_loss[loss=2.917, ArTop10Accuracy=0.7487, over 12615.00 frames. ], tot_loss[loss=2.92, ArTop10Accuracy=0.7477, over 10844.16 frames. ], batch size: 23, lr: 1.82e-02
2024-08-06 09:46:22,878 INFO [trainer.py:765] (3/8) Epoch 6, batch 600, train_loss[loss=2.954, ArTop10Accuracy=0.7389, over 11496.00 frames. ], tot_loss[loss=2.92, ArTop10Accuracy=0.7478, over 11382.94 frames. ], batch size: 18, lr: 1.81e-02
2024-08-06 09:46:37,225 INFO [optim.py:386] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.012e+02 1.339e+02 1.480e+02 1.701e+02 7.506e+02, threshold=2.959e+02, percent-clipped=1.1
2024-08-06 09:47:57,876 INFO [trainer.py:765] (3/8) Epoch 6, batch 700, train_loss[loss=2.81, ArTop10Accuracy=0.7641, over 9426.00 frames. ], tot_loss[loss=2.927, ArTop10Accuracy=0.7465, over 11545.50 frames. ], batch size: 11, lr: 1.80e-02
2024-08-06 09:49:15,960 INFO [trainer.py:765] (3/8) Epoch 6, batch 800, train_loss[loss=2.978, ArTop10Accuracy=0.738, over 10194.00 frames. ], tot_loss[loss=2.929, ArTop10Accuracy=0.7462, over 11653.83 frames. ], batch size: 12, lr: 1.79e-02
2024-08-06 09:50:32,141 INFO [trainer.py:765] (3/8) Epoch 6, batch 900, train_loss[loss=2.958, ArTop10Accuracy=0.7424, over 12861.00 frames. ], tot_loss[loss=2.923, ArTop10Accuracy=0.7476, over 11703.21 frames. ], batch size: 27, lr: 1.78e-02
2024-08-06 09:51:47,305 INFO [trainer.py:765] (3/8) Epoch 6, batch 1000, train_loss[loss=2.948, ArTop10Accuracy=0.7387, over 12723.00 frames. ], tot_loss[loss=2.922, ArTop10Accuracy=0.7476, over 11897.90 frames. ], batch size: 27, lr: 1.77e-02
2024-08-06 09:53:00,926 INFO [trainer.py:765] (3/8) Epoch 6, batch 1100, train_loss[loss=2.986, ArTop10Accuracy=0.7328, over 13701.00 frames. ], tot_loss[loss=2.928, ArTop10Accuracy=0.7464, over 11964.42 frames. ], batch size: 34, lr: 1.77e-02
2024-08-06 09:54:14,342 INFO [trainer.py:765] (3/8) Epoch 6, batch 1200, train_loss[loss=3.044, ArTop10Accuracy=0.7283, over 12891.00 frames. ], tot_loss[loss=2.927, ArTop10Accuracy=0.7465, over 11877.83 frames. ], batch size: 101, lr: 1.76e-02
2024-08-06 09:55:13,384 INFO [trainer.py:650] (3/8) Reaches end of dataloader.
2024-08-06 09:57:06,705 INFO [trainer.py:765] (3/8) Epoch 7, batch 100, train_loss[loss=2.902, ArTop10Accuracy=0.7499, over 14643.00 frames. ], tot_loss[loss=2.913, ArTop10Accuracy=0.7485, over 4780.37 frames. ], batch size: 63, lr: 1.64e-02
2024-08-06 09:58:39,432 INFO [trainer.py:765] (3/8) Epoch 7, batch 200, train_loss[loss=2.904, ArTop10Accuracy=0.7525, over 13992.00 frames. ], tot_loss[loss=2.905, ArTop10Accuracy=0.7503, over 7757.95 frames. ], batch size: 35, lr: 1.64e-02
2024-08-06 10:00:06,089 INFO [trainer.py:765] (3/8) Epoch 7, batch 300, train_loss[loss=2.937, ArTop10Accuracy=0.7427, over 14214.00 frames. ], tot_loss[loss=2.901, ArTop10Accuracy=0.7509, over 9379.79 frames. ], batch size: 45, lr: 1.63e-02
2024-08-06 10:00:40,516 INFO [trainer.py:803] (3/8) Computing validation loss
2024-08-06 10:00:50,245 INFO [trainer.py:811] (3/8) Epoch 7, validation: loss=2.88, ArTop10Accuracy=0.7554, over 1827537.00 frames. 
2024-08-06 10:00:50,246 INFO [trainer.py:814] (3/8) Maximum memory allocated so far is 30046MB
2024-08-06 10:00:50,983 INFO [optim.py:386] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.002e+02 1.286e+02 1.429e+02 1.605e+02 1.020e+03, threshold=2.857e+02, percent-clipped=1.5
2024-08-06 10:01:49,122 INFO [trainer.py:765] (3/8) Epoch 7, batch 400, train_loss[loss=2.901, ArTop10Accuracy=0.7524, over 10989.00 frames. ], tot_loss[loss=2.895, ArTop10Accuracy=0.7526, over 10298.60 frames. ], batch size: 15, lr: 1.62e-02
2024-08-06 10:03:21,465 INFO [trainer.py:765] (3/8) Epoch 7, batch 500, train_loss[loss=2.942, ArTop10Accuracy=0.7404, over 12828.00 frames. ], tot_loss[loss=2.891, ArTop10Accuracy=0.7535, over 10868.22 frames. ], batch size: 23, lr: 1.61e-02
2024-08-06 10:04:51,889 INFO [trainer.py:765] (3/8) Epoch 7, batch 600, train_loss[loss=2.859, ArTop10Accuracy=0.7626, over 11541.00 frames. ], tot_loss[loss=2.894, ArTop10Accuracy=0.7528, over 11398.95 frames. ], batch size: 18, lr: 1.61e-02
2024-08-06 10:06:25,118 INFO [trainer.py:765] (3/8) Epoch 7, batch 700, train_loss[loss=2.801, ArTop10Accuracy=0.7696, over 10050.00 frames. ], tot_loss[loss=2.897, ArTop10Accuracy=0.7524, over 11508.96 frames. ], batch size: 12, lr: 1.60e-02
2024-08-06 10:07:46,954 INFO [trainer.py:765] (3/8) Epoch 7, batch 800, train_loss[loss=2.846, ArTop10Accuracy=0.7637, over 10302.00 frames. ], tot_loss[loss=2.897, ArTop10Accuracy=0.7525, over 11633.29 frames. ], batch size: 12, lr: 1.59e-02
2024-08-06 10:09:02,830 INFO [trainer.py:765] (3/8) Epoch 7, batch 900, train_loss[loss=2.874, ArTop10Accuracy=0.7574, over 12936.00 frames. ], tot_loss[loss=2.888, ArTop10Accuracy=0.7541, over 11680.46 frames. ], batch size: 27, lr: 1.59e-02
2024-08-06 10:10:19,642 INFO [trainer.py:765] (3/8) Epoch 7, batch 1000, train_loss[loss=2.844, ArTop10Accuracy=0.7628, over 12774.00 frames. ], tot_loss[loss=2.897, ArTop10Accuracy=0.7524, over 11883.71 frames. ], batch size: 27, lr: 1.58e-02
2024-08-06 10:11:35,215 INFO [trainer.py:765] (3/8) Epoch 7, batch 1100, train_loss[loss=2.865, ArTop10Accuracy=0.7577, over 13803.00 frames. ], tot_loss[loss=2.9, ArTop10Accuracy=0.7517, over 11945.40 frames. ], batch size: 34, lr: 1.57e-02
2024-08-06 10:12:48,210 INFO [trainer.py:765] (3/8) Epoch 7, batch 1200, train_loss[loss=2.986, ArTop10Accuracy=0.7322, over 12135.00 frames. ], tot_loss[loss=2.897, ArTop10Accuracy=0.7522, over 11873.39 frames. ], batch size: 101, lr: 1.57e-02
2024-08-06 10:13:46,761 INFO [trainer.py:650] (3/8) Reaches end of dataloader.
2024-08-06 10:15:03,607 INFO [optim.py:386] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.017e+02 1.283e+02 1.410e+02 1.601e+02 1.017e+03, threshold=2.820e+02, percent-clipped=0.9
2024-08-06 10:15:40,826 INFO [trainer.py:765] (3/8) Epoch 8, batch 100, train_loss[loss=2.98, ArTop10Accuracy=0.7361, over 14430.00 frames. ], tot_loss[loss=2.887, ArTop10Accuracy=0.7539, over 4774.76 frames. ], batch size: 62, lr: 1.47e-02
2024-08-06 10:17:12,867 INFO [trainer.py:765] (3/8) Epoch 8, batch 200, train_loss[loss=2.818, ArTop10Accuracy=0.7666, over 13695.00 frames. ], tot_loss[loss=2.876, ArTop10Accuracy=0.7558, over 7759.07 frames. ], batch size: 34, lr: 1.46e-02
2024-08-06 10:18:37,904 INFO [trainer.py:765] (3/8) Epoch 8, batch 300, train_loss[loss=2.93, ArTop10Accuracy=0.7451, over 14208.00 frames. ], tot_loss[loss=2.868, ArTop10Accuracy=0.7573, over 9366.52 frames. ], batch size: 44, lr: 1.46e-02
2024-08-06 10:20:06,350 INFO [trainer.py:765] (3/8) Epoch 8, batch 400, train_loss[loss=2.762, ArTop10Accuracy=0.7819, over 10890.00 frames. ], tot_loss[loss=2.865, ArTop10Accuracy=0.7582, over 10274.78 frames. ], batch size: 15, lr: 1.45e-02
2024-08-06 10:21:32,417 INFO [trainer.py:765] (3/8) Epoch 8, batch 500, train_loss[loss=2.789, ArTop10Accuracy=0.7733, over 12693.00 frames. ], tot_loss[loss=2.859, ArTop10Accuracy=0.7593, over 10844.23 frames. ], batch size: 23, lr: 1.45e-02
2024-08-06 10:23:00,980 INFO [trainer.py:765] (3/8) Epoch 8, batch 600, train_loss[loss=2.75, ArTop10Accuracy=0.779, over 11301.00 frames. ], tot_loss[loss=2.864, ArTop10Accuracy=0.7585, over 11361.98 frames. ], batch size: 18, lr: 1.44e-02
2024-08-06 10:24:37,793 INFO [trainer.py:765] (3/8) Epoch 8, batch 700, train_loss[loss=2.707, ArTop10Accuracy=0.7901, over 10263.00 frames. ], tot_loss[loss=2.87, ArTop10Accuracy=0.7573, over 11497.41 frames. ], batch size: 12, lr: 1.43e-02
2024-08-06 10:25:56,091 INFO [trainer.py:765] (3/8) Epoch 8, batch 800, train_loss[loss=2.766, ArTop10Accuracy=0.7851, over 9222.00 frames. ], tot_loss[loss=2.875, ArTop10Accuracy=0.7566, over 11602.97 frames. ], batch size: 11, lr: 1.43e-02
2024-08-06 10:27:12,251 INFO [trainer.py:765] (3/8) Epoch 8, batch 900, train_loss[loss=2.853, ArTop10Accuracy=0.7619, over 12891.00 frames. ], tot_loss[loss=2.865, ArTop10Accuracy=0.7586, over 11666.59 frames. ], batch size: 27, lr: 1.42e-02
2024-08-06 10:28:25,269 INFO [trainer.py:765] (3/8) Epoch 8, batch 1000, train_loss[loss=2.875, ArTop10Accuracy=0.7523, over 12798.00 frames. ], tot_loss[loss=2.872, ArTop10Accuracy=0.7572, over 11869.24 frames. ], batch size: 27, lr: 1.42e-02
2024-08-06 10:29:07,161 INFO [trainer.py:803] (3/8) Computing validation loss
2024-08-06 10:29:16,830 INFO [trainer.py:811] (3/8) Epoch 8, validation: loss=2.858, ArTop10Accuracy=0.7594, over 1827537.00 frames. 
2024-08-06 10:29:16,831 INFO [trainer.py:814] (3/8) Maximum memory allocated so far is 33011MB
2024-08-06 10:29:17,497 INFO [optim.py:386] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.032e+02 1.275e+02 1.390e+02 1.547e+02 3.717e+02, threshold=2.781e+02, percent-clipped=0.7
2024-08-06 10:29:51,737 INFO [trainer.py:765] (3/8) Epoch 8, batch 1100, train_loss[loss=2.957, ArTop10Accuracy=0.737, over 13581.00 frames. ], tot_loss[loss=2.88, ArTop10Accuracy=0.7556, over 11946.69 frames. ], batch size: 34, lr: 1.41e-02
2024-08-06 10:31:05,952 INFO [trainer.py:765] (3/8) Epoch 8, batch 1200, train_loss[loss=2.957, ArTop10Accuracy=0.7357, over 12132.00 frames. ], tot_loss[loss=2.877, ArTop10Accuracy=0.7559, over 11833.29 frames. ], batch size: 101, lr: 1.40e-02
2024-08-06 10:32:05,668 INFO [trainer.py:650] (3/8) Reaches end of dataloader.
2024-08-06 10:34:01,262 INFO [trainer.py:765] (3/8) Epoch 9, batch 100, train_loss[loss=2.938, ArTop10Accuracy=0.7428, over 14349.00 frames. ], tot_loss[loss=2.857, ArTop10Accuracy=0.7592, over 4768.86 frames. ], batch size: 62, lr: 1.32e-02
2024-08-06 10:35:31,778 INFO [trainer.py:765] (3/8) Epoch 9, batch 200, train_loss[loss=2.876, ArTop10Accuracy=0.7568, over 13407.00 frames. ], tot_loss[loss=2.854, ArTop10Accuracy=0.7602, over 7749.90 frames. ], batch size: 34, lr: 1.32e-02
2024-08-06 10:36:57,933 INFO [trainer.py:765] (3/8) Epoch 9, batch 300, train_loss[loss=2.883, ArTop10Accuracy=0.7558, over 14283.00 frames. ], tot_loss[loss=2.851, ArTop10Accuracy=0.7606, over 9375.68 frames. ], batch size: 44, lr: 1.31e-02
2024-08-06 10:38:32,703 INFO [trainer.py:765] (3/8) Epoch 9, batch 400, train_loss[loss=2.786, ArTop10Accuracy=0.7768, over 10629.00 frames. ], tot_loss[loss=2.848, ArTop10Accuracy=0.7612, over 10299.19 frames. ], batch size: 14, lr: 1.31e-02
2024-08-06 10:39:59,262 INFO [trainer.py:765] (3/8) Epoch 9, batch 500, train_loss[loss=2.803, ArTop10Accuracy=0.7703, over 12306.00 frames. ], tot_loss[loss=2.844, ArTop10Accuracy=0.7624, over 10831.99 frames. ], batch size: 22, lr: 1.30e-02
2024-08-06 10:41:29,696 INFO [trainer.py:765] (3/8) Epoch 9, batch 600, train_loss[loss=2.849, ArTop10Accuracy=0.7581, over 11571.00 frames. ], tot_loss[loss=2.845, ArTop10Accuracy=0.7621, over 11341.92 frames. ], batch size: 18, lr: 1.30e-02
2024-08-06 10:42:58,446 INFO [trainer.py:765] (3/8) Epoch 9, batch 700, train_loss[loss=2.93, ArTop10Accuracy=0.7417, over 9555.00 frames. ], tot_loss[loss=2.847, ArTop10Accuracy=0.7619, over 11496.78 frames. ], batch size: 11, lr: 1.29e-02
2024-08-06 10:44:02,958 INFO [optim.py:386] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.039e+02 1.253e+02 1.352e+02 1.493e+02 7.010e+02, threshold=2.704e+02, percent-clipped=0.6
2024-08-06 10:44:19,675 INFO [trainer.py:765] (3/8) Epoch 9, batch 800, train_loss[loss=2.76, ArTop10Accuracy=0.7781, over 10053.00 frames. ], tot_loss[loss=2.853, ArTop10Accuracy=0.7607, over 11611.66 frames. ], batch size: 12, lr: 1.29e-02
2024-08-06 10:45:35,725 INFO [trainer.py:765] (3/8) Epoch 9, batch 900, train_loss[loss=2.775, ArTop10Accuracy=0.7765, over 13281.00 frames. ], tot_loss[loss=2.845, ArTop10Accuracy=0.7624, over 11670.68 frames. ], batch size: 28, lr: 1.28e-02
2024-08-06 10:46:51,277 INFO [trainer.py:765] (3/8) Epoch 9, batch 1000, train_loss[loss=2.849, ArTop10Accuracy=0.7631, over 12930.00 frames. ], tot_loss[loss=2.847, ArTop10Accuracy=0.7619, over 11874.30 frames. ], batch size: 27, lr: 1.28e-02
2024-08-06 10:48:06,252 INFO [trainer.py:765] (3/8) Epoch 9, batch 1100, train_loss[loss=2.854, ArTop10Accuracy=0.7604, over 13707.00 frames. ], tot_loss[loss=2.854, ArTop10Accuracy=0.7606, over 11943.12 frames. ], batch size: 34, lr: 1.28e-02
2024-08-06 10:49:21,059 INFO [trainer.py:765] (3/8) Epoch 9, batch 1200, train_loss[loss=2.971, ArTop10Accuracy=0.7419, over 12234.00 frames. ], tot_loss[loss=2.856, ArTop10Accuracy=0.76, over 11871.25 frames. ], batch size: 101, lr: 1.27e-02
2024-08-06 10:50:21,935 INFO [trainer.py:650] (3/8) Reaches end of dataloader.
2024-08-06 10:52:12,332 INFO [trainer.py:765] (3/8) Epoch 10, batch 100, train_loss[loss=2.91, ArTop10Accuracy=0.7482, over 14496.00 frames. ], tot_loss[loss=2.845, ArTop10Accuracy=0.7617, over 4757.99 frames. ], batch size: 62, lr: 1.20e-02
2024-08-06 10:53:44,592 INFO [trainer.py:765] (3/8) Epoch 10, batch 200, train_loss[loss=2.72, ArTop10Accuracy=0.79, over 13527.00 frames. ], tot_loss[loss=2.831, ArTop10Accuracy=0.7645, over 7751.44 frames. ], batch size: 34, lr: 1.20e-02
2024-08-06 10:55:08,096 INFO [trainer.py:765] (3/8) Epoch 10, batch 300, train_loss[loss=2.869, ArTop10Accuracy=0.757, over 14196.00 frames. ], tot_loss[loss=2.828, ArTop10Accuracy=0.7654, over 9370.30 frames. ], batch size: 44, lr: 1.19e-02
2024-08-06 10:56:41,182 INFO [trainer.py:765] (3/8) Epoch 10, batch 400, train_loss[loss=2.822, ArTop10Accuracy=0.7626, over 10914.00 frames. ], tot_loss[loss=2.824, ArTop10Accuracy=0.7661, over 10267.64 frames. ], batch size: 15, lr: 1.19e-02
2024-08-06 10:58:04,944 INFO [trainer.py:803] (3/8) Computing validation loss
2024-08-06 10:58:14,559 INFO [trainer.py:811] (3/8) Epoch 10, validation: loss=2.842, ArTop10Accuracy=0.7624, over 1827537.00 frames. 
2024-08-06 10:58:14,560 INFO [trainer.py:814] (3/8) Maximum memory allocated so far is 33011MB
2024-08-06 10:58:15,578 INFO [optim.py:386] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.035e+02 1.228e+02 1.320e+02 1.458e+02 6.096e+02, threshold=2.641e+02, percent-clipped=0.6
2024-08-06 10:58:15,585 INFO [trainer.py:765] (3/8) Epoch 10, batch 500, train_loss[loss=2.765, ArTop10Accuracy=0.7807, over 12336.00 frames. ], tot_loss[loss=2.819, ArTop10Accuracy=0.7668, over 10858.82 frames. ], batch size: 22, lr: 1.19e-02
2024-08-06 10:59:42,820 INFO [trainer.py:765] (3/8) Epoch 10, batch 600, train_loss[loss=2.878, ArTop10Accuracy=0.7531, over 11310.00 frames. ], tot_loss[loss=2.822, ArTop10Accuracy=0.7663, over 11375.27 frames. ], batch size: 18, lr: 1.18e-02
2024-08-06 11:01:18,113 INFO [trainer.py:765] (3/8) Epoch 10, batch 700, train_loss[loss=2.746, ArTop10Accuracy=0.7848, over 9285.00 frames. ], tot_loss[loss=2.828, ArTop10Accuracy=0.7654, over 11525.92 frames. ], batch size: 11, lr: 1.18e-02
2024-08-06 11:02:36,923 INFO [trainer.py:765] (3/8) Epoch 10, batch 800, train_loss[loss=2.666, ArTop10Accuracy=0.7947, over 10209.00 frames. ], tot_loss[loss=2.829, ArTop10Accuracy=0.7651, over 11652.28 frames. ], batch size: 12, lr: 1.17e-02
2024-08-06 11:03:51,217 INFO [trainer.py:765] (3/8) Epoch 10, batch 900, train_loss[loss=2.843, ArTop10Accuracy=0.764, over 12903.00 frames. ], tot_loss[loss=2.824, ArTop10Accuracy=0.766, over 11698.22 frames. ], batch size: 27, lr: 1.17e-02
2024-08-06 11:05:06,357 INFO [trainer.py:765] (3/8) Epoch 10, batch 1000, train_loss[loss=2.825, ArTop10Accuracy=0.7681, over 12927.00 frames. ], tot_loss[loss=2.831, ArTop10Accuracy=0.7648, over 11889.20 frames. ], batch size: 27, lr: 1.17e-02
2024-08-06 11:06:21,727 INFO [trainer.py:765] (3/8) Epoch 10, batch 1100, train_loss[loss=2.881, ArTop10Accuracy=0.7536, over 13725.00 frames. ], tot_loss[loss=2.837, ArTop10Accuracy=0.7633, over 11982.89 frames. ], batch size: 34, lr: 1.16e-02
2024-08-06 11:07:34,778 INFO [trainer.py:765] (3/8) Epoch 10, batch 1200, train_loss[loss=2.931, ArTop10Accuracy=0.7462, over 12060.00 frames. ], tot_loss[loss=2.836, ArTop10Accuracy=0.7637, over 11894.39 frames. ], batch size: 101, lr: 1.16e-02
2024-08-06 11:08:33,651 INFO [trainer.py:650] (3/8) Reaches end of dataloader.
2024-08-06 11:10:29,961 INFO [trainer.py:765] (3/8) Epoch 11, batch 100, train_loss[loss=2.872, ArTop10Accuracy=0.7564, over 14637.00 frames. ], tot_loss[loss=2.823, ArTop10Accuracy=0.7654, over 4739.55 frames. ], batch size: 63, lr: 1.10e-02
2024-08-06 11:12:04,681 INFO [trainer.py:765] (3/8) Epoch 11, batch 200, train_loss[loss=2.874, ArTop10Accuracy=0.755, over 13779.00 frames. ], tot_loss[loss=2.816, ArTop10Accuracy=0.767, over 7746.84 frames. ], batch size: 34, lr: 1.10e-02
2024-08-06 11:12:22,831 INFO [optim.py:386] (3/8) Clipping_scale=2.0, grad-norm quartiles 9.884e+01 1.240e+02 1.333e+02 1.457e+02 6.939e+02, threshold=2.667e+02, percent-clipped=0.1
2024-08-06 11:13:31,554 INFO [trainer.py:765] (3/8) Epoch 11, batch 300, train_loss[loss=2.886, ArTop10Accuracy=0.7523, over 14241.00 frames. ], tot_loss[loss=2.811, ArTop10Accuracy=0.7679, over 9372.27 frames. ], batch size: 44, lr: 1.09e-02
2024-08-06 11:15:03,275 INFO [trainer.py:765] (3/8) Epoch 11, batch 400, train_loss[loss=2.797, ArTop10Accuracy=0.7679, over 10716.00 frames. ], tot_loss[loss=2.809, ArTop10Accuracy=0.7686, over 10280.11 frames. ], batch size: 15, lr: 1.09e-02
2024-08-06 11:16:29,643 INFO [trainer.py:765] (3/8) Epoch 11, batch 500, train_loss[loss=2.834, ArTop10Accuracy=0.7655, over 12348.00 frames. ], tot_loss[loss=2.806, ArTop10Accuracy=0.7691, over 10847.00 frames. ], batch size: 22, lr: 1.09e-02
2024-08-06 11:18:00,524 INFO [trainer.py:765] (3/8) Epoch 11, batch 600, train_loss[loss=2.729, ArTop10Accuracy=0.7832, over 11418.00 frames. ], tot_loss[loss=2.809, ArTop10Accuracy=0.7687, over 11384.89 frames. ], batch size: 18, lr: 1.08e-02
2024-08-06 11:19:34,519 INFO [trainer.py:765] (3/8) Epoch 11, batch 700, train_loss[loss=2.69, ArTop10Accuracy=0.7945, over 10086.00 frames. ], tot_loss[loss=2.815, ArTop10Accuracy=0.7676, over 11518.99 frames. ], batch size: 12, lr: 1.08e-02
2024-08-06 11:20:55,489 INFO [trainer.py:765] (3/8) Epoch 11, batch 800, train_loss[loss=2.668, ArTop10Accuracy=0.7995, over 10152.00 frames. ], tot_loss[loss=2.815, ArTop10Accuracy=0.7676, over 11650.91 frames. ], batch size: 12, lr: 1.07e-02
2024-08-06 11:22:13,711 INFO [trainer.py:765] (3/8) Epoch 11, batch 900, train_loss[loss=2.8, ArTop10Accuracy=0.7695, over 12996.00 frames. ], tot_loss[loss=2.814, ArTop10Accuracy=0.7678, over 11706.55 frames. ], batch size: 27, lr: 1.07e-02
2024-08-06 11:23:31,804 INFO [trainer.py:765] (3/8) Epoch 11, batch 1000, train_loss[loss=2.759, ArTop10Accuracy=0.7778, over 12738.00 frames. ], tot_loss[loss=2.816, ArTop10Accuracy=0.7677, over 11899.04 frames. ], batch size: 27, lr: 1.07e-02
2024-08-06 11:24:46,908 INFO [trainer.py:765] (3/8) Epoch 11, batch 1100, train_loss[loss=2.76, ArTop10Accuracy=0.7812, over 13725.00 frames. ], tot_loss[loss=2.82, ArTop10Accuracy=0.7667, over 11968.01 frames. ], batch size: 34, lr: 1.06e-02
2024-08-06 11:26:00,739 INFO [trainer.py:765] (3/8) Epoch 11, batch 1200, train_loss[loss=2.928, ArTop10Accuracy=0.7442, over 12543.00 frames. ], tot_loss[loss=2.82, ArTop10Accuracy=0.7664, over 11889.44 frames. ], batch size: 101, lr: 1.06e-02
2024-08-06 11:26:15,853 INFO [trainer.py:803] (3/8) Computing validation loss
2024-08-06 11:26:25,556 INFO [trainer.py:811] (3/8) Epoch 11, validation: loss=2.831, ArTop10Accuracy=0.7643, over 1827537.00 frames. 
2024-08-06 11:26:25,556 INFO [trainer.py:814] (3/8) Maximum memory allocated so far is 33011MB
2024-08-06 11:26:26,191 INFO [optim.py:386] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.029e+02 1.251e+02 1.335e+02 1.441e+02 2.942e+02, threshold=2.669e+02, percent-clipped=0.1
2024-08-06 11:27:09,618 INFO [trainer.py:650] (3/8) Reaches end of dataloader.
2024-08-06 11:29:03,457 INFO [trainer.py:765] (3/8) Epoch 12, batch 100, train_loss[loss=2.846, ArTop10Accuracy=0.7631, over 14118.00 frames. ], tot_loss[loss=2.806, ArTop10Accuracy=0.7687, over 4773.56 frames. ], batch size: 62, lr: 1.01e-02
2024-08-06 11:30:30,680 INFO [trainer.py:765] (3/8) Epoch 12, batch 200, train_loss[loss=2.78, ArTop10Accuracy=0.7769, over 13701.00 frames. ], tot_loss[loss=2.799, ArTop10Accuracy=0.7701, over 7763.97 frames. ], batch size: 34, lr: 1.01e-02
2024-08-06 11:31:57,661 INFO [trainer.py:765] (3/8) Epoch 12, batch 300, train_loss[loss=2.809, ArTop10Accuracy=0.7704, over 14559.00 frames. ], tot_loss[loss=2.795, ArTop10Accuracy=0.7712, over 9355.86 frames. ], batch size: 45, lr: 1.01e-02
2024-08-06 11:33:30,744 INFO [trainer.py:765] (3/8) Epoch 12, batch 400, train_loss[loss=2.739, ArTop10Accuracy=0.7841, over 10419.00 frames. ], tot_loss[loss=2.791, ArTop10Accuracy=0.7719, over 10277.50 frames. ], batch size: 14, lr: 1.00e-02
2024-08-06 11:34:55,739 INFO [trainer.py:765] (3/8) Epoch 12, batch 500, train_loss[loss=2.723, ArTop10Accuracy=0.7865, over 12318.00 frames. ], tot_loss[loss=2.785, ArTop10Accuracy=0.7731, over 10829.72 frames. ], batch size: 22, lr: 1.00e-02
2024-08-06 11:36:29,367 INFO [trainer.py:765] (3/8) Epoch 12, batch 600, train_loss[loss=2.752, ArTop10Accuracy=0.7847, over 12060.00 frames. ], tot_loss[loss=2.787, ArTop10Accuracy=0.7728, over 11358.64 frames. ], batch size: 19, lr: 9.97e-03
2024-08-06 11:38:00,349 INFO [trainer.py:765] (3/8) Epoch 12, batch 700, train_loss[loss=2.783, ArTop10Accuracy=0.7784, over 10062.00 frames. ], tot_loss[loss=2.791, ArTop10Accuracy=0.772, over 11511.89 frames. ], batch size: 12, lr: 9.93e-03
2024-08-06 11:39:23,617 INFO [trainer.py:765] (3/8) Epoch 12, batch 800, train_loss[loss=2.752, ArTop10Accuracy=0.785, over 9558.00 frames. ], tot_loss[loss=2.797, ArTop10Accuracy=0.7708, over 11639.07 frames. ], batch size: 11, lr: 9.90e-03
2024-08-06 11:40:39,895 INFO [trainer.py:765] (3/8) Epoch 12, batch 900, train_loss[loss=2.831, ArTop10Accuracy=0.7677, over 12933.00 frames. ], tot_loss[loss=2.794, ArTop10Accuracy=0.7714, over 11694.97 frames. ], batch size: 27, lr: 9.87e-03
2024-08-06 11:41:14,001 INFO [optim.py:386] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.041e+02 1.248e+02 1.348e+02 1.459e+02 5.540e+02, threshold=2.695e+02, percent-clipped=0.3
2024-08-06 11:41:56,195 INFO [trainer.py:765] (3/8) Epoch 12, batch 1000, train_loss[loss=2.767, ArTop10Accuracy=0.7761, over 12909.00 frames. ], tot_loss[loss=2.797, ArTop10Accuracy=0.7708, over 11880.84 frames. ], batch size: 27, lr: 9.85e-03
2024-08-06 11:43:14,326 INFO [trainer.py:765] (3/8) Epoch 12, batch 1100, train_loss[loss=2.787, ArTop10Accuracy=0.7675, over 13797.00 frames. ], tot_loss[loss=2.802, ArTop10Accuracy=0.77, over 11937.72 frames. ], batch size: 34, lr: 9.82e-03
2024-08-06 11:44:26,162 INFO [trainer.py:765] (3/8) Epoch 12, batch 1200, train_loss[loss=2.935, ArTop10Accuracy=0.7424, over 12132.00 frames. ], tot_loss[loss=2.801, ArTop10Accuracy=0.7701, over 11867.60 frames. ], batch size: 101, lr: 9.79e-03
2024-08-06 11:45:26,537 INFO [trainer.py:650] (3/8) Reaches end of dataloader.
2024-08-06 11:47:26,606 INFO [trainer.py:765] (3/8) Epoch 13, batch 100, train_loss[loss=2.859, ArTop10Accuracy=0.7575, over 14892.00 frames. ], tot_loss[loss=2.794, ArTop10Accuracy=0.7707, over 4740.93 frames. ], batch size: 62, lr: 9.37e-03
2024-08-06 11:48:54,785 INFO [trainer.py:765] (3/8) Epoch 13, batch 200, train_loss[loss=2.754, ArTop10Accuracy=0.7835, over 14028.00 frames. ], tot_loss[loss=2.785, ArTop10Accuracy=0.7729, over 7741.31 frames. ], batch size: 35, lr: 9.34e-03
2024-08-06 11:50:20,521 INFO [trainer.py:765] (3/8) Epoch 13, batch 300, train_loss[loss=2.751, ArTop10Accuracy=0.7841, over 14265.00 frames. ], tot_loss[loss=2.774, ArTop10Accuracy=0.7751, over 9363.54 frames. ], batch size: 44, lr: 9.31e-03
2024-08-06 11:51:48,772 INFO [trainer.py:765] (3/8) Epoch 13, batch 400, train_loss[loss=2.735, ArTop10Accuracy=0.7855, over 10485.00 frames. ], tot_loss[loss=2.776, ArTop10Accuracy=0.775, over 10270.67 frames. ], batch size: 14, lr: 9.28e-03
2024-08-06 11:53:13,413 INFO [trainer.py:765] (3/8) Epoch 13, batch 500, train_loss[loss=2.73, ArTop10Accuracy=0.7846, over 12183.00 frames. ], tot_loss[loss=2.776, ArTop10Accuracy=0.7753, over 10846.70 frames. ], batch size: 22, lr: 9.26e-03
2024-08-06 11:54:52,229 INFO [trainer.py:765] (3/8) Epoch 13, batch 600, train_loss[loss=2.769, ArTop10Accuracy=0.7767, over 11760.00 frames. ], tot_loss[loss=2.775, ArTop10Accuracy=0.7753, over 11382.98 frames. ], batch size: 19, lr: 9.23e-03
2024-08-06 11:55:47,087 INFO [trainer.py:803] (3/8) Computing validation loss
2024-08-06 11:55:56,835 INFO [trainer.py:811] (3/8) Epoch 13, validation: loss=2.824, ArTop10Accuracy=0.7662, over 1827537.00 frames. 
2024-08-06 11:55:56,835 INFO [trainer.py:814] (3/8) Maximum memory allocated so far is 33011MB
2024-08-06 11:55:57,718 INFO [optim.py:386] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.064e+02 1.255e+02 1.343e+02 1.452e+02 4.888e+02, threshold=2.687e+02, percent-clipped=0.1
2024-08-06 11:56:28,472 INFO [trainer.py:765] (3/8) Epoch 13, batch 700, train_loss[loss=2.726, ArTop10Accuracy=0.7868, over 10173.00 frames. ], tot_loss[loss=2.779, ArTop10Accuracy=0.7747, over 11525.72 frames. ], batch size: 12, lr: 9.20e-03
2024-08-06 11:57:46,689 INFO [trainer.py:765] (3/8) Epoch 13, batch 800, train_loss[loss=2.736, ArTop10Accuracy=0.7815, over 10086.00 frames. ], tot_loss[loss=2.784, ArTop10Accuracy=0.7734, over 11612.96 frames. ], batch size: 12, lr: 9.18e-03
2024-08-06 11:59:03,292 INFO [trainer.py:765] (3/8) Epoch 13, batch 900, train_loss[loss=2.768, ArTop10Accuracy=0.7799, over 12903.00 frames. ], tot_loss[loss=2.782, ArTop10Accuracy=0.7737, over 11684.55 frames. ], batch size: 27, lr: 9.15e-03
2024-08-06 12:00:19,180 INFO [trainer.py:765] (3/8) Epoch 13, batch 1000, train_loss[loss=2.768, ArTop10Accuracy=0.7764, over 12711.00 frames. ], tot_loss[loss=2.786, ArTop10Accuracy=0.773, over 11887.42 frames. ], batch size: 27, lr: 9.13e-03
2024-08-06 12:01:34,887 INFO [trainer.py:765] (3/8) Epoch 13, batch 1100, train_loss[loss=2.819, ArTop10Accuracy=0.7671, over 13707.00 frames. ], tot_loss[loss=2.791, ArTop10Accuracy=0.772, over 11953.14 frames. ], batch size: 34, lr: 9.10e-03
2024-08-06 12:02:48,671 INFO [trainer.py:765] (3/8) Epoch 13, batch 1200, train_loss[loss=2.951, ArTop10Accuracy=0.7404, over 11715.00 frames. ], tot_loss[loss=2.793, ArTop10Accuracy=0.7714, over 11872.35 frames. ], batch size: 101, lr: 9.08e-03
2024-08-06 12:03:48,181 INFO [trainer.py:650] (3/8) Reaches end of dataloader.
2024-08-06 12:05:45,340 INFO [trainer.py:765] (3/8) Epoch 14, batch 100, train_loss[loss=2.837, ArTop10Accuracy=0.7671, over 14409.00 frames. ], tot_loss[loss=2.776, ArTop10Accuracy=0.7742, over 4764.27 frames. ], batch size: 62, lr: 8.71e-03
2024-08-06 12:07:16,610 INFO [trainer.py:765] (3/8) Epoch 14, batch 200, train_loss[loss=2.752, ArTop10Accuracy=0.7797, over 13509.00 frames. ], tot_loss[loss=2.767, ArTop10Accuracy=0.7761, over 7760.98 frames. ], batch size: 34, lr: 8.69e-03
2024-08-06 12:08:44,317 INFO [trainer.py:765] (3/8) Epoch 14, batch 300, train_loss[loss=2.763, ArTop10Accuracy=0.7764, over 14157.00 frames. ], tot_loss[loss=2.762, ArTop10Accuracy=0.7774, over 9387.83 frames. ], batch size: 44, lr: 8.66e-03
2024-08-06 12:10:01,137 INFO [optim.py:386] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.072e+02 1.266e+02 1.374e+02 1.483e+02 6.480e+02, threshold=2.748e+02, percent-clipped=0.2
2024-08-06 12:10:10,233 INFO [trainer.py:765] (3/8) Epoch 14, batch 400, train_loss[loss=2.688, ArTop10Accuracy=0.7928, over 10413.00 frames. ], tot_loss[loss=2.759, ArTop10Accuracy=0.7781, over 10296.32 frames. ], batch size: 14, lr: 8.64e-03
2024-08-06 12:11:36,157 INFO [trainer.py:765] (3/8) Epoch 14, batch 500, train_loss[loss=2.743, ArTop10Accuracy=0.7808, over 12153.00 frames. ], tot_loss[loss=2.759, ArTop10Accuracy=0.7781, over 10850.15 frames. ], batch size: 22, lr: 8.62e-03
2024-08-06 12:13:05,999 INFO [trainer.py:765] (3/8) Epoch 14, batch 600, train_loss[loss=2.798, ArTop10Accuracy=0.7717, over 11496.00 frames. ], tot_loss[loss=2.762, ArTop10Accuracy=0.7775, over 11365.20 frames. ], batch size: 18, lr: 8.59e-03
2024-08-06 12:14:38,559 INFO [trainer.py:765] (3/8) Epoch 14, batch 700, train_loss[loss=2.733, ArTop10Accuracy=0.7762, over 10191.00 frames. ], tot_loss[loss=2.767, ArTop10Accuracy=0.7765, over 11522.21 frames. ], batch size: 12, lr: 8.57e-03
2024-08-06 12:15:58,076 INFO [trainer.py:765] (3/8) Epoch 14, batch 800, train_loss[loss=2.694, ArTop10Accuracy=0.793, over 9315.00 frames. ], tot_loss[loss=2.769, ArTop10Accuracy=0.7762, over 11630.20 frames. ], batch size: 11, lr: 8.55e-03
2024-08-06 12:17:12,872 INFO [trainer.py:765] (3/8) Epoch 14, batch 900, train_loss[loss=2.755, ArTop10Accuracy=0.7847, over 12834.00 frames. ], tot_loss[loss=2.766, ArTop10Accuracy=0.7768, over 11689.30 frames. ], batch size: 27, lr: 8.52e-03
2024-08-06 12:18:29,620 INFO [trainer.py:765] (3/8) Epoch 14, batch 1000, train_loss[loss=2.773, ArTop10Accuracy=0.7776, over 12876.00 frames. ], tot_loss[loss=2.772, ArTop10Accuracy=0.7755, over 11875.98 frames. ], batch size: 27, lr: 8.50e-03
2024-08-06 12:19:45,382 INFO [trainer.py:765] (3/8) Epoch 14, batch 1100, train_loss[loss=2.798, ArTop10Accuracy=0.7736, over 13728.00 frames. ], tot_loss[loss=2.781, ArTop10Accuracy=0.7738, over 11970.38 frames. ], batch size: 34, lr: 8.48e-03
2024-08-06 12:20:59,285 INFO [trainer.py:765] (3/8) Epoch 14, batch 1200, train_loss[loss=2.908, ArTop10Accuracy=0.7524, over 12462.00 frames. ], tot_loss[loss=2.783, ArTop10Accuracy=0.7734, over 11864.36 frames. ], batch size: 101, lr: 8.46e-03
2024-08-06 12:21:58,393 INFO [trainer.py:650] (3/8) Reaches end of dataloader.
2024-08-06 12:23:51,968 INFO [trainer.py:765] (3/8) Epoch 15, batch 100, train_loss[loss=2.804, ArTop10Accuracy=0.7692, over 14889.00 frames. ], tot_loss[loss=2.761, ArTop10Accuracy=0.7771, over 4760.90 frames. ], batch size: 65, lr: 8.14e-03
2024-08-06 12:24:00,605 INFO [trainer.py:803] (3/8) Computing validation loss
2024-08-06 12:24:10,290 INFO [trainer.py:811] (3/8) Epoch 15, validation: loss=2.819, ArTop10Accuracy=0.7675, over 1827537.00 frames. 
2024-08-06 12:24:10,291 INFO [trainer.py:814] (3/8) Maximum memory allocated so far is 33011MB
2024-08-06 12:24:11,100 INFO [optim.py:386] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.080e+02 1.284e+02 1.371e+02 1.488e+02 4.667e+02, threshold=2.743e+02, percent-clipped=0.2
2024-08-06 12:25:29,995 INFO [trainer.py:765] (3/8) Epoch 15, batch 200, train_loss[loss=2.662, ArTop10Accuracy=0.7959, over 13275.00 frames. ], tot_loss[loss=2.752, ArTop10Accuracy=0.7792, over 7758.27 frames. ], batch size: 34, lr: 8.12e-03
2024-08-06 12:26:58,702 INFO [trainer.py:765] (3/8) Epoch 15, batch 300, train_loss[loss=2.777, ArTop10Accuracy=0.772, over 13758.00 frames. ], tot_loss[loss=2.75, ArTop10Accuracy=0.7796, over 9380.39 frames. ], batch size: 44, lr: 8.09e-03
2024-08-06 12:28:28,540 INFO [trainer.py:765] (3/8) Epoch 15, batch 400, train_loss[loss=2.664, ArTop10Accuracy=0.7956, over 10533.00 frames. ], tot_loss[loss=2.75, ArTop10Accuracy=0.7797, over 10288.58 frames. ], batch size: 14, lr: 8.07e-03
2024-08-06 12:29:54,038 INFO [trainer.py:765] (3/8) Epoch 15, batch 500, train_loss[loss=2.802, ArTop10Accuracy=0.7713, over 12066.00 frames. ], tot_loss[loss=2.749, ArTop10Accuracy=0.7799, over 10860.11 frames. ], batch size: 22, lr: 8.05e-03
2024-08-06 12:31:23,299 INFO [trainer.py:765] (3/8) Epoch 15, batch 600, train_loss[loss=2.704, ArTop10Accuracy=0.7892, over 11985.00 frames. ], tot_loss[loss=2.751, ArTop10Accuracy=0.7796, over 11365.41 frames. ], batch size: 19, lr: 8.03e-03
2024-08-06 12:32:53,182 INFO [trainer.py:765] (3/8) Epoch 15, batch 700, train_loss[loss=2.61, ArTop10Accuracy=0.8052, over 10227.00 frames. ], tot_loss[loss=2.756, ArTop10Accuracy=0.7784, over 11530.22 frames. ], batch size: 12, lr: 8.01e-03
2024-08-06 12:34:18,261 INFO [trainer.py:765] (3/8) Epoch 15, batch 800, train_loss[loss=2.733, ArTop10Accuracy=0.7805, over 10134.00 frames. ], tot_loss[loss=2.762, ArTop10Accuracy=0.7775, over 11637.96 frames. ], batch size: 12, lr: 7.99e-03
2024-08-06 12:35:34,733 INFO [trainer.py:765] (3/8) Epoch 15, batch 900, train_loss[loss=2.748, ArTop10Accuracy=0.779, over 12927.00 frames. ], tot_loss[loss=2.757, ArTop10Accuracy=0.7785, over 11705.92 frames. ], batch size: 27, lr: 7.97e-03
2024-08-06 12:36:50,547 INFO [trainer.py:765] (3/8) Epoch 15, batch 1000, train_loss[loss=2.757, ArTop10Accuracy=0.7794, over 12849.00 frames. ], tot_loss[loss=2.762, ArTop10Accuracy=0.7777, over 11893.13 frames. ], batch size: 27, lr: 7.95e-03
2024-08-06 12:38:05,186 INFO [trainer.py:765] (3/8) Epoch 15, batch 1100, train_loss[loss=2.801, ArTop10Accuracy=0.7653, over 13638.00 frames. ], tot_loss[loss=2.767, ArTop10Accuracy=0.7765, over 11952.94 frames. ], batch size: 34, lr: 7.93e-03
2024-08-06 12:38:12,847 INFO [optim.py:386] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.080e+02 1.293e+02 1.379e+02 1.467e+02 2.824e+02, threshold=2.759e+02, percent-clipped=0.1
2024-08-06 12:39:18,795 INFO [trainer.py:765] (3/8) Epoch 15, batch 1200, train_loss[loss=2.927, ArTop10Accuracy=0.7428, over 12492.00 frames. ], tot_loss[loss=2.768, ArTop10Accuracy=0.7763, over 11874.98 frames. ], batch size: 101, lr: 7.91e-03
2024-08-06 12:40:18,968 INFO [trainer.py:650] (3/8) Reaches end of dataloader.
2024-08-06 12:42:17,625 INFO [trainer.py:765] (3/8) Epoch 16, batch 100, train_loss[loss=2.855, ArTop10Accuracy=0.7584, over 14046.00 frames. ], tot_loss[loss=2.755, ArTop10Accuracy=0.7784, over 4773.43 frames. ], batch size: 62, lr: 7.63e-03
2024-08-06 12:43:49,570 INFO [trainer.py:765] (3/8) Epoch 16, batch 200, train_loss[loss=2.744, ArTop10Accuracy=0.7782, over 13509.00 frames. ], tot_loss[loss=2.741, ArTop10Accuracy=0.7811, over 7745.98 frames. ], batch size: 34, lr: 7.61e-03
2024-08-06 12:45:18,508 INFO [trainer.py:765] (3/8) Epoch 16, batch 300, train_loss[loss=2.82, ArTop10Accuracy=0.7672, over 14163.00 frames. ], tot_loss[loss=2.733, ArTop10Accuracy=0.7828, over 9367.42 frames. ], batch size: 44, lr: 7.59e-03
2024-08-06 12:46:45,215 INFO [trainer.py:765] (3/8) Epoch 16, batch 400, train_loss[loss=2.704, ArTop10Accuracy=0.7902, over 10095.00 frames. ], tot_loss[loss=2.738, ArTop10Accuracy=0.7819, over 10274.41 frames. ], batch size: 14, lr: 7.58e-03
2024-08-06 12:48:16,317 INFO [trainer.py:765] (3/8) Epoch 16, batch 500, train_loss[loss=2.739, ArTop10Accuracy=0.7879, over 12132.00 frames. ], tot_loss[loss=2.735, ArTop10Accuracy=0.7824, over 10839.76 frames. ], batch size: 22, lr: 7.56e-03
2024-08-06 12:49:46,648 INFO [trainer.py:765] (3/8) Epoch 16, batch 600, train_loss[loss=2.728, ArTop10Accuracy=0.7856, over 11466.00 frames. ], tot_loss[loss=2.74, ArTop10Accuracy=0.7816, over 11376.36 frames. ], batch size: 18, lr: 7.54e-03
2024-08-06 12:51:23,687 INFO [trainer.py:765] (3/8) Epoch 16, batch 700, train_loss[loss=2.741, ArTop10Accuracy=0.7755, over 10176.00 frames. ], tot_loss[loss=2.743, ArTop10Accuracy=0.7809, over 11521.68 frames. ], batch size: 12, lr: 7.52e-03
2024-08-06 12:52:43,507 INFO [trainer.py:765] (3/8) Epoch 16, batch 800, train_loss[loss=2.648, ArTop10Accuracy=0.8006, over 9210.00 frames. ], tot_loss[loss=2.748, ArTop10Accuracy=0.7801, over 11645.23 frames. ], batch size: 11, lr: 7.51e-03
2024-08-06 12:53:06,022 INFO [trainer.py:803] (3/8) Computing validation loss
2024-08-06 12:53:15,496 INFO [trainer.py:811] (3/8) Epoch 16, validation: loss=2.816, ArTop10Accuracy=0.7678, over 1827537.00 frames. 
2024-08-06 12:53:15,497 INFO [trainer.py:814] (3/8) Maximum memory allocated so far is 33011MB
2024-08-06 12:53:16,192 INFO [optim.py:386] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.112e+02 1.291e+02 1.391e+02 1.487e+02 3.459e+02, threshold=2.783e+02, percent-clipped=0.1
2024-08-06 12:54:06,486 INFO [trainer.py:765] (3/8) Epoch 16, batch 900, train_loss[loss=2.737, ArTop10Accuracy=0.7819, over 12945.00 frames. ], tot_loss[loss=2.743, ArTop10Accuracy=0.781, over 11677.13 frames. ], batch size: 27, lr: 7.49e-03
2024-08-06 12:55:19,797 INFO [trainer.py:765] (3/8) Epoch 16, batch 1000, train_loss[loss=2.689, ArTop10Accuracy=0.7908, over 13293.00 frames. ], tot_loss[loss=2.748, ArTop10Accuracy=0.78, over 11861.96 frames. ], batch size: 28, lr: 7.47e-03
2024-08-06 12:56:33,168 INFO [trainer.py:765] (3/8) Epoch 16, batch 1100, train_loss[loss=2.79, ArTop10Accuracy=0.7754, over 13731.00 frames. ], tot_loss[loss=2.757, ArTop10Accuracy=0.7783, over 11946.49 frames. ], batch size: 34, lr: 7.45e-03
2024-08-06 12:57:48,491 INFO [trainer.py:765] (3/8) Epoch 16, batch 1200, train_loss[loss=2.893, ArTop10Accuracy=0.7484, over 12243.00 frames. ], tot_loss[loss=2.757, ArTop10Accuracy=0.7782, over 11839.50 frames. ], batch size: 101, lr: 7.44e-03
2024-08-06 12:58:48,504 INFO [trainer.py:650] (3/8) Reaches end of dataloader.
2024-08-06 13:00:47,906 INFO [trainer.py:765] (3/8) Epoch 17, batch 100, train_loss[loss=2.789, ArTop10Accuracy=0.7714, over 14616.00 frames. ], tot_loss[loss=2.74, ArTop10Accuracy=0.7809, over 4765.07 frames. ], batch size: 63, lr: 7.18e-03
2024-08-06 13:02:19,309 INFO [trainer.py:765] (3/8) Epoch 17, batch 200, train_loss[loss=2.732, ArTop10Accuracy=0.7834, over 13602.00 frames. ], tot_loss[loss=2.733, ArTop10Accuracy=0.7823, over 7748.91 frames. ], batch size: 34, lr: 7.17e-03
2024-08-06 13:03:45,523 INFO [trainer.py:765] (3/8) Epoch 17, batch 300, train_loss[loss=2.795, ArTop10Accuracy=0.7725, over 13977.00 frames. ], tot_loss[loss=2.728, ArTop10Accuracy=0.7836, over 9359.46 frames. ], batch size: 44, lr: 7.15e-03
2024-08-06 13:05:21,766 INFO [trainer.py:765] (3/8) Epoch 17, batch 400, train_loss[loss=2.692, ArTop10Accuracy=0.7914, over 10851.00 frames. ], tot_loss[loss=2.729, ArTop10Accuracy=0.7836, over 10278.48 frames. ], batch size: 15, lr: 7.14e-03
2024-08-06 13:06:47,027 INFO [trainer.py:765] (3/8) Epoch 17, batch 500, train_loss[loss=2.7, ArTop10Accuracy=0.7846, over 12618.00 frames. ], tot_loss[loss=2.723, ArTop10Accuracy=0.7847, over 10854.47 frames. ], batch size: 23, lr: 7.12e-03
2024-08-06 13:07:39,884 INFO [optim.py:386] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.140e+02 1.293e+02 1.386e+02 1.488e+02 3.253e+02, threshold=2.772e+02, percent-clipped=0.1
2024-08-06 13:08:22,694 INFO [trainer.py:765] (3/8) Epoch 17, batch 600, train_loss[loss=2.655, ArTop10Accuracy=0.8002, over 11547.00 frames. ], tot_loss[loss=2.726, ArTop10Accuracy=0.784, over 11386.03 frames. ], batch size: 18, lr: 7.10e-03
2024-08-06 13:09:54,842 INFO [trainer.py:765] (3/8) Epoch 17, batch 700, train_loss[loss=2.522, ArTop10Accuracy=0.8214, over 10182.00 frames. ], tot_loss[loss=2.728, ArTop10Accuracy=0.7836, over 11523.38 frames. ], batch size: 12, lr: 7.09e-03
2024-08-06 13:11:19,487 INFO [trainer.py:765] (3/8) Epoch 17, batch 800, train_loss[loss=2.54, ArTop10Accuracy=0.8153, over 10068.00 frames. ], tot_loss[loss=2.732, ArTop10Accuracy=0.7831, over 11645.95 frames. ], batch size: 12, lr: 7.07e-03
2024-08-06 13:12:35,676 INFO [trainer.py:765] (3/8) Epoch 17, batch 900, train_loss[loss=2.767, ArTop10Accuracy=0.7784, over 12936.00 frames. ], tot_loss[loss=2.733, ArTop10Accuracy=0.7831, over 11694.16 frames. ], batch size: 27, lr: 7.06e-03
2024-08-06 13:13:53,067 INFO [trainer.py:765] (3/8) Epoch 17, batch 1000, train_loss[loss=2.722, ArTop10Accuracy=0.7889, over 12888.00 frames. ], tot_loss[loss=2.739, ArTop10Accuracy=0.782, over 11891.73 frames. ], batch size: 27, lr: 7.04e-03
2024-08-06 13:15:08,490 INFO [trainer.py:765] (3/8) Epoch 17, batch 1100, train_loss[loss=2.697, ArTop10Accuracy=0.7897, over 13773.00 frames. ], tot_loss[loss=2.744, ArTop10Accuracy=0.7811, over 11951.73 frames. ], batch size: 34, lr: 7.02e-03
2024-08-06 13:16:22,395 INFO [trainer.py:765] (3/8) Epoch 17, batch 1200, train_loss[loss=2.828, ArTop10Accuracy=0.7645, over 12483.00 frames. ], tot_loss[loss=2.746, ArTop10Accuracy=0.7804, over 11856.50 frames. ], batch size: 101, lr: 7.01e-03
2024-08-06 13:17:21,345 INFO [trainer.py:650] (3/8) Reaches end of dataloader.
2024-08-06 13:19:16,000 INFO [trainer.py:765] (3/8) Epoch 18, batch 100, train_loss[loss=2.748, ArTop10Accuracy=0.7797, over 14040.00 frames. ], tot_loss[loss=2.732, ArTop10Accuracy=0.7818, over 4757.97 frames. ], batch size: 62, lr: 6.78e-03
2024-08-06 13:20:46,607 INFO [trainer.py:765] (3/8) Epoch 18, batch 200, train_loss[loss=2.765, ArTop10Accuracy=0.7724, over 13836.00 frames. ], tot_loss[loss=2.728, ArTop10Accuracy=0.7828, over 7748.93 frames. ], batch size: 34, lr: 6.77e-03
2024-08-06 13:21:55,111 INFO [trainer.py:803] (3/8) Computing validation loss
2024-08-06 13:22:04,751 INFO [trainer.py:811] (3/8) Epoch 18, validation: loss=2.817, ArTop10Accuracy=0.768, over 1827537.00 frames. 
2024-08-06 13:22:04,752 INFO [trainer.py:814] (3/8) Maximum memory allocated so far is 33011MB
2024-08-06 13:22:05,480 INFO [optim.py:386] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.131e+02 1.323e+02 1.409e+02 1.514e+02 3.209e+02, threshold=2.818e+02, percent-clipped=0.1
2024-08-06 13:22:26,587 INFO [trainer.py:765] (3/8) Epoch 18, batch 300, train_loss[loss=2.719, ArTop10Accuracy=0.7826, over 14019.00 frames. ], tot_loss[loss=2.719, ArTop10Accuracy=0.7851, over 9382.02 frames. ], batch size: 44, lr: 6.76e-03
2024-08-06 13:23:57,935 INFO [trainer.py:765] (3/8) Epoch 18, batch 400, train_loss[loss=2.703, ArTop10Accuracy=0.7932, over 10263.00 frames. ], tot_loss[loss=2.714, ArTop10Accuracy=0.7862, over 10301.86 frames. ], batch size: 14, lr: 6.74e-03
2024-08-06 13:25:34,019 INFO [trainer.py:765] (3/8) Epoch 18, batch 500, train_loss[loss=2.681, ArTop10Accuracy=0.7939, over 12159.00 frames. ], tot_loss[loss=2.716, ArTop10Accuracy=0.7859, over 10841.49 frames. ], batch size: 22, lr: 6.73e-03
2024-08-06 13:27:00,639 INFO [trainer.py:765] (3/8) Epoch 18, batch 600, train_loss[loss=2.633, ArTop10Accuracy=0.8064, over 11547.00 frames. ], tot_loss[loss=2.721, ArTop10Accuracy=0.785, over 11359.83 frames. ], batch size: 18, lr: 6.71e-03
2024-08-06 13:28:33,588 INFO [trainer.py:765] (3/8) Epoch 18, batch 700, train_loss[loss=2.561, ArTop10Accuracy=0.814, over 10119.00 frames. ], tot_loss[loss=2.721, ArTop10Accuracy=0.7849, over 11503.18 frames. ], batch size: 12, lr: 6.70e-03
2024-08-06 13:29:54,991 INFO [trainer.py:765] (3/8) Epoch 18, batch 800, train_loss[loss=2.603, ArTop10Accuracy=0.8063, over 10134.00 frames. ], tot_loss[loss=2.725, ArTop10Accuracy=0.7842, over 11650.29 frames. ], batch size: 12, lr: 6.68e-03
2024-08-06 13:31:12,525 INFO [trainer.py:765] (3/8) Epoch 18, batch 900, train_loss[loss=2.859, ArTop10Accuracy=0.7596, over 12912.00 frames. ], tot_loss[loss=2.721, ArTop10Accuracy=0.7852, over 11680.26 frames. ], batch size: 27, lr: 6.67e-03
2024-08-06 13:32:26,557 INFO [trainer.py:765] (3/8) Epoch 18, batch 1000, train_loss[loss=2.705, ArTop10Accuracy=0.7892, over 13044.00 frames. ], tot_loss[loss=2.724, ArTop10Accuracy=0.7844, over 11875.32 frames. ], batch size: 27, lr: 6.66e-03
2024-08-06 13:33:41,503 INFO [trainer.py:765] (3/8) Epoch 18, batch 1100, train_loss[loss=2.799, ArTop10Accuracy=0.7727, over 13740.00 frames. ], tot_loss[loss=2.733, ArTop10Accuracy=0.7828, over 11949.67 frames. ], batch size: 34, lr: 6.64e-03
2024-08-06 13:34:54,679 INFO [trainer.py:765] (3/8) Epoch 18, batch 1200, train_loss[loss=2.883, ArTop10Accuracy=0.7567, over 12735.00 frames. ], tot_loss[loss=2.738, ArTop10Accuracy=0.7818, over 11873.72 frames. ], batch size: 101, lr: 6.63e-03
2024-08-06 13:35:51,070 INFO [optim.py:386] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.124e+02 1.340e+02 1.433e+02 1.533e+02 2.444e+02, threshold=2.867e+02, percent-clipped=0.0
2024-08-06 13:35:54,614 INFO [trainer.py:650] (3/8) Reaches end of dataloader.
2024-08-06 13:37:48,631 INFO [trainer.py:765] (3/8) Epoch 19, batch 100, train_loss[loss=2.814, ArTop10Accuracy=0.7616, over 14505.00 frames. ], tot_loss[loss=2.714, ArTop10Accuracy=0.786, over 4772.08 frames. ], batch size: 62, lr: 6.43e-03
2024-08-06 13:39:23,263 INFO [trainer.py:765] (3/8) Epoch 19, batch 200, train_loss[loss=2.737, ArTop10Accuracy=0.7761, over 13692.00 frames. ], tot_loss[loss=2.714, ArTop10Accuracy=0.786, over 7752.10 frames. ], batch size: 34, lr: 6.41e-03
2024-08-06 13:40:48,365 INFO [trainer.py:765] (3/8) Epoch 19, batch 300, train_loss[loss=2.774, ArTop10Accuracy=0.7773, over 13896.00 frames. ], tot_loss[loss=2.71, ArTop10Accuracy=0.7867, over 9365.27 frames. ], batch size: 44, lr: 6.40e-03
2024-08-06 13:42:21,074 INFO [trainer.py:765] (3/8) Epoch 19, batch 400, train_loss[loss=2.689, ArTop10Accuracy=0.7877, over 10443.00 frames. ], tot_loss[loss=2.705, ArTop10Accuracy=0.7878, over 10266.11 frames. ], batch size: 14, lr: 6.39e-03
2024-08-06 13:43:44,962 INFO [trainer.py:765] (3/8) Epoch 19, batch 500, train_loss[loss=2.773, ArTop10Accuracy=0.7813, over 12192.00 frames. ], tot_loss[loss=2.707, ArTop10Accuracy=0.7875, over 10846.68 frames. ], batch size: 22, lr: 6.37e-03
2024-08-06 13:45:16,688 INFO [trainer.py:765] (3/8) Epoch 19, batch 600, train_loss[loss=2.611, ArTop10Accuracy=0.8074, over 11343.00 frames. ], tot_loss[loss=2.71, ArTop10Accuracy=0.7871, over 11365.54 frames. ], batch size: 18, lr: 6.36e-03
2024-08-06 13:46:48,330 INFO [trainer.py:765] (3/8) Epoch 19, batch 700, train_loss[loss=2.617, ArTop10Accuracy=0.8064, over 9999.00 frames. ], tot_loss[loss=2.714, ArTop10Accuracy=0.7864, over 11505.08 frames. ], batch size: 12, lr: 6.35e-03
2024-08-06 13:48:11,890 INFO [trainer.py:765] (3/8) Epoch 19, batch 800, train_loss[loss=2.594, ArTop10Accuracy=0.8097, over 9999.00 frames. ], tot_loss[loss=2.72, ArTop10Accuracy=0.7854, over 11627.45 frames. ], batch size: 12, lr: 6.34e-03
2024-08-06 13:49:27,265 INFO [trainer.py:765] (3/8) Epoch 19, batch 900, train_loss[loss=2.669, ArTop10Accuracy=0.7927, over 13359.00 frames. ], tot_loss[loss=2.714, ArTop10Accuracy=0.7863, over 11681.49 frames. ], batch size: 28, lr: 6.32e-03
2024-08-06 13:50:40,660 INFO [trainer.py:803] (3/8) Computing validation loss
2024-08-06 13:50:50,537 INFO [trainer.py:811] (3/8) Epoch 19, validation: loss=2.818, ArTop10Accuracy=0.7679, over 1827537.00 frames. 
2024-08-06 13:50:50,537 INFO [trainer.py:814] (3/8) Maximum memory allocated so far is 33011MB
2024-08-06 13:50:51,496 INFO [optim.py:386] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.161e+02 1.371e+02 1.455e+02 1.550e+02 3.697e+02, threshold=2.909e+02, percent-clipped=0.2
2024-08-06 13:50:52,922 INFO [trainer.py:765] (3/8) Epoch 19, batch 1000, train_loss[loss=2.747, ArTop10Accuracy=0.782, over 12804.00 frames. ], tot_loss[loss=2.721, ArTop10Accuracy=0.7851, over 11885.30 frames. ], batch size: 27, lr: 6.31e-03
2024-08-06 13:52:08,271 INFO [trainer.py:765] (3/8) Epoch 19, batch 1100, train_loss[loss=2.752, ArTop10Accuracy=0.7856, over 13539.00 frames. ], tot_loss[loss=2.728, ArTop10Accuracy=0.7839, over 11950.86 frames. ], batch size: 34, lr: 6.30e-03
2024-08-06 13:53:22,318 INFO [trainer.py:765] (3/8) Epoch 19, batch 1200, train_loss[loss=2.822, ArTop10Accuracy=0.7666, over 12456.00 frames. ], tot_loss[loss=2.728, ArTop10Accuracy=0.7836, over 11872.46 frames. ], batch size: 101, lr: 6.28e-03
2024-08-06 13:54:21,909 INFO [trainer.py:650] (3/8) Reaches end of dataloader.
2024-08-06 13:56:12,911 INFO [trainer.py:765] (3/8) Epoch 20, batch 100, train_loss[loss=2.75, ArTop10Accuracy=0.7818, over 14421.00 frames. ], tot_loss[loss=2.709, ArTop10Accuracy=0.7869, over 4732.71 frames. ], batch size: 62, lr: 6.10e-03
2024-08-06 13:57:42,502 INFO [trainer.py:765] (3/8) Epoch 20, batch 200, train_loss[loss=2.767, ArTop10Accuracy=0.7772, over 13650.00 frames. ], tot_loss[loss=2.704, ArTop10Accuracy=0.788, over 7747.52 frames. ], batch size: 34, lr: 6.09e-03
2024-08-06 13:59:15,437 INFO [trainer.py:765] (3/8) Epoch 20, batch 300, train_loss[loss=2.739, ArTop10Accuracy=0.7793, over 14226.00 frames. ], tot_loss[loss=2.697, ArTop10Accuracy=0.7894, over 9362.38 frames. ], batch size: 44, lr: 6.08e-03
2024-08-06 14:00:44,363 INFO [trainer.py:765] (3/8) Epoch 20, batch 400, train_loss[loss=2.645, ArTop10Accuracy=0.7971, over 10269.00 frames. ], tot_loss[loss=2.695, ArTop10Accuracy=0.7897, over 10275.36 frames. ], batch size: 14, lr: 6.07e-03
2024-08-06 14:02:14,860 INFO [trainer.py:765] (3/8) Epoch 20, batch 500, train_loss[loss=2.717, ArTop10Accuracy=0.786, over 12165.00 frames. ], tot_loss[loss=2.693, ArTop10Accuracy=0.7903, over 10821.64 frames. ], batch size: 22, lr: 6.06e-03
2024-08-06 14:03:40,861 INFO [trainer.py:765] (3/8) Epoch 20, batch 600, train_loss[loss=2.75, ArTop10Accuracy=0.78, over 11436.00 frames. ], tot_loss[loss=2.695, ArTop10Accuracy=0.7902, over 11359.02 frames. ], batch size: 18, lr: 6.04e-03
2024-08-06 14:05:13,871 INFO [trainer.py:765] (3/8) Epoch 20, batch 700, train_loss[loss=2.699, ArTop10Accuracy=0.7882, over 9402.00 frames. ], tot_loss[loss=2.702, ArTop10Accuracy=0.7887, over 11513.76 frames. ], batch size: 11, lr: 6.03e-03
2024-08-06 14:05:30,797 INFO [optim.py:386] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.180e+02 1.365e+02 1.456e+02 1.550e+02 3.525e+02, threshold=2.913e+02, percent-clipped=0.1
2024-08-06 14:06:34,515 INFO [trainer.py:765] (3/8) Epoch 20, batch 800, train_loss[loss=2.575, ArTop10Accuracy=0.8192, over 10212.00 frames. ], tot_loss[loss=2.706, ArTop10Accuracy=0.7877, over 11662.80 frames. ], batch size: 12, lr: 6.02e-03
2024-08-06 14:07:50,950 INFO [trainer.py:765] (3/8) Epoch 20, batch 900, train_loss[loss=2.687, ArTop10Accuracy=0.7923, over 12840.00 frames. ], tot_loss[loss=2.703, ArTop10Accuracy=0.7885, over 11704.16 frames. ], batch size: 27, lr: 6.01e-03
2024-08-06 14:09:07,179 INFO [trainer.py:765] (3/8) Epoch 20, batch 1000, train_loss[loss=2.767, ArTop10Accuracy=0.776, over 13011.00 frames. ], tot_loss[loss=2.705, ArTop10Accuracy=0.7881, over 11884.06 frames. ], batch size: 27, lr: 6.00e-03
2024-08-06 14:10:21,215 INFO [trainer.py:765] (3/8) Epoch 20, batch 1100, train_loss[loss=2.713, ArTop10Accuracy=0.7884, over 13860.00 frames. ], tot_loss[loss=2.712, ArTop10Accuracy=0.7867, over 11958.77 frames. ], batch size: 34, lr: 5.99e-03
2024-08-06 14:11:37,819 INFO [trainer.py:765] (3/8) Epoch 20, batch 1200, train_loss[loss=2.817, ArTop10Accuracy=0.7719, over 11877.00 frames. ], tot_loss[loss=2.713, ArTop10Accuracy=0.7866, over 11870.45 frames. ], batch size: 101, lr: 5.98e-03
2024-08-06 14:12:36,807 INFO [trainer.py:650] (3/8) Reaches end of dataloader.
2024-08-06 14:12:36,810 INFO [trainer.py:1069] (3/8) Done!